WO2021051877A1 - Method for obtaining input text in artificial intelligence interview, and related apparatus - Google Patents

Method for obtaining input text in artificial intelligence interview, and related apparatus Download PDF

Info

Publication number
WO2021051877A1
WO2021051877A1 PCT/CN2020/093597 CN2020093597W WO2021051877A1 WO 2021051877 A1 WO2021051877 A1 WO 2021051877A1 CN 2020093597 W CN2020093597 W CN 2020093597W WO 2021051877 A1 WO2021051877 A1 WO 2021051877A1
Authority
WO
WIPO (PCT)
Prior art keywords
word
mutual information
information value
word segmentation
homophone
Prior art date
Application number
PCT/CN2020/093597
Other languages
French (fr)
Chinese (zh)
Inventor
郑立颖
徐亮
阮晓雯
Original Assignee
平安科技（深圳）有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 平安科技（深圳）有限公司 filed Critical 平安科技（深圳）有限公司
Publication of WO2021051877A1 publication Critical patent/WO2021051877A1/en

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G06F16/332—Query formulation
    • G06F16/3329—Natural language query formulation or dialogue systems
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G06F16/3331—Query processing
    • G06F16/334—Query execution
    • G06F16/3343—Query execution using phonetics
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G06F16/3331—Query processing
    • G06F16/334—Query execution
    • G06F16/3344—Query execution using natural language analysis
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management

Definitions

  • This application relates to the field of artificial intelligence technology, and in particular to a method, device, equipment, and computer-readable storage medium for obtaining input text in artificial intelligence interviews.
  • the intelligent interview is that the artificial intelligence interviewer replaces the traditional interviewer to interview the candidate.
  • the artificial intelligence interviewer integrates voice recognition, facial recognition and other functions, can comprehensively evaluate the interview performance of the candidate, and pass the corresponding candidate Perform rankings to determine ideal candidates.
  • this application provides a method, device, equipment, and computer-readable storage medium for obtaining input text in an artificial intelligence interview.
  • a method for obtaining input text in an artificial intelligence interview including: in the artificial intelligence interview, calling a preset general word database and a general word pronunciation comparison table to correct the input voice, and perform voice on the corrected voice Recognize and obtain the recognized text; obtain a word segmentation set by performing word segmentation processing on the recognized text; for the word segmentation in the word segmentation set, respectively calculate the mutual information value of the word segmentation relative to the left and right participles, and according to the obtained mutual information
  • the value locates the homophone error words in the recognized text, the mutual information value includes the left mutual information value and the right mutual information value; the target word is extracted from the preset homophone word database to replace the homophone error words to obtain all the homophone error words.
  • the pronunciation of the target word is the same as the homophone wrong word.
  • a device for acquiring input text in an artificial intelligence interview including: a device for acquiring input text in an artificial intelligence interview, wherein the device includes: a speech recognition module, used to call pre-programs in the artificial intelligence interview.
  • the universal word database and the universal word pronunciation comparison table are set to correct the input speech, and perform speech recognition on the corrected speech to obtain the recognized text;
  • the word segmentation processing module is used to obtain the word segmentation set by performing word segmentation processing on the recognized text; homophone
  • the error word positioning module is used to calculate the mutual information value of the word segmentation relative to the left and right participles for the word segmentation in the word segmentation set, and locate the homophone error words in the recognized text according to the obtained mutual information value ,
  • the mutual information value includes a left mutual information value and a right mutual information value;
  • the input text acquisition module is used to extract the target word from the preset homophone word database to replace the homophone error word, and obtain the intelligent interview In the input text of, the pronunciation of the target word is the same as the
  • a device for acquiring input text in an artificial intelligence interview comprising a processor and a memory, and computer-readable instructions are stored on the memory.
  • the computer-readable instructions are executed by the processor, the implementation of The method of obtaining input text in the artificial intelligence interview described in the item.
  • a computer-readable storage medium has a computer program stored thereon, and when the computer program is executed by a processor, the method for obtaining input text in an artificial intelligence interview as described in any one of the preceding items is realized.
  • the input voice in the artificial intelligence interview is corrected by calling the preset general word database and the general word pronunciation comparison table, which can accurately identify the general vocabulary in the interview question by the applicant, thereby correcting Correct the speech for speech recognition to obtain accurate recognized text, and then perform word segmentation on the recognized text.
  • After obtaining the word segmentation set calculate the mutual information value of the relative left participle and the right participle by dividing the words in the word segmentation set to calculate the mutual information value according to the obtained mutual information Value positioning recognizes homophone wrong words in the text, and finally replaces homophone wrong words by extracting the target word from the preset homophone word database to obtain the input text in the smart interview.
  • this application can not only accurately identify the general vocabulary in the interview field, but also correct the homophonic wrong words in the recognized text.
  • the input text obtained is largely close to the applicant’s real expression, making the artificial intelligence interview The officer can accurately obtain the answer content of the applicant, so that the current intelligent interview can be carried out effectively.
  • Fig. 1 is a schematic diagram showing an implementation environment involved in this application according to an exemplary embodiment.
  • Fig. 2 is a hardware block diagram showing a server according to an exemplary embodiment.
  • Fig. 3 is a flow chart showing a method for obtaining input text in an artificial intelligence interview according to an exemplary embodiment.
  • Fig. 4 is a flowchart showing a method for obtaining input text in an artificial intelligence interview according to another exemplary embodiment.
  • FIG. 5 is a flowchart of step 350 shown in FIG. 3 in an embodiment.
  • FIG. 6 is a flowchart of step 350 shown in FIG. 3 in another embodiment.
  • FIG. 7 is a flowchart of step 370 shown in FIG. 3 in an embodiment.
  • Fig. 8 is a block diagram of a device for acquiring input text in an artificial intelligence interview according to an exemplary embodiment.
  • Fig. 1 is a schematic diagram showing an implementation environment involved in this application according to an exemplary embodiment. As shown in Figure 1, the implementation environment includes an interview client 100 and an interview server 200.
  • a wired or wireless network connection is established in advance between the interview client 100 and the interview server 200 to realize the interaction between the interview client 100 and the interview server 200.
  • the interview client 100 is used for displaying the interview questions, and correspondingly obtain the voice of the applicant for answering the interview questions, so as to transmit the obtained input voice to the interview server 200 for corresponding processing.
  • the interview server 200 receives the voice input by the interview client 100, it needs to perform voice recognition on the input voice to obtain the candidate's voice answering the interview question as input text, and correspond to the candidate for the obtained input text Evaluation of interview performance.
  • the interview server 200 assumes the role of the artificial intelligence interviewer.
  • the interview client 100 may be an electronic device such as a smart phone, a tablet computer, a notebook computer, a computer, etc., and the number thereof is not limited (only two are shown in FIG. 1).
  • the interview server 200 may be a server or a server cluster composed of several servers, and there is no restriction here.
  • Fig. 2 is a block diagram of a server according to an exemplary embodiment.
  • the server can be specifically implemented as the interview server 200 in the implementation environment shown in FIG. 1.
  • server is only an example adapted to this application, and cannot be considered as providing any restriction on the scope of use of this application.
  • the server also cannot be interpreted as needing to rely on or have one or more components in the exemplary server shown in FIG. 2.
  • the server includes: a power supply 210, an interface 230, at least one memory 250, and at least one central processing unit (CPU) Units) 270.
  • the power supply 210 is used to provide working voltage for each hardware device on the server.
  • the interface 230 includes at least one wired or wireless network interface 231, at least one serial-to-parallel conversion interface 233, at least one input/output interface 235, at least one USB interface 237, etc., for communicating with external devices.
  • the memory 250 can be a read-only memory, a random access memory, a magnetic disk or an optical disc, etc.
  • the resources stored on it include the operating system 251, application programs 253 or data 255, etc.
  • the storage method can be short-term storage or permanent storage. .
  • the operating system 251 is used to manage and control the hardware devices and application programs 253 on the server to realize the calculation and processing of the massive data 255 by the central processing unit 270. It can be Windows ServerTM, Mac OS XTM, UnixTM, LinuxTM, etc. .
  • the application program 253 is a computer program that completes at least one specific task based on the operating system 251. It may include at least one module (not shown in FIG. 2), and each module may include a series of computer programs for the server. Read instructions.
  • the data 255 may be interface metadata stored in a disk or the like.
  • the central processing unit 270 may include one or more processors, and is configured to communicate with the memory 250 via a bus for computing and processing the massive data 255 in the memory 250.
  • the server applicable to this application will read a series of computer-readable instructions stored in the memory 250 through the central processing unit 270 to complete the method for obtaining input text in the artificial intelligence interview described in the following embodiments .
  • this application can also be implemented by hardware circuits or hardware circuits in combination with software instructions. Therefore, implementation of this application is not limited to any specific hardware circuits, software, and combinations of the two.
  • Fig. 3 is a flowchart showing a method for acquiring input text in an artificial intelligence interview according to an exemplary embodiment. The method is suitable for the interview server 200 in the implementation environment shown in Fig. 1 to achieve accurate acquisition of the input text.
  • the method for obtaining input text in the artificial intelligence interview includes at least the following steps: Step 310: In the artificial intelligence interview, a preset general vocabulary is called Correct the input voice with the common word pronunciation comparison table, and perform voice recognition on the corrected voice to obtain the recognized text.
  • the input voice refers to the content of the applicant's answer to the interview question. Since the artificial intelligence interviewer needs to evaluate the interview performance of the candidate based on the content of the candidate’s answer, the artificial intelligence interviewer cannot directly score the interview performance of the candidate based on the input voice, and must convert the input voice into input text. Therefore, it is necessary to accurately convert the input voice.
  • Passive words refer to professional vocabulary such as professional nouns and spoken words involved in artificial intelligence interviews.
  • the common words contained should be different, and some common words can also contain corresponding Chinese and English expressions.
  • it can include general terms such as Convolutional Neural Networks (CNN), TensorFlow (a translation model based on neural networks), and K-means algorithm.
  • CNN Convolutional Neural Networks
  • TensorFlow a translation model based on neural networks
  • K-means algorithm K-means algorithm
  • the general word library refers to a set of general words composed of several general words. According to the phonetic information corresponding to the general words in the input speech, the corresponding general words can be found in the general word vocabulary.
  • the common word pronunciation comparison table refers to the common word pronunciation set composed of the pronunciation corresponding to each common word in the common word database.
  • the pronunciation comparison table may include initial and final information corresponding to the general word; when the general word is an English word, the pronunciation comparison table may be the pronunciation of the English word.
  • the pronunciation comparison table contains each kind of pronunciation information corresponding to the general word, and each kind of pronunciation information is set to correspond to the same general word. Therefore, the common word pronunciation comparison table and the common word database are mapped to each other.
  • this application collects in advance the common word database and common word pronunciation comparison table involved in different interview scenarios, and calls the pre-collected common word database and common words in the artificial intelligence interview.
  • the pronunciation comparison table corrects and recognizes the input speech, and can accurately recognize the common words in the input speech, so as to obtain accurate input text.
  • Step 330 Obtain a word segmentation set by performing word segmentation processing on the recognized text.
  • common word segmentation algorithms can be used to perform word segmentation on the recognized text. For example, you can use forward maximum matching segmentation algorithm, reverse maximum matching segmentation algorithm, two-way maximum matching segmentation algorithm, and other word segmentation algorithms based on string matching. You can also use word segmentation algorithms based on Statistical word segmentation algorithm, this section will not go into details here.
  • the stop words contained in the recognized text may be located according to a preset stop word database, and the stop words obtained by the positioning may be filtered. To perform word segmentation processing on the recognized text obtained by performing stop word filtering.
  • the word segmentation set corresponding to the recognized text can be obtained.
  • Step 350 For the word segmentation in the word segmentation set, respectively calculate the mutual information value of the word segmentation relative to the left and right side participles, and locate and identify homophone wrong words in the text according to the obtained mutual information value.
  • mutual information is the amount of information about another random variable contained in one random variable, and it is the mutual information between two random variables. According to the mutual information value between two random variables, the degree of correlation between the two random variables can be reflected.
  • Homophonic wrong words refer to words in the recognized text that have the correct pronunciation but the meaning of the text does not conform to the contextual information of the entire recognized text. For example, if the recognition text contains a sentence of "artificial intelligence technology field", the "function" is a homophone wrong word.
  • the mutual information value corresponding to the word segmentation includes the left mutual information value between the word segmentation and the participle located on its left side, and the right mutual information value between the word segmentation and the word segmentation located on its right side.
  • the calculation of the left and right mutual information values of the word segmentation in this embodiment is realized by combining the overall context information of the recognized text, when the word segmentation corresponds to the left and right mutual information lower than When the threshold is set, it means that the word segmentation does not meet the overall context information of the recognized text, and thus the word segmentation is positioned as a homophone error word.
  • Step 370 Extract the target word from the preset homophone word library, replace the homophone wrong word, and obtain the input text in the smart interview.
  • the homophone error word needs to be corrected, that is, the homophone error word in the recognized text is replaced with the correct word, so that the word conforms to the recognition
  • the preset homophone word database refers to a collection of homophone words collected in advance, that is, the homophone word database contains several words with the same pronunciation.
  • the homonymous erroneous words in the recognized text are replaced by extracting the target words with the same pronunciation from the homophone lexicon, so that the recognition after the homophone replacement is performed
  • the text acquisition is the input text in the smart interview.
  • the method provided in this embodiment not only can the general vocabulary in the interview field be accurately recognized, but also the homophones in the recognized text are corrected, and the input text obtained is largely consistent with the truthfulness of the applicant.
  • the expression is similar, so that the artificial intelligence interviewer can accurately obtain the answer content of the applicant, so that the current intelligent interview can be carried out effectively.
  • the method for obtaining input text in an artificial intelligence interview further includes the following steps: step 410, according to a preset modal particle dictionary, search for the recognized text Repeated modal particles; step 430, delete the repeated modal particles from the recognized text.
  • modal particles are words that applicants often use as connections and pauses when speaking. Especially when the interview is relatively tight, applicants will use modal particle pairs more frequently, and modal particles are useful for artificial intelligence interviews. The official has little effect on the interview performance of the candidate, and may even interfere with the candidate's evaluation due to the appearance of a large number of modal particles. Therefore, it is necessary to delete the repetitive modal particles in the recognition text accordingly.
  • the modal particle dictionary refers to a pre-collected collection containing several different modal particles.
  • the process of searching and recognizing repeated modal particles in the text according to the preset modal particle dictionary may be: according to the preset
  • the modal particle dictionary recognizes the modal particles in the recognized text word by word according to the text sequence of the recognized text, thereby obtaining the modal particles that appear repeatedly in the recognized text, and deletes the repeated modal words from the recognized text.
  • the repetitive words in the recognized text may be recognized word by word according to the text sequence of the recognized text, and then the repetitive modal particles may be determined according to the preset modal particle dictionary, thereby obtaining Recognize repeated modal particles in the text.
  • the process of acquiring the repetitive words in the recognized text and the process of judging whether the repetitive words are modal particles may be performed at the same time.
  • the process of locating and recognizing homophones in the text according to the mutual information value corresponding to the word segmentation in the word segmentation set may include the following steps: Step 351, according to each of the word segmentation sets The left mutual information value and right mutual information value corresponding to the word segmentation are respectively calculated for the mean value and standard deviation of the left mutual information value and the right mutual information value of the word segmentation set.
  • each participle in the word participle set contains the corresponding left mutual information value and right mutual information value.
  • the left mutual information value and/or right mutual information value are counted to obtain the normal distribution formed by the left mutual information value corresponding to each participle in the word segmentation set, and the normal distribution formed by the right mutual information value corresponding to each participle in the word participle set distributed.
  • the mean value and standard deviation of the left mutual information value of the word segmentation set, and the mean value of the right mutual information value of the word segmentation set and the right mutual information value can be calculated respectively. Standard deviation.
  • four related parameters can be obtained: the mean deviation of the left mutual information value, the standard deviation of the left mutual information, the mean difference of the right mutual information and the standard deviation of the left mutual information.
  • Step 353 Perform a difference operation on the mean and standard deviation of the left mutual information value and the right mutual information value, respectively, to obtain the threshold value of the left mutual information value and the right mutual information value of the word segmentation set.
  • the threshold value of the left mutual information value and the right mutual information value of the word segmentation set is calculated according to the 3-sigma principle in the normal distribution.
  • the 3-sigma principle indicates that the probability of the value distribution in ( ⁇ -3 ⁇ , ⁇ +3 ⁇ ) is 0.9974, where ⁇ represents the standard deviation in the normal distribution, that is, the standard deviation of the left mutual information or the standard deviation of the right mutual information, and ⁇ represents The mean value in the normal distribution is the mean value of the left mutual information or the mean value of the right mutual information.
  • left mutual trust value threshold left mutual trust value average-3 left mutual information standard deviation
  • right mutual trust value threshold right mutual trust value average-3 right mutual information Standard deviation
  • Step 355 Obtain the word segmentation whose left mutual information value and right mutual information value are less than the corresponding threshold in the word segmentation set as homophone error words.
  • the left mutual information value and right mutual information value corresponding to each word segmentation in the word segmentation set are counted, and the average value and standard deviation of the left mutual information value and the right mutual information value of the word segmentation set are calculated according to The obtained mean value and standard deviation are respectively calculated for the threshold value of the left mutual information value and the threshold value of the right mutual information value of the word segmentation set, so that the homophone error words in the recognized text can be accurately located according to the obtained threshold value.
  • the process of locating and recognizing homophones in the text may further include the following steps: step 354, according to the left mutual information value and the right mutual information
  • the weight given by the value is weighted and calculated on the thresholds of the left and right mutual information values of the word segmentation set to obtain the mutual information threshold corresponding to the word segmentation set.
  • the weights assigned to the left mutual information value and the right mutual information value are based on the importance of the left mutual information value for the recognition of homophone wrong words and the importance of the right mutual information value for the recognition of homophone wrong words. Yes, the weights of the two can be the same or different. According to the assigned weight, the threshold value of the left mutual information value and the right mutual information value of the word segmentation set is weighted and operated, and the unique mutual information threshold corresponding to the word segmentation set can be obtained.
  • step 355 includes at least the following steps: step 3551, for the word segmentation in the word segmentation set, perform a weighted sum operation on the left mutual information value and the right mutual information value corresponding to the word segmentation according to the weight to obtain The mutual information value of the word segmentation; step 3553, acquiring the word segmentation whose mutual information value is less than the mutual information threshold is a homophone error word.
  • the left and right mutual information values corresponding to each participle are still weighted and calculated according to the assigned weight to obtain the corresponding value of each participle.
  • the value of mutual information is a value of mutual information.
  • the mutual information value corresponding to each participle can be obtained through the weighted sum calculation.
  • this application calculates the mutual information threshold corresponding to the word segmentation set according to the pre-configured weights, and calculates the mutual information value of each word segmentation in the word segmentation set, so as to judge by the mutual information value and the mutual information threshold value of each word segmentation You can quickly get the homophone wrong words in the recognized text.
  • the process of extracting the target word from the preset homophone word library to replace the homophone error word includes the following steps: step 371, according to the pinyin corresponding to the homophone error word , Extract a number of candidate words that are homophones with the homonymous wrong words from the homophone word database.
  • the homophone dictionary refers to a collection of homophones collected in advance. Therefore, according to the corresponding pinyin of the homophone error word, several words that are homophones with the homophone error word can be extracted from the homophone word library as candidate words.
  • the homophone lexicon In the homophone lexicon, a two-dimensional table is used to realize the correspondence between words and pinyin. Since the computer cannot directly recognize the text, it can only recognize the computer code corresponding to the text. Therefore, the words contained in the homophone lexicon should be understood as It is computer code.
  • a pinyin syllable table is established in the homophone syllable to realize the mapping of multiple homophones through the pinyin syllable table, and the index of the pinyin syllable table is saved in the homophone syllable.
  • the index corresponds to the pinyin corresponding to the pinyin syllable table.
  • the pinyin syllable table for a specific pinyin, several homophones corresponding to the pinyin can be found.
  • the pinyin corresponding to the homonymous wrong word refers to the complete pinyin.
  • the index value corresponding to the homophone word database can be searched for the homophone word database, and the corresponding phonetic syllable can be obtained according to the searched index value. In this way, a number of candidate words that are homophones with the homophone of the wrong word can be queried from the phonetic syllable table.
  • Step 373 Replace the homophone error words one by one through the candidate words, and calculate the mutual information value corresponding to the candidate words.
  • Step 375 Extract the candidate word with the highest mutual information value as the target word to replace the homophone wrong word.
  • the candidate word with the highest mutual trust value indicates that the candidate word best fits the contextual information of the overall recognition text.
  • the candidate word with the highest mutual information value can be extracted as the target word to replace the homophone word, thereby obtaining the input text .
  • the homophone error word in the recognized text is replaced, thereby obtaining an accurate input text.
  • Fig. 8 is a block diagram showing a device for acquiring input text in an artificial intelligence interview according to an exemplary embodiment.
  • the device includes a speech recognition module 510, a word segmentation processing module 530, a homophone wrong word location module 550, and an input text acquisition module 570.
  • the voice recognition module 510 is used to call a preset general word database and a general word pronunciation comparison table to correct the input voice during the artificial intelligence interview, and perform voice recognition on the corrected voice to obtain a recognized text.
  • the word segmentation processing module 530 is configured to obtain a word segmentation set by performing word segmentation processing on the recognized text.
  • the homophone error word positioning module 550 is configured to calculate the mutual information value of the word segmentation relative to the left and right participles for the word segmentation in the word segmentation set, and locate the homophone error in the recognized text according to the obtained mutual information value Word, the mutual information value includes a left mutual information value and a right mutual information value.
  • the input text acquisition module 570 is configured to extract a target word from a preset homophone word library to replace the homophone error word, and obtain the input text in the smart interview.
  • the pronunciation of the target word is the same as the homophone error word .
  • the device further includes a modal particle search module and a modal particle deletion module, wherein the modal particle search module is used to search for repetitive modal particles in the recognized text according to a preset modal particle dictionary, The modal particle deletion module is used to delete the repeated modal particle from the recognized text.
  • the homophone error word location module 550 includes a mutual information acquisition unit, a mutual information threshold acquisition unit, and a homophone error word acquisition unit.
  • the mutual information acquiring unit is configured to calculate the mean value and standard deviation of the left mutual information value and the right mutual information value of the word segmentation set according to the left mutual information value and the right mutual information value corresponding to each word segmentation in the word segmentation set.
  • the mutual information threshold obtaining unit is configured to perform difference operations on the mean value and standard deviation of the left mutual information value and the right mutual information value, respectively, to obtain the threshold value of the word segmentation set with respect to the left mutual information value and the right mutual information value.
  • the homophone error word acquisition unit is used to acquire the word segmentation whose left mutual information value and right mutual information value are less than the corresponding threshold in the word segmentation set as homophone error words.
  • the input text acquisition module 570 includes a candidate word acquisition unit, a candidate word mutual information calculation unit, and a homophone error word correction unit.
  • the candidate word acquiring unit is configured to extract several candidate words that are homophones with the homophone error word from the homophone word database according to the pinyin corresponding to the homophone error word.
  • the candidate word mutual information calculation unit is configured to replace the homophone error words one by one through the candidate words, and calculate the mutual information value corresponding to the candidate words.
  • the homophone error word correction unit is used to extract the candidate word with the highest mutual information value as the target word to replace the homophone error word.
  • the candidate word obtaining unit includes an index value searching subunit and a Pinyin syllable table searching subunit.
  • the index value searching subunit is used to query the index value corresponding to the pinyin in the homophone word library according to the pinyin of the homophone error word, and the index value corresponds to the pinyin syllable table set in the homophone word library,
  • the pinyin syllable table is used to realize the mapping of multiple homophones.
  • the pinyin syllable table query subunit is used for querying the spliced syllable table for several candidate words that are homophones with the homophone error word according to the found index value.
  • the present application also provides a device for obtaining input text in an artificial intelligence interview.
  • the device includes: a processor; a memory, and computer-readable instructions are stored on the memory, and the computer-readable instructions are When the processor is executed, the method for obtaining input text in the artificial intelligence interview as described above is realized.
  • the present application also provides a computer-readable storage medium.
  • the computer-readable storage medium may be non-volatile or volatile.
  • a computer program is stored thereon, and when the computer program is executed by the processor, the method for obtaining input text in an artificial intelligence interview as described above is realized.

Abstract

A method and apparatus for obtaining input text in an artificial intelligence interview, which can be implemented by artificial intelligence machine learning. The method comprises: in an on-going artificial intelligence interview, invoking a preset general word bank and a general word pronunciation comparison table to correct an input speech, and performing speech recognition on the corrected speech to obtain recognized text (310); performing word segmentation processing on the recognized text to obtain a segmented word set (330); for segmented words in the segmented word set, calculating mutual information values of a segmented word with respect to a left segmented word and a right segmented word, and positioning a homophone error word in the recognized text according to the obtained mutual information values (350); and extracting a target word from a preset homophone word bank to replace the homophone error word, so as to obtain input text in the intelligent interview (370), the pronunciation of the target word being the same as that of the homophone error word. The input text obtained in the method is close to the real expression of the applicant to a great extent.

Description

人工智能面试中获取输入文本和相关装置Obtain input text and related devices in artificial intelligence interviews
本申请要求于2019年09月17日提交中国专利局、申请号为2019108770924，发明名称为“人工智能面试中获取输入文本和相关装置”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。This application claims the priority of a Chinese patent application filed with the Chinese Patent Office on September 17, 2019, with application number 2019108770924, and the title of the invention is "Acquisition of input text and related devices in artificial intelligence interviews", the entire content of which is incorporated by reference In this application.
技术领域Technical field
本申请涉及人工智能技术领域，特别涉及一种人工智能面试中获取输入文本的方法及装置、设备、计算机可读存储介质。This application relates to the field of artificial intelligence technology, and in particular to a method, device, equipment, and computer-readable storage medium for obtaining input text in artificial intelligence interviews.
背景技术Background technique
随着电子技术的不断发展，人工智能技术逐渐活跃在各种技术领域，例如智能汽车驾驶、智能客服、智能扫地机器人、智能面试等等。With the continuous development of electronic technology, artificial intelligence technology is gradually active in various technical fields, such as smart car driving, smart customer service, smart sweeping robots, smart interviews, and so on.
其中，智能面试是由人工智能面试官代替传统面试官对应聘者进行面试，人工智能面试官融合了语音识别、面部识别等功能，能够对应聘者的面试表现进行综合评价，并通过对应聘者进行排名来确定理想人选。Among them, the intelligent interview is that the artificial intelligence interviewer replaces the traditional interviewer to interview the candidate. The artificial intelligence interviewer integrates voice recognition, facial recognition and other functions, can comprehensively evaluate the interview performance of the candidate, and pass the corresponding candidate Perform rankings to determine ideal candidates.
发明人意识到，在智能面试中，应聘者对于面试题的回答仍是人工智能面试官评价应聘者的重要内容。由此，人工智能面试官能够准确获取应聘者的回答内容，决定了当前所进行智能面试的有效性。The inventor realized that in the smart interview, the applicant's answer to the interview questions is still an important part of the artificial intelligence interviewer's evaluation of the applicant. As a result, the artificial intelligence interviewer can accurately obtain the answer content of the applicant, which determines the effectiveness of the current intelligent interview.
需要说明的是，在上述背景技术部分公开的信息仅用于加强对本申请的背景的理解，因此可以包括不构成对本领域普通技术人员已知的现有技术的信息。It should be noted that the information disclosed in the background art section above is only used to enhance the understanding of the background of the application, and therefore may include information that does not constitute the prior art known to those of ordinary skill in the art.
技术问题technical problem
为了使得人工智能面试官能够准确获取应聘者针对面试题的回答内容，本申请提供了一种人工智能面试中获取输入文本的方法及装置、设备、计算机可读存储介质。In order to enable the artificial intelligence interviewer to accurately obtain the candidates' answers to the interview questions, this application provides a method, device, equipment, and computer-readable storage medium for obtaining input text in an artificial intelligence interview.
技术解决方案Technical solutions
一种人工智能面试中获取输入文本的方法，包括：在所进行的人工智能面试中，调用预设的通用词词库和通用词发音对照表对输入语音进行校正，并对校正所得语音进行语音识别获得识别文本；通过对所述识别文本进行分词处理获得分词集合；对所述分词集合中的分词，分别计算所述分词相对左侧分词和右侧分词的互信息值，且根据所得互信息值定位所述识别文本中的同音错误词，所述互信息值包括左互信息值和右互信息值；从预设的同音词词库中提取目标词对所述同音错误词进行替换，获得所述智能面试中的输入文本，所述目标词的读音与所述同音错误词相同。A method for obtaining input text in an artificial intelligence interview, including: in the artificial intelligence interview, calling a preset general word database and a general word pronunciation comparison table to correct the input voice, and perform voice on the corrected voice Recognize and obtain the recognized text; obtain a word segmentation set by performing word segmentation processing on the recognized text; for the word segmentation in the word segmentation set, respectively calculate the mutual information value of the word segmentation relative to the left and right participles, and according to the obtained mutual information The value locates the homophone error words in the recognized text, the mutual information value includes the left mutual information value and the right mutual information value; the target word is extracted from the preset homophone word database to replace the homophone error words to obtain all the homophone error words. Speaking of the input text in the smart interview, the pronunciation of the target word is the same as the homophone wrong word.
一种人工智能面试中获取输入文本的装置，包括：一种人工智能面试中获取输入文本的装置，其中，所述装置包括：语音识别模块，用于在所进行的人工智能面试中，调用预设的通用词词库和通用词发音对照表对输入语音进行校正，并对校正所得语音进行语音识别获得识别文本；分词处理模块，用于通过对所述识别文本进行分词处理获得分词集合；同音错误词定位模块，用于对所述分词集合中的分词，分别计算所述分词相对左侧分词和右侧分词的互信息值，且根据所得互信息值定位所述识别文本中的同音错误词，所述互信息值包括左互信息值和右互信息值；输入文本获取模块，用于从预设的同音词词库中提取目标词对所述同音错误词进行替换，获得所述智能面试中的输入文本，所述目标词的读音与所述同音错误词相同。A device for acquiring input text in an artificial intelligence interview, including: a device for acquiring input text in an artificial intelligence interview, wherein the device includes: a speech recognition module, used to call pre-programs in the artificial intelligence interview. The universal word database and the universal word pronunciation comparison table are set to correct the input speech, and perform speech recognition on the corrected speech to obtain the recognized text; the word segmentation processing module is used to obtain the word segmentation set by performing word segmentation processing on the recognized text; homophone The error word positioning module is used to calculate the mutual information value of the word segmentation relative to the left and right participles for the word segmentation in the word segmentation set, and locate the homophone error words in the recognized text according to the obtained mutual information value , The mutual information value includes a left mutual information value and a right mutual information value; the input text acquisition module is used to extract the target word from the preset homophone word database to replace the homophone error word, and obtain the intelligent interview In the input text of, the pronunciation of the target word is the same as the homonymous wrong word.
一种人工智能面试中获取输入文本的设备，所述设备包括处理器和存储器，所述存储器上存储有计算机可读指令，所述计算机可读指令被所述处理器执行时，实现如前任一项所述的人工智能面试中获取输入文本的方法。A device for acquiring input text in an artificial intelligence interview, the device comprising a processor and a memory, and computer-readable instructions are stored on the memory. When the computer-readable instructions are executed by the processor, the implementation of The method of obtaining input text in the artificial intelligence interview described in the item.
一种计算机可读存储介质，其上存储有计算机程序，所述计算机程序被处理器执行时，实现如前任一项所述的人工智能面试中获取输入文本的方法。A computer-readable storage medium has a computer program stored thereon, and when the computer program is executed by a processor, the method for obtaining input text in an artificial intelligence interview as described in any one of the preceding items is realized.
有益效果Beneficial effect
在上述技术方案中，通过调用预设的通用词词库和通用词发音对照表对人工智能面试中的输入语音进行校正，能够准确识别应聘者针对面试题所进行回答中的通用词汇，从而对校正语音进行语音识别获得准确的识别文本，然后对识别文本进行分词处理获得分词集合后，通过对分词集合中的分词分别计算相对左侧分词和右侧分词的互信息值，以根据所得互信息值定位识别文本中的同音错误词，最后通过从预设的同音词词库中提取目标词对同音错误词进行替换，获得智能面试中的输入文本。In the above technical solution, the input voice in the artificial intelligence interview is corrected by calling the preset general word database and the general word pronunciation comparison table, which can accurately identify the general vocabulary in the interview question by the applicant, thereby correcting Correct the speech for speech recognition to obtain accurate recognized text, and then perform word segmentation on the recognized text. After obtaining the word segmentation set, calculate the mutual information value of the relative left participle and the right participle by dividing the words in the word segmentation set to calculate the mutual information value according to the obtained mutual information Value positioning recognizes homophone wrong words in the text, and finally replaces homophone wrong words by extracting the target word from the preset homophone word database to obtain the input text in the smart interview.
由此，本申请不仅能够准确识别面试领域中的通用词汇，还修正了识别文本中的同音错误词，所得到的输入文本在很大程度上与应聘者的真实表达相接近，使得人工智能面试官能够准确获取应聘者的回答内容，使得当前所进行的智能面试能够有效进行。As a result, this application can not only accurately identify the general vocabulary in the interview field, but also correct the homophonic wrong words in the recognized text. The input text obtained is largely close to the applicant’s real expression, making the artificial intelligence interview The officer can accurately obtain the answer content of the applicant, so that the current intelligent interview can be carried out effectively.
应当理解的是，以上的一般描述和后文的细节描述仅是示例性的，并不能限制本申请。It should be understood that the above general description and the following detailed description are only exemplary and cannot limit the application.
附图说明Description of the drawings
图1是根据一示例性实施例示出的本申请所涉及实施环境的示意图。Fig. 1 is a schematic diagram showing an implementation environment involved in this application according to an exemplary embodiment.
图2是根据一示例性实施例示出的一种服务器的硬件框图。Fig. 2 is a hardware block diagram showing a server according to an exemplary embodiment.
图3是根据一示例性实施例示出的一种人工智能面试中获取输入文本的方法的流程图。Fig. 3 is a flow chart showing a method for obtaining input text in an artificial intelligence interview according to an exemplary embodiment.
图4是根据另一示例性实施例示出的一种人工智能面试中获取输入文本的方法的流程图。Fig. 4 is a flowchart showing a method for obtaining input text in an artificial intelligence interview according to another exemplary embodiment.
图5是图3所示步骤350在一个实施例的流程图。FIG. 5 is a flowchart of step 350 shown in FIG. 3 in an embodiment.
图6是图3所示步骤350在另一个实施例的流程图。FIG. 6 is a flowchart of step 350 shown in FIG. 3 in another embodiment.
图7是图3所示步骤370在一个实施例的流程图。FIG. 7 is a flowchart of step 370 shown in FIG. 3 in an embodiment.
图8是根据一示例性实施例所示出的一种人工智能面试中获取输入文本的装置的框图。Fig. 8 is a block diagram of a device for acquiring input text in an artificial intelligence interview according to an exemplary embodiment.
本发明的实施方式Embodiments of the present invention
这里将详细地对示例性实施例执行说明，其示例表示在附图中。下面的描述涉及附图时，除非另有表示，不同附图中的相同数字表示相同或相似的要素。以下示例性实施例中所描述的实施方式并不代表与本申请相一致的所有实施方式。相反，它们仅是与如所附权利要求书中所详述的、本申请的一些方面相一致的装置和方法的例子。Here, an exemplary embodiment will be described in detail, and examples thereof are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings indicate the same or similar elements. The implementation manners described in the following exemplary embodiments do not represent all implementation manners consistent with the present application. On the contrary, they are merely examples of devices and methods consistent with some aspects of the application as detailed in the appended claims.
图1是根据一示例性实施例示出的一种本申请所涉及实施环境的示意图。如图1所示，该实施环境包括面试客户端100和面试服务端200。Fig. 1 is a schematic diagram showing an implementation environment involved in this application according to an exemplary embodiment. As shown in Figure 1, the implementation environment includes an interview client 100 and an interview server 200.
其中，面试客户端100与面试服务端200之间预先建立有线或者无线网络连接，以实现面试客户端100与面试服务端200之间的交互。Wherein, a wired or wireless network connection is established in advance between the interview client 100 and the interview server 200 to realize the interaction between the interview client 100 and the interview server 200.
面试客户端100用于对面试题进行展示，并相应获取应聘者针对面试题进行回答的语音，以将获取的输入语音传输至面试服务端200进行相应处理。例如，面试服务端200接收到面试客户端100输入的语音后，需要对输入语音进行语音识别，以将应聘者针对面试题进行回答的语音获取为输入文本，并针对获取的输入文本对应聘者的面试表现进行评价。也即是说，在智能面试场景中，由面试服务端200担任人工智能面试官的角色。The interview client 100 is used for displaying the interview questions, and correspondingly obtain the voice of the applicant for answering the interview questions, so as to transmit the obtained input voice to the interview server 200 for corresponding processing. For example, after the interview server 200 receives the voice input by the interview client 100, it needs to perform voice recognition on the input voice to obtain the candidate's voice answering the interview question as input text, and correspond to the candidate for the obtained input text Evaluation of interview performance. In other words, in the intelligent interview scenario, the interview server 200 assumes the role of the artificial intelligence interviewer.
示例性的，面试客户端100可以是智能手机、平板电脑、笔记本电脑、计算机等电子设备，其数量不作限制（图1仅示出2个）。面试服务端200可以是一台服务器，也可以是由若干服务器构成的服务器集群，本处也不进行限制。Exemplarily, the interview client 100 may be an electronic device such as a smart phone, a tablet computer, a notebook computer, a computer, etc., and the number thereof is not limited (only two are shown in FIG. 1). The interview server 200 may be a server or a server cluster composed of several servers, and there is no restriction here.
图2是根据一示例性实施例所示出的一种服务器的框图。该服务器可以被具体实现为图1所示实施环境中的面试服务端200。Fig. 2 is a block diagram of a server according to an exemplary embodiment. The server can be specifically implemented as the interview server 200 in the implementation environment shown in FIG. 1.
需要说明的是，该服务器只是一个适配于本申请的示例，不能认为是提供了对本申请的使用范围的任何限制。该服务器也不能解释为需要依赖于或者必须具有图2中示出的示例性的服务器中的一个或者多个组件。It should be noted that the server is only an example adapted to this application, and cannot be considered as providing any restriction on the scope of use of this application. The server also cannot be interpreted as needing to rely on or have one or more components in the exemplary server shown in FIG. 2.
该服务器的硬件结构可因配置或者性能的不同而产生较大的差异，如图7所示，服务器包括：电源210、接口230、至少一存储器250、以及至少一中央处理器（CPU ，Central Processing Units）270。The hardware structure of the server may vary greatly due to differences in configuration or performance. As shown in FIG. 7, the server includes: a power supply 210, an interface 230, at least one memory 250, and at least one central processing unit (CPU) Units) 270.
其中，电源210用于为服务器上的各硬件设备提供工作电压。Wherein, the power supply 210 is used to provide working voltage for each hardware device on the server.
接口230包括至少一有线或无线网络接口231、至少一串并转换接口233、至少一输入输出接口235以及至少一USB接口237等，用于与外部设备通信。The interface 230 includes at least one wired or wireless network interface 231, at least one serial-to-parallel conversion interface 233, at least one input/output interface 235, at least one USB interface 237, etc., for communicating with external devices.
存储器250作为资源存储的载体，可以是只读存储器、随机存储器、磁盘或者光盘等，其上所存储的资源包括操作***251、应用程序253或者数据255等，存储方式可以是短暂存储或者永久存储。As a carrier for resource storage, the memory 250 can be a read-only memory, a random access memory, a magnetic disk or an optical disc, etc. The resources stored on it include the operating system 251, application programs 253 or data 255, etc. The storage method can be short-term storage or permanent storage. .
其中，操作***251用于管理与控制服务器上的各硬件设备以及应用程序253，以实现中央处理器270对海量数据255的计算与处理，其可以是Windows ServerTM、Mac OS XTM、UnixTM、LinuxTM等。应用程序253是基于操作***251之上完成至少一项特定工作的计算机程序，其可以包括至少一模块（图2中未示出），每个模块都可以分别包含有对服务器的一系列计算机可读指令。数据255可以是存储于磁盘中的接口元数据等。Among them, the operating system 251 is used to manage and control the hardware devices and application programs 253 on the server to realize the calculation and processing of the massive data 255 by the central processing unit 270. It can be Windows ServerTM, Mac OS XTM, UnixTM, LinuxTM, etc. . The application program 253 is a computer program that completes at least one specific task based on the operating system 251. It may include at least one module (not shown in FIG. 2), and each module may include a series of computer programs for the server. Read instructions. The data 255 may be interface metadata stored in a disk or the like.
中央处理器270可以包括一个或多个以上的处理器，并设置为通过总线与存储器250通信，用于运算与处理存储器250中的海量数据255。The central processing unit 270 may include one or more processors, and is configured to communicate with the memory 250 via a bus for computing and processing the massive data 255 in the memory 250.
如上面所详细描述的，适用本申请的服务器将通过中央处理器270读取存储器250中存储的一系列计算机可读指令的形式来完成以下实施例所述的人工智能面试中获取输入文本的方法。As described in detail above, the server applicable to this application will read a series of computer-readable instructions stored in the memory 250 through the central processing unit 270 to complete the method for obtaining input text in the artificial intelligence interview described in the following embodiments .
此外，通过硬件电路或者硬件电路结合软件指令也能同样实现本申请，因此，实现本申请并不限于任何特定硬件电路、软件以及两者的组合。In addition, this application can also be implemented by hardware circuits or hardware circuits in combination with software instructions. Therefore, implementation of this application is not limited to any specific hardware circuits, software, and combinations of the two.
图3是根据一例性实施例示出的一种人工智能面试中获取输入文本的方法的流程图，该方法适用于图1所示实施环境中的面试服务端200，以实现输入文本的准确获取。Fig. 3 is a flowchart showing a method for acquiring input text in an artificial intelligence interview according to an exemplary embodiment. The method is suitable for the interview server 200 in the implementation environment shown in Fig. 1 to achieve accurate acquisition of the input text.
如图3所示，在一示例性的实施例中，该人工智能面试中获取输入文本的方法至少包括以下步骤：步骤310，在所进行的人工智能面试中，调用预设的通用词词库和通用词发音对照表对输入语音进行校正，并对校正所得语音进行语音识别获得识别文本。As shown in FIG. 3, in an exemplary embodiment, the method for obtaining input text in the artificial intelligence interview includes at least the following steps: Step 310: In the artificial intelligence interview, a preset general vocabulary is called Correct the input voice with the common word pronunciation comparison table, and perform voice recognition on the corrected voice to obtain the recognized text.
其中，在所进行的人工智能面试中，输入语音是指应聘者针对面试题所进行的回答内容。由于人工智能面试官需要根据应聘者的回答内容对应聘者的面试表现进行评价，但人工智能面试官无法直接根据输入语音来对应聘者的面试表现进行打分，必须将输入语音转换为输入文本，由此，需要对输入语音进行准确转换。Among them, in the conducted artificial intelligence interview, the input voice refers to the content of the applicant's answer to the interview question. Since the artificial intelligence interviewer needs to evaluate the interview performance of the candidate based on the content of the candidate’s answer, the artificial intelligence interviewer cannot directly score the interview performance of the candidate based on the input voice, and must convert the input voice into input text. Therefore, it is necessary to accurately convert the input voice.
通过词是指，人工智能面试中所涉及到的专业名词、口语词等专业词汇。在人工智能面试所属的不同技术领域中，所含有的通用词应当不同，并且某些通用词还可以包含有对应的中英文表达。例如，在人工智能技术领域中，可以包括卷积神经网络（Convolutional Neural Networks, CNN）、TensorFlow（一种基于神经网络构建的翻译模型）、K均值算法等通用词。Passive words refer to professional vocabulary such as professional nouns and spoken words involved in artificial intelligence interviews. In the different technical fields of the artificial intelligence interview, the common words contained should be different, and some common words can also contain corresponding Chinese and English expressions. For example, in the field of artificial intelligence technology, it can include general terms such as Convolutional Neural Networks (CNN), TensorFlow (a translation model based on neural networks), and K-means algorithm.
通用词词库则是指由若干通用词构成的通用词集合，根据输入语音中对应于通用词的语音信息，可以在通用词词库中查找到对应的通用词。The general word library refers to a set of general words composed of several general words. According to the phonetic information corresponding to the general words in the input speech, the corresponding general words can be found in the general word vocabulary.
通用词发音对照表则是指，由通用词词库中每一通用词所对应发音构成的通用词发音集合。其中，当通用词为中文词语时，发音对照表中可以包括该通用词对应的声母信息和韵母信息；当通用词为英文词语来说，发音对照表可以是该英文词语的读音。此外，当同一通用词具有多种读法时，发音对照表中包含有该通用词对应的每种发音信息，并且设置每种发音信息共同对应于同一通用词。由此，通用词发音对照表与通用词词库之间是相互映射的。The common word pronunciation comparison table refers to the common word pronunciation set composed of the pronunciation corresponding to each common word in the common word database. Wherein, when the general word is a Chinese word, the pronunciation comparison table may include initial and final information corresponding to the general word; when the general word is an English word, the pronunciation comparison table may be the pronunciation of the English word. In addition, when the same general word has multiple pronunciations, the pronunciation comparison table contains each kind of pronunciation information corresponding to the general word, and each kind of pronunciation information is set to correspond to the same general word. Therefore, the common word pronunciation comparison table and the common word database are mapped to each other.
在进行输入语音的识别中，需要先根据通用词发音对照表对应聘者所输入语音中对应于通用词的语音信息进行校正，以得到发音准确的输入语音，然后对校正所得到的输入语音进行语音识别，获得识别文本。In the recognition of the input speech, it is necessary to correct the speech information corresponding to the common words in the speech input by the applicant according to the common word pronunciation comparison table to obtain the input speech with accurate pronunciation, and then perform the correction on the input speech obtained by the correction. Voice recognition to obtain recognized text.
在人工智能面试场景中，对应聘者的输入语音进行识别的难度之一即在于识别面试所属行业或者技术领域中涉及的通用词，如果不能准确识别这些通用词，则不能准确地理解应聘者的语义表达，由此，本申请通过预先收集不同面试场景所涉及的通用词词库以及通用词发音对照表，并在所进行的人工智能面试中，通过调用预先收集的通用词词库和通用词发音对照表对输入语音进行校正和识别，能够对输入语音中的通用词进行准确识别，从而能够获得准确的输入文本。In the artificial intelligence interview scenario, one of the difficulties in recognizing the input voice of the candidate is to recognize the common words involved in the industry or technical field of the interview. If these common words cannot be accurately recognized, the candidate's voice cannot be accurately understood. Semantic expression. Therefore, this application collects in advance the common word database and common word pronunciation comparison table involved in different interview scenarios, and calls the pre-collected common word database and common words in the artificial intelligence interview. The pronunciation comparison table corrects and recognizes the input speech, and can accurately recognize the common words in the input speech, so as to obtain accurate input text.
步骤330，通过对识别文本进行分词处理获得分词集合。Step 330: Obtain a word segmentation set by performing word segmentation processing on the recognized text.
其中，对识别文本进行分词处理可以选用常用的分词算法进行，例如可以选用正向最大匹配分词算法、逆向最大匹配分词算法、双向最大匹配分词算法等基于字符串匹配的分词算法，还可以选用基于统计的分词算法，本处不对此进行赘述。Among them, common word segmentation algorithms can be used to perform word segmentation on the recognized text. For example, you can use forward maximum matching segmentation algorithm, reverse maximum matching segmentation algorithm, two-way maximum matching segmentation algorithm, and other word segmentation algorithms based on string matching. You can also use word segmentation algorithms based on Statistical word segmentation algorithm, this section will not go into details here.
此外，在一个实施例中，在对识别文本进行分词处理之前，可以先根据预设的停用词词库来定位识别文本中含有的停用词，并对定位得到的停用词进行过滤，以对进行停用词过滤所得的识别文本进行分词处理。In addition, in one embodiment, before performing word segmentation processing on the recognized text, the stop words contained in the recognized text may be located according to a preset stop word database, and the stop words obtained by the positioning may be filtered. To perform word segmentation processing on the recognized text obtained by performing stop word filtering.
由此，本实施例通过对识别文本进行分词处理，能够获得识别文本所对应的分词集合。Therefore, in this embodiment, by performing word segmentation processing on the recognized text, the word segmentation set corresponding to the recognized text can be obtained.
步骤350，对分词集合中的分词，分别计算分词相对左侧分词和右侧分词的互信息值，且根据所得互信息值定位识别文本中的同音错误词。Step 350: For the word segmentation in the word segmentation set, respectively calculate the mutual information value of the word segmentation relative to the left and right side participles, and locate and identify homophone wrong words in the text according to the obtained mutual information value.
其中，互信息是一个随机变量中包含的关于另一个随机变量的信息量，是两个随机变量之间的互享信息。根据两个随机变量之间的互信息值，可以反映得出两个随机变量之间的关联程度。Among them, mutual information is the amount of information about another random variable contained in one random variable, and it is the mutual information between two random variables. According to the mutual information value between two random variables, the degree of correlation between the two random variables can be reflected.
同音错误词是指，识别文本中读音正确但是文字含义不符合识别文本整体的语境信息的词语。例如，若识别文本中含有“人工智能技术领域”的语句，该“职能”则为同音错误词。Homophonic wrong words refer to words in the recognized text that have the correct pronunciation but the meaning of the text does not conform to the contextual information of the entire recognized text. For example, if the recognition text contains a sentence of "artificial intelligence technology field", the "function" is a homophone wrong word.
在本实施例中，分词所对应的互信息值包括分词与位于其左侧的分词之间的左互信息值，以及分词与位于其右侧的分词之间的右互信息值。通过对分词集合中的每一个分词，分别计算该分词相对其左侧分词的左互信息值，以及计算该分词相对其右侧分词的右信息值，从而根据该分词所对应的左、右互信息值判断该分词是否为同音错误词。In this embodiment, the mutual information value corresponding to the word segmentation includes the left mutual information value between the word segmentation and the participle located on its left side, and the right mutual information value between the word segmentation and the word segmentation located on its right side. Through each participle in the word segmentation set, calculate the left mutual information value of the participle relative to its left participle, and calculate the right information value of the participle relative to its right participle, so as to calculate the left and right mutual information corresponding to the participle. The information value judges whether the word segmentation is a homophone wrong word.
需要说明的是，本实施例对分词进行左、右互信息值的计算是根据常用的互信息值计算方法所实现的，本处不对具体的计算过程进行赘述。It should be noted that the calculation of the left and right mutual information values of the word segmentation in this embodiment is implemented according to the commonly used mutual information value calculation method, and the specific calculation process is not repeated here.
此外，还需要说明的是，由于本实施例对分词进行左、右互信息值的计算是结合识别文本的整体语境信息所实现的，因此，当分词所对应的左、右互信息低于设定阈值时，则表示该分词不符合识别文本的整体语境信息，由此定位该分词为同音错误词。In addition, it should be noted that, since the calculation of the left and right mutual information values of the word segmentation in this embodiment is realized by combining the overall context information of the recognized text, when the word segmentation corresponds to the left and right mutual information lower than When the threshold is set, it means that the word segmentation does not meet the overall context information of the recognized text, and thus the word segmentation is positioned as a homophone error word.
步骤370，从预设的同音词词库中提取目标词对同音错误词进行替换，获得智能面试中的输入文本。Step 370: Extract the target word from the preset homophone word library, replace the homophone wrong word, and obtain the input text in the smart interview.
其中，为了得到准确的输入文本，在定位得到识别文本中的同音错误词后，需要对该同音错误词进行修正，也即将识别文本中的同音错误词替换为正确的词语，使得该词语符合识别文本的整体语境信息，由此将进行同音错误词替换后的识别文本获取为输入文本。Among them, in order to obtain accurate input text, after locating the homophone error word in the recognized text, the homophone error word needs to be corrected, that is, the homophone error word in the recognized text is replaced with the correct word, so that the word conforms to the recognition The overall context information of the text, from which the recognized text after the replacement of the homophone error word is obtained as the input text.
预设的同音词词库是指，预先收集的同音词集合，也即是说，该同音词词库中含有读音相同的若干词语。The preset homophone word database refers to a collection of homophone words collected in advance, that is, the homophone word database contains several words with the same pronunciation.
由此，在定位得到识别文本中的同音错误词之后，通过从同音词词库中提取读音与同音错误词相同的目标词对识别文本中的同音错误词进行替换，从而将进行同音词替换后的识别文本获取为智能面试中的输入文本。Therefore, after locating the homonymous erroneous words in the recognized text, the homonymous erroneous words in the recognized text are replaced by extracting the target words with the same pronunciation from the homophone lexicon, so that the recognition after the homophone replacement is performed The text acquisition is the input text in the smart interview.
由此，在本实施例所提供的方法中，不仅能够准确识别面试领域中的通用词汇，还修正了识别文本中的同音错误词，所得到的输入文本在很大程度上与应聘者的真实表达相接近，使得人工智能面试官能够准确获取应聘者的回答内容，使得当前所进行的智能面试能够有效进行。Therefore, in the method provided in this embodiment, not only can the general vocabulary in the interview field be accurately recognized, but also the homophones in the recognized text are corrected, and the input text obtained is largely consistent with the truthfulness of the applicant. The expression is similar, so that the artificial intelligence interviewer can accurately obtain the answer content of the applicant, so that the current intelligent interview can be carried out effectively.
在另一示例性的实施例中，如图4所示，在步骤330之前，人工智能面试中获取输入文本的方法还包括以下步骤：步骤410，根据预先设置的语气词词典，查找识别文本中重复出现的语气词；步骤430，将重复出现的语气词从识别文本中删除。In another exemplary embodiment, as shown in FIG. 4, before step 330, the method for obtaining input text in an artificial intelligence interview further includes the following steps: step 410, according to a preset modal particle dictionary, search for the recognized text Repeated modal particles; step 430, delete the repeated modal particles from the recognized text.
其中，语气词是应聘者在说话时经常用来作为连接、停顿的词，尤其是在面试环节相对紧张的情况下，应聘者使用语气词对的频率会更高，而语气词对于人工智能面试官对应聘者的面试表现的评价作用不大，甚至可能由于大量语气词的出现对应聘者的评价造成干扰，由此，有必要将识别文本中的重复出现的语气词相应删除。Among them, modal particles are words that applicants often use as connections and pauses when speaking. Especially when the interview is relatively tight, applicants will use modal particle pairs more frequently, and modal particles are useful for artificial intelligence interviews. The official has little effect on the interview performance of the candidate, and may even interfere with the candidate's evaluation due to the appearance of a large number of modal particles. Therefore, it is necessary to delete the repetitive modal particles in the recognition text accordingly.
语气词词典是指预先收集的含有若干不同语气词的集合，在一示例性的实施例中，根据预先设置的语气词词典查找识别文本中重复出现的语气词的过程可以是：根据预先设置的语气词词典且按照识别文本的文本先后顺序，逐字识别该识别文本中的语气词，由此获取识别文本中重复出现的语气词，并将重复出现的语气词从识别文本中删除。The modal particle dictionary refers to a pre-collected collection containing several different modal particles. In an exemplary embodiment, the process of searching and recognizing repeated modal particles in the text according to the preset modal particle dictionary may be: according to the preset The modal particle dictionary recognizes the modal particles in the recognized text word by word according to the text sequence of the recognized text, thereby obtaining the modal particles that appear repeatedly in the recognized text, and deletes the repeated modal words from the recognized text.
在另一示例性的实施例中，可以先按照识别文本的文本先后顺序，逐字识别该识别文本中重复出现的词语，然后根据预先设置的语气词词典确定重复出现的语气词，由此获取识别文本中重复出现的语气词。In another exemplary embodiment, the repetitive words in the recognized text may be recognized word by word according to the text sequence of the recognized text, and then the repetitive modal particles may be determined according to the preset modal particle dictionary, thereby obtaining Recognize repeated modal particles in the text.
而在另外的实施例中，对识别文本中重复出现的词语的获取过程以及对该重复出现的词语是否为语气词的判断过程可以是同时进行的。In another embodiment, the process of acquiring the repetitive words in the recognized text and the process of judging whether the repetitive words are modal particles may be performed at the same time.
由此，本实施例通过对识别文本中重复出现的语气词进行识别，并将识别文本中重复出现的语气词相应删除，有利于人工智能面试官对所获取输入文本的后续处理。Therefore, in this embodiment, by recognizing the modal particles that appear repeatedly in the recognized text, and correspondingly deleting the modal particles that appear repeatedly in the recognized text, it is beneficial to the artificial intelligence interviewer's subsequent processing of the acquired input text.
在另一示例性的实施例中，如图5所示，根据分词集合中分词所对应的互信息值定位识别文本中的同音错误词的过程可以包括以下步骤：步骤351，根据分词集合中各分词对应的左互信息值和右互信息值，分别计算分词集合关于左互信息值和右互信息值的均值以及标准差。In another exemplary embodiment, as shown in FIG. 5, the process of locating and recognizing homophones in the text according to the mutual information value corresponding to the word segmentation in the word segmentation set may include the following steps: Step 351, according to each of the word segmentation sets The left mutual information value and right mutual information value corresponding to the word segmentation are respectively calculated for the mean value and standard deviation of the left mutual information value and the right mutual information value of the word segmentation set.
其中，除分词集合中的第一个分词以及最后一个分词外，分词集合中的每一个分词均含有对应的左互信息值和右互信息值，由此，通过对分词集合中各分词对应的左互信息值和/或右互信息值进行统计，获得由分词集合中各分词所对应左互信息值构成的正态分布，以及由分词集合中各分词所对应右互信息值构成的正态分布。Among them, except for the first participle and the last participle in the word participle set, each participle in the word participle set contains the corresponding left mutual information value and right mutual information value. The left mutual information value and/or right mutual information value are counted to obtain the normal distribution formed by the left mutual information value corresponding to each participle in the word segmentation set, and the normal distribution formed by the right mutual information value corresponding to each participle in the word participle set distributed.
通过对所统计的分词集合中各分词对应的左互信息值以及右互信息值，可以分别计算得到分词集合关于左互信息值的均值和标准差，以及分词集合关于右互信息值的均值和标准差。By calculating the left mutual information value and the right mutual information value corresponding to each word in the word segmentation set, the mean value and standard deviation of the left mutual information value of the word segmentation set, and the mean value of the right mutual information value of the word segmentation set and the right mutual information value can be calculated respectively. Standard deviation.
也即是说，在本实施例中，可以得到左互信息值均差、左互信息标准差、右互信息值均差和左互信息标准差四个相关参数。That is to say, in this embodiment, four related parameters can be obtained: the mean deviation of the left mutual information value, the standard deviation of the left mutual information, the mean difference of the right mutual information and the standard deviation of the left mutual information.
步骤353，分别对左互信息值和右互信息值的均值以及标准差进行差值运算，获得分词集合关于左互信息值和右互信息值的阈值。Step 353: Perform a difference operation on the mean and standard deviation of the left mutual information value and the right mutual information value, respectively, to obtain the threshold value of the left mutual information value and the right mutual information value of the word segmentation set.
其中，分词集合关于左互信息值和右互信息值的阈值是根据正态分布中的3-sigma原则所计算得到的。3-sigma原则表示数值分布在（μ—3σ，μ+3σ)中的概率为0.9974 ，其中σ代表正态分布中的标准差，即上述左互信息标准差或者右互信息标准差，μ代表正态分布中的均值，即上述左互信息均值或者右互信息均值。Among them, the threshold value of the left mutual information value and the right mutual information value of the word segmentation set is calculated according to the 3-sigma principle in the normal distribution. The 3-sigma principle indicates that the probability of the value distribution in (μ-3σ, μ+3σ) is 0.9974, where σ represents the standard deviation in the normal distribution, that is, the standard deviation of the left mutual information or the standard deviation of the right mutual information, and μ represents The mean value in the normal distribution is the mean value of the left mutual information or the mean value of the right mutual information.
根据3-sigma原则计算分词集合关于左互信息值的阈值的公式为：左互信值阈值=左互信值均值-3左互信息标准差，右互信值阈值=右互信值均值-3右互信息标准差。According to the 3-sigma principle, the formula for calculating the threshold of the left mutual information value of the word segmentation set is: left mutual trust value threshold = left mutual trust value average-3 left mutual information standard deviation, right mutual trust value threshold = right mutual trust value average-3 right mutual information Standard deviation.
步骤355，获取分词集合中左互信息值和右互信息值小于对应阈值的分词为同音错误词。Step 355: Obtain the word segmentation whose left mutual information value and right mutual information value are less than the corresponding threshold in the word segmentation set as homophone error words.
其中，对于分词集合中分词对应的左互信息值小于步骤353计算所得的左互信息阈值、以及分词所对应右互信息值小于步骤353计算所得的右互信息阈值的分词，可以判断为是识别文本中的同音错误词。Among them, for a word segmentation whose left mutual information value corresponding to the word segmentation in the word segmentation set is less than the left mutual information threshold calculated in step 353, and the right mutual information value corresponding to the word segmentation is less than the right mutual information threshold calculated in step 353, it can be judged as recognition Homonymous wrong words in the text.
由此，本实施例通过对分词集合中每一分词对应的左互信息值和右互信息值进行统计，并计算分词集合关于左互信息值和右互信息值的均值以及标准差，以根据所得均值和标准差分别计算分词集合关于左互信息值的阈值以及关于右互信息值的阈值，从而能够根据所得阈值准确定位得到识别文本中的同音错误词。Therefore, in this embodiment, the left mutual information value and right mutual information value corresponding to each word segmentation in the word segmentation set are counted, and the average value and standard deviation of the left mutual information value and the right mutual information value of the word segmentation set are calculated according to The obtained mean value and standard deviation are respectively calculated for the threshold value of the left mutual information value and the threshold value of the right mutual information value of the word segmentation set, so that the homophone error words in the recognized text can be accurately located according to the obtained threshold value.
在另一示例性的实施例中，如图6所示，在步骤355之前，定位识别文本中的同音错误词的过程还可以包括以下步骤：步骤354，根据为左互信息值和右互信息值所赋予的权重，对分词集合关于左互信息值和右互信息值的阈值进行加权和运算，获得分词集合对应的互信息阈值。In another exemplary embodiment, as shown in FIG. 6, before step 355, the process of locating and recognizing homophones in the text may further include the following steps: step 354, according to the left mutual information value and the right mutual information The weight given by the value is weighted and calculated on the thresholds of the left and right mutual information values of the word segmentation set to obtain the mutual information threshold corresponding to the word segmentation set.
其中，为左互信息值和右互信息值所赋予的权重，是根据左互信息值对于同音错误词识别的重要性、以及右互信息值对于同音错误词识别的重要性所进行针对性赋予的，二者权重可以相同，也可以不同。根据所赋予的权重对分词集合关于左互信息值和右互信息值的阈值进行加权和运算，能够得到分词集合所对应唯一的互信息阈值。Among them, the weights assigned to the left mutual information value and the right mutual information value are based on the importance of the left mutual information value for the recognition of homophone wrong words and the importance of the right mutual information value for the recognition of homophone wrong words. Yes, the weights of the two can be the same or different. According to the assigned weight, the threshold value of the left mutual information value and the right mutual information value of the word segmentation set is weighted and operated, and the unique mutual information threshold corresponding to the word segmentation set can be obtained.
相应的，在一示例性的实施例中，步骤355至少包括以下步骤：步骤3551，对分词集合中的分词，按照权重对分词对应的左互信息值和右互信息值进行加权和运算，获得分词的互信息值；步骤3553，获取所述互信息值小于所述互信息阈值的分词为同音错误词。Correspondingly, in an exemplary embodiment, step 355 includes at least the following steps: step 3551, for the word segmentation in the word segmentation set, perform a weighted sum operation on the left mutual information value and the right mutual information value corresponding to the word segmentation according to the weight to obtain The mutual information value of the word segmentation; step 3553, acquiring the word segmentation whose mutual information value is less than the mutual information threshold is a homophone error word.
其中，在对分词集合中的分词进行同音错误词的判断时，仍按照所赋予的权重对每一分词对应的左互信息值和右互信息值进行加权和运算，以获取每一分词所对应的互信息值。Among them, when judging homophone wrong words for the participles in the word segmentation set, the left and right mutual information values corresponding to each participle are still weighted and calculated according to the assigned weight to obtain the corresponding value of each participle. The value of mutual information.
也即是说，对于分词集合中的每一分词，通过所进行的加权和计算，能够获得各分词对应的互信息值。In other words, for each participle in the word segmentation set, the mutual information value corresponding to each participle can be obtained through the weighted sum calculation.
由此，本申请按照预先配设的权重计算获得分词集合对应的互信息阈值，以及计算获得分词集合中每一分词的互信息值，从而通过对各分词的互信息值与互信息阈值进行判断即可快速获得识别文本中的同音错误词。Therefore, this application calculates the mutual information threshold corresponding to the word segmentation set according to the pre-configured weights, and calculates the mutual information value of each word segmentation in the word segmentation set, so as to judge by the mutual information value and the mutual information threshold value of each word segmentation You can quickly get the homophone wrong words in the recognized text.
在另一示例性的实施例中，如图7所示，从预设的同音词词库中提取目标词对同音错误词进行替换的过程包括如下步骤：步骤371，根据同音错误词所对应的拼音，从同音词词库中提取与同音错误词同音的若干候选词。In another exemplary embodiment, as shown in FIG. 7, the process of extracting the target word from the preset homophone word library to replace the homophone error word includes the following steps: step 371, according to the pinyin corresponding to the homophone error word , Extract a number of candidate words that are homophones with the homonymous wrong words from the homophone word database.
如前所述，同音词词库是指预先收集的同音词集合，由此，根据同音错误词对应的拼音，能够从同音词词库中提取得到与同音错误词同音的若干词语为候选词。As mentioned above, the homophone dictionary refers to a collection of homophones collected in advance. Therefore, according to the corresponding pinyin of the homophone error word, several words that are homophones with the homophone error word can be extracted from the homophone word library as candidate words.
在同音词词库中，通过一个二维表来实现词语和拼音的对应关系，由于计算机并不能直接识别文字，只能识别文字所对应的计算机编码，因此同音词词库中所含有的词语应当理解为是计算机编码。In the homophone lexicon, a two-dimensional table is used to realize the correspondence between words and pinyin. Since the computer cannot directly recognize the text, it can only recognize the computer code corresponding to the text. Therefore, the words contained in the homophone lexicon should be understood as It is computer code.
对于同一个拼音对应于多个词语的情况，同音词词库中通过建立一个拼音音节表，以通过该拼音音节表实现多个同音词语的映射，并在同音词词库中保存拼音音节表的索引，该索引对应于拼音音节表对应的拼音。在拼音音节表中，对于某一特定拼音，能够查找到对应于该拼音的若干同音词。For the case that the same pinyin corresponds to multiple words, a pinyin syllable table is established in the homophone syllable to realize the mapping of multiple homophones through the pinyin syllable table, and the index of the pinyin syllable table is saved in the homophone syllable. The index corresponds to the pinyin corresponding to the pinyin syllable table. In the pinyin syllable table, for a specific pinyin, several homophones corresponding to the pinyin can be found.
由于识别文本中仅含有同音错误词对应的文字，并不含有该同音错误的拼音，因此需要获取同音错误词对应的拼音。需要说明的是，在本实施例中，同音错误词所对应的拼音是指完整的拼音。Since the recognized text only contains the text corresponding to the homonymous wrong word, and does not contain the homonymous wrong pinyin, it is necessary to obtain the corresponding pinyin of the homonymous wrong word. It should be noted that, in this embodiment, the pinyin corresponding to the homonymous wrong word refers to the complete pinyin.
实现文字与拼音的转换，需要结合文字的计算机编码技术，其中不同的文字格式对应于不同的编码格式，本处不进行限制。To realize the conversion between text and pinyin, it is necessary to combine text computer coding technology. Different text formats correspond to different coding formats, and there is no restriction here.
由此，在对同音错误词进行文字与拼音的转换之后，根据同音错误词的拼音，可以在同音词词库查找该拼音所对应的索引值，以根据查找到的索引值获取得到相应的拼音音节表，从而可以从拼音音节表中查询到与同音错误词同音的若干候选词。Therefore, after the text and pinyin conversion of the homophone error word, according to the phonetic alphabet of the homophone error word, the index value corresponding to the homophone word database can be searched for the homophone word database, and the corresponding phonetic syllable can be obtained according to the searched index value. In this way, a number of candidate words that are homophones with the homophone of the wrong word can be queried from the phonetic syllable table.
在所得到的若干候选词中，存在一正确的词语为目标词，以对识别文本中的同音错误词进行替换，以得到正确的输入文本。Among the obtained candidate words, there is a correct word as the target word, and the homophone wrong word in the recognized text is replaced to obtain the correct input text.
步骤373，通过候选词逐一对同音错误词进行替换，并计算候选词对应的互信息值。Step 373: Replace the homophone error words one by one through the candidate words, and calculate the mutual information value corresponding to the candidate words.
其中，为了从候选词中确定对同音错误词进行替换的目标词，先将候选词逐个对识别文本中的同音错误词进行替换，并计算所替换的候选词相对左侧分词的左互信息值以及相对右侧分词的右互信息值。Among them, in order to determine the target word to replace the homophone error word from the candidate words, first replace the homophone error word in the recognized text one by one with the candidate words, and calculate the left mutual information value of the replaced candidate word relative to the left participle And the value of the right mutual information relative to the right participle.
步骤375，提取互信息值最高的候选词作为目标词对同音错误词进行替换。Step 375: Extract the candidate word with the highest mutual information value as the target word to replace the homophone wrong word.
其中，互信值最高的候选词表示该候选词最为贴合于识别文本整体的语境信息，由此，可以提取互信息值最高的候选词为目标词对同音错误词进行替换，从而得到输入文本。Among them, the candidate word with the highest mutual trust value indicates that the candidate word best fits the contextual information of the overall recognition text. As a result, the candidate word with the highest mutual information value can be extracted as the target word to replace the homophone word, thereby obtaining the input text .
由此，根据本实施例所提供的方法，通过从同音词词库中提取正确的同音词对识别文本中的同音错误词进行替换，由此获得准确的输入文本。Therefore, according to the method provided in this embodiment, by extracting the correct homophone from the homophone word database, the homophone error word in the recognized text is replaced, thereby obtaining an accurate input text.
图8是根据一示例性实施例示出的一种人工智能面试中获取输入文本的装置的框图。如图8所示，该装置包括语音识别模块510、分词处理模块530、同音错误词定位模块550和输入文本获取模块570。Fig. 8 is a block diagram showing a device for acquiring input text in an artificial intelligence interview according to an exemplary embodiment. As shown in FIG. 8, the device includes a speech recognition module 510, a word segmentation processing module 530, a homophone wrong word location module 550, and an input text acquisition module 570.
语音识别模块510用于在所进行的人工智能面试中，调用预设的通用词词库和通用词发音对照表对输入语音进行校正，并对校正所得语音进行语音识别获得识别文本。The voice recognition module 510 is used to call a preset general word database and a general word pronunciation comparison table to correct the input voice during the artificial intelligence interview, and perform voice recognition on the corrected voice to obtain a recognized text.
分词处理模块530用于通过对所述识别文本进行分词处理获得分词集合。The word segmentation processing module 530 is configured to obtain a word segmentation set by performing word segmentation processing on the recognized text.
同音错误词定位模块550用于对所述分词集合中的分词，分别计算所述分词相对左侧分词和右侧分词的互信息值，且根据所得互信息值定位所述识别文本中的同音错误词，所述互信息值包括左互信息值和右互信息值。The homophone error word positioning module 550 is configured to calculate the mutual information value of the word segmentation relative to the left and right participles for the word segmentation in the word segmentation set, and locate the homophone error in the recognized text according to the obtained mutual information value Word, the mutual information value includes a left mutual information value and a right mutual information value.
输入文本获取模块570用于从预设的同音词词库中提取目标词对所述同音错误词进行替换，获得所述智能面试中的输入文本，所述目标词的读音与所述同音错误词相同。The input text acquisition module 570 is configured to extract a target word from a preset homophone word library to replace the homophone error word, and obtain the input text in the smart interview. The pronunciation of the target word is the same as the homophone error word .
在一示例性的实施例中，该装置还包括语气词查找模块和语气词删除模块，其中语气词查找模块用于根据预先设置的语气词词典，查找所述识别文本中重复出现的语气词，语气词删除模块用于将所述重复出现的语气词从所述识别文本中删除。In an exemplary embodiment, the device further includes a modal particle search module and a modal particle deletion module, wherein the modal particle search module is used to search for repetitive modal particles in the recognized text according to a preset modal particle dictionary, The modal particle deletion module is used to delete the repeated modal particle from the recognized text.
在一示例性的实施例中，同音错误词定位模块550包括互信息获取单元、互信息阈值获取单元和同音错误词获取单元。In an exemplary embodiment, the homophone error word location module 550 includes a mutual information acquisition unit, a mutual information threshold acquisition unit, and a homophone error word acquisition unit.
互信息获取单元用于根据所述分词集合中各分词对应的左互信息值和右互信息值，分别计算所述分词集合关于所述左互信息值和右互信息值的均值以及标准差。The mutual information acquiring unit is configured to calculate the mean value and standard deviation of the left mutual information value and the right mutual information value of the word segmentation set according to the left mutual information value and the right mutual information value corresponding to each word segmentation in the word segmentation set.
互信息阈值获取单元用于分别对所述左互信息值和右互信息值的均值以及标准差进行差值运算，获得所述分词集合关于所述左互信息值和右互信息值的阈值。The mutual information threshold obtaining unit is configured to perform difference operations on the mean value and standard deviation of the left mutual information value and the right mutual information value, respectively, to obtain the threshold value of the word segmentation set with respect to the left mutual information value and the right mutual information value.
同音错误词获取单元用于获取所述分词集合中左互信息值和右互信息值小于对应阈值的分词为同音错误词。The homophone error word acquisition unit is used to acquire the word segmentation whose left mutual information value and right mutual information value are less than the corresponding threshold in the word segmentation set as homophone error words.
在一示例性的实施例中，输入文本获取模块570包括候选词获取单元、候选词互信息计算单元和同音错误词修正单元。In an exemplary embodiment, the input text acquisition module 570 includes a candidate word acquisition unit, a candidate word mutual information calculation unit, and a homophone error word correction unit.
候选词获取单元用于根据所述同音错误词所对应的拼音，从所述同音词词库中提取与所述同音错误词同音的若干候选词。The candidate word acquiring unit is configured to extract several candidate words that are homophones with the homophone error word from the homophone word database according to the pinyin corresponding to the homophone error word.
候选词互信息计算单元用于通过所述候选词逐一对所述同音错误词进行替换，并计算所述候选词对应的互信息值。The candidate word mutual information calculation unit is configured to replace the homophone error words one by one through the candidate words, and calculate the mutual information value corresponding to the candidate words.
同音错误词修正单元用于提取互信息值最高的候选词作为目标词对所述同音错误词进行替换。The homophone error word correction unit is used to extract the candidate word with the highest mutual information value as the target word to replace the homophone error word.
在一示例性的实施例中，候选词获取单元包括索引值查找子单元和拼音音节表查询子单元。In an exemplary embodiment, the candidate word obtaining unit includes an index value searching subunit and a Pinyin syllable table searching subunit.
索引值查找子单元用于根据所述同音错误词的拼音，在所述同音词词库中查询所述拼音对应的索引值，所述索引值对应于所述同音词词库中设置的拼音音节表，所述拼音音节表用于实现多个同音词语的映射。The index value searching subunit is used to query the index value corresponding to the pinyin in the homophone word library according to the pinyin of the homophone error word, and the index value corresponds to the pinyin syllable table set in the homophone word library, The pinyin syllable table is used to realize the mapping of multiple homophones.
拼音音节表查询子单元用于根据所查找到的索引值，在所述拼接音节表中查询与所述同音错误词同音的若干候选词。The pinyin syllable table query subunit is used for querying the spliced syllable table for several candidate words that are homophones with the homophone error word according to the found index value.
需要说明的是，上述实施例所提供的装置与上述实施例所提供的方法属于同一构思，其中各个模块执行操作的具体方式已经在方法实施例中进行了详细描述，此处不再赘述。It should be noted that the device provided in the foregoing embodiment and the method provided in the foregoing embodiment belong to the same concept, and the specific manners for performing operations of each module have been described in detail in the method embodiment, and will not be repeated here.
在一示例性的实施例中，本申请还提供一种人工智能面试中获取输入文本的设备，该设备包括：处理器；存储器，该存储器上存储有计算机可读指令，该计算机可读指令被处理器执行时，实现如前所述的人工智能面试中获取输入文本的方法。In an exemplary embodiment, the present application also provides a device for obtaining input text in an artificial intelligence interview. The device includes: a processor; a memory, and computer-readable instructions are stored on the memory, and the computer-readable instructions are When the processor is executed, the method for obtaining input text in the artificial intelligence interview as described above is realized.
在一示例性的实施例中，本申请还提供一种计算机可读存储介质，所述计算机可读存储介质可以是非易失性，也可以是易失性。其上存储有计算机程序，该计算机程序被处理器执行时，实现如前所述的人工智能面试中获取输入文本的方法。In an exemplary embodiment, the present application also provides a computer-readable storage medium. The computer-readable storage medium may be non-volatile or volatile. A computer program is stored thereon, and when the computer program is executed by the processor, the method for obtaining input text in an artificial intelligence interview as described above is realized.
以上所述，仅为本申请的具体实施方式，但本申请的保护范围并不局限于此，任何熟悉本技术领域的技术人员在本申请揭露的技术范围内，可轻易想到变化或替换，都应涵盖在本申请的保护范围之内。因此，本申请的保护范围应以所述权利要求的保护范围为准。The above are only specific implementations of this application, but the protection scope of this application is not limited to this. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in this application. Should be covered within the scope of protection of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

Claims (20)

  1. 一种人工智能面试中获取输入文本的方法，其中，所述方法包括：A method for obtaining input text in an artificial intelligence interview, wherein the method includes:
    在所进行的人工智能面试中，调用预设的通用词词库和通用词发音对照表对输入语音进行校正，并对校正所得语音进行语音识别获得识别文本；In the conducted artificial intelligence interview, call the preset general word database and general word pronunciation comparison table to correct the input voice, and perform voice recognition on the corrected voice to obtain the recognized text;
    通过对所述识别文本进行分词处理获得分词集合；Obtaining a word segmentation set by performing word segmentation processing on the recognized text;
    对所述分词集合中的分词，分别计算所述分词相对左侧分词和右侧分词的互信息值，且根据所得互信息值定位所述识别文本中的同音错误词，所述互信息值包括左互信息值和右互信息值；For the word segmentation in the word segmentation set, the mutual information value of the word segmentation relative to the left and right participles is calculated, and the homophone error words in the recognized text are located according to the obtained mutual information value, and the mutual information value includes Left mutual information value and right mutual information value;
    从预设的同音词词库中提取目标词对所述同音错误词进行替换，获得所述智能面试中的输入文本，所述目标词的读音与所述同音错误词相同。The target word is extracted from a preset homophone word database to replace the homophone error word to obtain the input text in the smart interview, and the pronunciation of the target word is the same as the homophone error word.
  2. 根据权利要求1所述的方法，其中，在所述通过对所述识别文本进行分词处理获得分词集合之前，所述方法还包括：The method according to claim 1, wherein before said obtaining a word segmentation set by performing word segmentation processing on the recognized text, the method further comprises:
    根据预先设置的语气词词典，查找所述识别文本中重复出现的语气词；According to a preset modal particle dictionary, search for repetitive modal particles in the recognized text;
    将所述重复出现的语气词从所述识别文本中删除。The repeated modal particles are deleted from the recognized text.
  3. 根据权利要求1所述的方法，其中，所述根据所得互信息值定位所述识别文本中的同音错误词，包括：The method according to claim 1, wherein the locating the homophone wrong words in the recognized text according to the obtained mutual information value comprises:
    根据所述分词集合中各分词对应的左互信息值和右互信息值，分别计算所述分词集合关于所述左互信息值和右互信息值的均值以及标准差；According to the left mutual information value and the right mutual information value corresponding to each word in the word segmentation set, respectively calculating the mean value and the standard deviation of the left mutual information value and the right mutual information value of the word segmentation set;
    分别对所述左互信息值和右互信息值的均值以及标准差进行差值运算，获得所述分词集合关于所述左互信息值和右互信息值的阈值；Performing a difference operation on the mean and standard deviation of the left mutual information value and the right mutual information value, respectively, to obtain the threshold value of the left mutual information value and the right mutual information value of the word segmentation set;
    获取所述分词集合中左互信息值和右互信息值小于对应阈值的分词为同音错误词。The word segmentation whose left mutual information value and right mutual information value are less than the corresponding threshold in the word segmentation set is obtained as homophone error words.
  4. 根据权利要求3所述的方法，其中，在所述获取所述分词集合中左互信息值和右互信息值小于对应阈值的分词获取为同音错误词之前，所述方法还包括：The method according to claim 3, wherein before said acquiring the word segmentation in the word segmentation set whose left mutual information value and right mutual information value are less than the corresponding threshold value is acquired as a homophone error word, the method further comprises:
    根据为所述左互信息值和右互信息值所赋予的权重，对所述分词集合关于所述左互信息值和右互信息值的阈值进行加权和运算，获得所述分词集合对应的互信息阈值；According to the weights assigned to the left mutual information value and the right mutual information value, the threshold value of the left mutual information value and the right mutual information value of the word segmentation set is weighted and operated to obtain the mutual information corresponding to the word segmentation set. Information threshold
    所述获取所述分词集合中左互信息值和右互信息值小于对应阈值的分词获取为同音错误词，包括：The acquiring the word segmentation whose left mutual information value and right mutual information value in the word segmentation set are less than a corresponding threshold value is acquired as a homophone error word, including:
    对所述分词集合中的分词，按照所述权重对所述分词对应的左互信息值和右互信息值进行加权和运算，获得所述分词的互信息值；For the word segmentation in the word segmentation set, weighting and calculating the left mutual information value and the right mutual information value corresponding to the word segmentation according to the weight to obtain the mutual information value of the word segmentation;
    获取所述互信息值小于所述互信息阈值的分词为同音错误词。The word segmentation whose mutual information value is less than the mutual information threshold is acquired as homophone error words.
  5. 根据权利要求1所述的方法，其中，所述从预设的同音词词库中提取目标词对所述同音错误词进行替换，包括：The method according to claim 1, wherein said extracting a target word from a preset homophone word library to replace said homophone error word comprises:
    根据所述同音错误词所对应的拼音，从所述同音词词库中提取与所述同音错误词同音的若干候选词；According to the pinyin corresponding to the homophone error word, extract several candidate words that are homophones with the homophone error word from the homophone word library;
    通过所述候选词逐一对所述同音错误词进行替换，并计算所述候选词对应的互信息值；Replace the homophone error words one by one through the candidate words, and calculate the mutual information value corresponding to the candidate words;
    提取互信息值最高的候选词作为目标词对所述同音错误词进行替换。The candidate word with the highest mutual information value is extracted as the target word to replace the homophone error word.
  6. 根据权利要求5所述的方法，其中，所述根据所述同音错误词所对应的拼音，从所述同音词词库中提取与所述同音错误词同音的若干候选词，包括：The method according to claim 5, wherein the extracting from the homophone word library according to the pinyin corresponding to the homophone error word, a number of candidate words that are homophones with the homophone error word comprises:
    根据所述同音错误词的拼音，在所述同音词词库中查询所述拼音对应的索引值，所述索引值对应于所述同音词词库中设置的拼音音节表，所述拼音音节表用于实现多个同音词语的映射；According to the pinyin of the homophone error word, the index value corresponding to the pinyin is queried in the homophone word database, and the index value corresponds to the pinyin syllable table set in the homophone word database, and the pinyin syllable table is used for Realize the mapping of multiple homophones;
    根据所查找到的索引值，在所述拼接音节表中查询与所述同音错误词同音的若干候选词。According to the searched index value, the concatenated syllable table is searched for several candidate words that are homophones with the homophone of the wrong word.
  7. 根据权利要求1所述的方法，其中，所述通过对所述识别文本进行分词处理获得分词集合，包括：The method according to claim 1, wherein the obtaining a word segmentation set by performing word segmentation processing on the recognized text comprises:
    采用分词算法对所述识别文本进行分词处理，其中，所述分词算法包括正向最大匹配分词算法、逆向最大匹配分词算法、双向最大匹配分词算法。A word segmentation algorithm is used to perform word segmentation processing on the recognized text, where the word segmentation algorithm includes a forward maximum matching word segmentation algorithm, a reverse maximum matching word segmentation algorithm, and a two-way maximum matching word segmentation algorithm.
  8. 一种人工智能面试中获取输入文本的装置，其中，所述装置包括：A device for acquiring input text in an artificial intelligence interview, wherein the device includes:
    语音识别模块，用于在所进行的人工智能面试中，调用预设的通用词词库和通用词发音对照表对输入语音进行校正，并对校正所得语音进行语音识别获得识别文本；The voice recognition module is used to call the preset general word database and general word pronunciation comparison table to correct the input voice in the artificial intelligence interview, and perform voice recognition on the corrected voice to obtain the recognized text;
    分词处理模块，用于通过对所述识别文本进行分词处理获得分词集合；A word segmentation processing module, configured to obtain a word segmentation set by performing word segmentation processing on the recognized text;
    同音错误词定位模块，用于对所述分词集合中的分词，分别计算所述分词相对左侧分词和右侧分词的互信息值，且根据所得互信息值定位所述识别文本中的同音错误词，所述互信息值包括左互信息值和右互信息值；The homophone error word positioning module is used to calculate the mutual information value of the word segmentation relative to the left and right participles for the word segmentation in the word segmentation set, and locate the homophone error in the recognized text according to the obtained mutual information value Word, the mutual information value includes a left mutual information value and a right mutual information value;
    输入文本获取模块，用于从预设的同音词词库中提取目标词对所述同音错误词进行替换，获得所述智能面试中的输入文本，所述目标词的读音与所述同音错误词相同。The input text acquisition module is used to extract a target word from a preset homophone word library to replace the homophone error word to obtain the input text in the smart interview, and the pronunciation of the target word is the same as the homophone error word .
  9. 一种人工智能面试中获取输入文本的设备，其中，包括存储器和处理器，所述处理器、和所述存储器相互连接，其中，所述存储器用于存储计算机程序，所述计算机程序包括程序指令，所述处理器用于执行所述存储器的所述程序指令，其中：A device for acquiring input text in an artificial intelligence interview, including a memory and a processor, the processor and the memory are connected to each other, wherein the memory is used to store a computer program, and the computer program includes program instructions , The processor is configured to execute the program instructions of the memory, wherein:
    在所进行的人工智能面试中，调用预设的通用词词库和通用词发音对照表对输入语音进行校正，并对校正所得语音进行语音识别获得识别文本；In the conducted artificial intelligence interview, call the preset general word database and general word pronunciation comparison table to correct the input voice, and perform voice recognition on the corrected voice to obtain the recognized text;
    通过对所述识别文本进行分词处理获得分词集合；Obtaining a word segmentation set by performing word segmentation processing on the recognized text;
    对所述分词集合中的分词，分别计算所述分词相对左侧分词和右侧分词的互信息值，且根据所得互信息值定位所述识别文本中的同音错误词，所述互信息值包括左互信息值和右互信息值；For the word segmentation in the word segmentation set, the mutual information value of the word segmentation relative to the left and right participles is calculated, and the homophone error words in the recognized text are located according to the obtained mutual information value, and the mutual information value includes Left mutual information value and right mutual information value;
    从预设的同音词词库中提取目标词对所述同音错误词进行替换，获得所述智能面试中的输入文本，所述目标词的读音与所述同音错误词相同。The target word is extracted from a preset homophone word database to replace the homophone error word to obtain the input text in the smart interview, and the pronunciation of the target word is the same as the homophone error word.
  10. 根据权利要求9所述的设备，其中，所述处理器用于：The device according to claim 9, wherein the processor is configured to:
    根据预先设置的语气词词典，查找所述识别文本中重复出现的语气词；According to a preset modal particle dictionary, search for repetitive modal particles in the recognized text;
    将所述重复出现的语气词从所述识别文本中删除。The repeated modal particles are deleted from the recognized text.
  11. 根据权利要求9所述的设备，其中，所述处理器用于：The device according to claim 9, wherein the processor is configured to:
    根据所述分词集合中各分词对应的左互信息值和右互信息值，分别计算所述分词集合关于所述左互信息值和右互信息值的均值以及标准差；According to the left mutual information value and the right mutual information value corresponding to each word in the word segmentation set, respectively calculating the mean value and the standard deviation of the left mutual information value and the right mutual information value of the word segmentation set;
    分别对所述左互信息值和右互信息值的均值以及标准差进行差值运算，获得所述分词集合关于所述左互信息值和右互信息值的阈值；Performing a difference operation on the mean value and standard deviation of the left mutual information value and the right mutual information value respectively to obtain the threshold value of the left mutual information value and the right mutual information value of the word segmentation set;
    获取所述分词集合中左互信息值和右互信息值小于对应阈值的分词为同音错误词。The word segmentation whose left mutual information value and right mutual information value are less than the corresponding threshold in the word segmentation set is obtained as homophone error words.
  12. 根据权利要求11所述的设备，其中，所述处理器用于：The device according to claim 11, wherein the processor is configured to:
    根据为所述左互信息值和右互信息值所赋予的权重，对所述分词集合关于所述左互信息值和右互信息值的阈值进行加权和运算，获得所述分词集合对应的互信息阈值；According to the weights assigned to the left mutual information value and the right mutual information value, the threshold value of the left mutual information value and the right mutual information value of the word segmentation set is weighted and operated to obtain the mutual information corresponding to the word segmentation set. Information threshold
    所述处理器用于：The processor is used for:
    对所述分词集合中的分词，按照所述权重对所述分词对应的左互信息值和右互信息值进行加权和运算，获得所述分词的互信息值；For the word segmentation in the word segmentation set, weighting and calculating the left mutual information value and the right mutual information value corresponding to the word segmentation according to the weight to obtain the mutual information value of the word segmentation;
    获取所述互信息值小于所述互信息阈值的分词为同音错误词。The word segmentation whose mutual information value is less than the mutual information threshold is acquired as homophone error words.
  13. 根据权利要求9所述的设备，其中，所述处理器用于：The device according to claim 9, wherein the processor is configured to:
    根据所述同音错误词所对应的拼音，从所述同音词词库中提取与所述同音错误词同音的若干候选词；According to the pinyin corresponding to the homophone error word, extract several candidate words that are homophones with the homophone error word from the homophone word library;
    通过所述候选词逐一对所述同音错误词进行替换，并计算所述候选词对应的互信息值；Replace the homophone error words one by one through the candidate words, and calculate the mutual information value corresponding to the candidate words;
    提取互信息值最高的候选词作为目标词对所述同音错误词进行替换。The candidate word with the highest mutual information value is extracted as the target word to replace the homophone error word.
  14. 根据权利要求13所述的设备，其中，所述处理器还用于：The device according to claim 13, wherein the processor is further configured to:
    根据所述同音错误词的拼音，在所述同音词词库中查询所述拼音对应的索引值，所述索引值对应于所述同音词词库中设置的拼音音节表，所述拼音音节表用于实现多个同音词语的映射；According to the pinyin of the homophone error word, the index value corresponding to the pinyin is queried in the homophone word database, and the index value corresponds to the pinyin syllable table set in the homophone word database, and the pinyin syllable table is used for Realize the mapping of multiple homophones;
    根据所查找到的索引值，在所述拼接音节表中查询与所述同音错误词同音的若干候选词。According to the searched index value, the concatenated syllable table is searched for several candidate words that are homophones with the homophone of the wrong word.
  15. 根据权利要求9所述的设备，其中，所述处理器还用于：The device according to claim 9, wherein the processor is further configured to:
    采用分词算法对所述识别文本进行分词处理，其中，所述分词算法包括正向最大匹配分词算法、逆向最大匹配分词算法、双向最大匹配分词算法。A word segmentation algorithm is used to perform word segmentation processing on the recognized text, where the word segmentation algorithm includes a forward maximum matching word segmentation algorithm, a reverse maximum matching word segmentation algorithm, and a two-way maximum matching word segmentation algorithm.
  16. 一种计算机可读存储介质，其中，所述计算机可读存储介质存储有计算机程序，所述计算机程序包括程序指令，被处理器执行时，用于实现以下步骤：A computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, the computer program includes program instructions, when executed by a processor, is used to implement the following steps:
    在所进行的人工智能面试中，调用预设的通用词词库和通用词发音对照表对输入语音进行校正，并对校正所得语音进行语音识别获得识别文本；In the conducted artificial intelligence interview, call the preset general word database and general word pronunciation comparison table to correct the input voice, and perform voice recognition on the corrected voice to obtain the recognized text;
    通过对所述识别文本进行分词处理获得分词集合；Obtaining a word segmentation set by performing word segmentation processing on the recognized text;
    对所述分词集合中的分词，分别计算所述分词相对左侧分词和右侧分词的互信息值，且根据所得互信息值定位所述识别文本中的同音错误词，所述互信息值包括左互信息值和右互信息值；For the word segmentation in the word segmentation set, the mutual information value of the word segmentation relative to the left and right participles is calculated, and the homophone error words in the recognized text are located according to the obtained mutual information value, and the mutual information value includes Left mutual information value and right mutual information value;
    从预设的同音词词库中提取目标词对所述同音错误词进行替换，获得所述智能面试中的输入文本，所述目标词的读音与所述同音错误词相同。The target word is extracted from a preset homophone word database to replace the homophone error word to obtain the input text in the smart interview, and the pronunciation of the target word is the same as the homophone error word.
  17. 根据权利要求16所述的计算机可读存储介质，其中，所述程序指令被处理器执行时，还用于实现以下步骤：The computer-readable storage medium according to claim 16, wherein when the program instructions are executed by the processor, they are further used to implement the following steps:
    根据预先设置的语气词词典，查找所述识别文本中重复出现的语气词；According to a preset modal particle dictionary, search for repetitive modal particles in the recognized text;
    将所述重复出现的语气词从所述识别文本中删除。The repeated modal particles are deleted from the recognized text.
  18. 根据权利要求16所述的计算机可读存储介质，其中，所述程序指令被处理器执行时，还用于实现以下步骤：The computer-readable storage medium according to claim 16, wherein when the program instructions are executed by the processor, they are further used to implement the following steps:
    根据所述分词集合中各分词对应的左互信息值和右互信息值，分别计算所述分词集合关于所述左互信息值和右互信息值的均值以及标准差；According to the left mutual information value and the right mutual information value corresponding to each word in the word segmentation set, respectively calculating the mean value and the standard deviation of the left mutual information value and the right mutual information value of the word segmentation set;
    分别对所述左互信息值和右互信息值的均值以及标准差进行差值运算，获得所述分词集合关于所述左互信息值和右互信息值的阈值；Performing a difference operation on the mean value and standard deviation of the left mutual information value and the right mutual information value respectively to obtain the threshold value of the left mutual information value and the right mutual information value of the word segmentation set;
    获取所述分词集合中左互信息值和右互信息值小于对应阈值的分词为同音错误词。The word segmentation whose left mutual information value and right mutual information value are less than the corresponding threshold in the word segmentation set is obtained as homophone error words.
  19. 根据权利要求18所述的计算机可读存储介质，其中，所述程序指令被处理器执行时，还用于实现以下步骤：The computer-readable storage medium according to claim 18, wherein, when the program instructions are executed by the processor, they are further used to implement the following steps:
    根据为所述左互信息值和右互信息值所赋予的权重，对所述分词集合关于所述左互信息值和右互信息值的阈值进行加权和运算，获得所述分词集合对应的互信息阈值；According to the weights assigned to the left mutual information value and the right mutual information value, the threshold value of the left mutual information value and the right mutual information value of the word segmentation set is weighted and operated to obtain the mutual information corresponding to the word segmentation set. Information threshold
    所述获取所述分词集合中左互信息值和右互信息值小于对应阈值的分词获取为同音错误词，包括：The acquiring the word segmentation whose left mutual information value and right mutual information value in the word segmentation set are less than a corresponding threshold value is acquired as a homophone error word, including:
    对所述分词集合中的分词，按照所述权重对所述分词对应的左互信息值和右互信息值进行加权和运算，获得所述分词的互信息值；For the word segmentation in the word segmentation set, weighting and calculating the left mutual information value and the right mutual information value corresponding to the word segmentation according to the weight to obtain the mutual information value of the word segmentation;
    获取所述互信息值小于所述互信息阈值的分词为同音错误词。The word segmentation whose mutual information value is less than the mutual information threshold is acquired as homophone error words.
  20. 根据权利要求16所述的计算机可读存储介质，其中，所述程序指令被处理器执行时，还用于实现以下步骤：The computer-readable storage medium according to claim 16, wherein when the program instructions are executed by the processor, they are further used to implement the following steps:
    根据所述同音错误词所对应的拼音，从所述同音词词库中提取与所述同音错误词同音的若干候选词；According to the pinyin corresponding to the homophone error word, extract several candidate words that are homophones with the homophone error word from the homophone word library;
    通过所述候选词逐一对所述同音错误词进行替换，并计算所述候选词对应的互信息值；Replace the homophone error words one by one through the candidate words, and calculate the mutual information value corresponding to the candidate words;
    提取互信息值最高的候选词作为目标词对所述同音错误词进行替换。The candidate word with the highest mutual information value is extracted as the target word to replace the homophone error word.
PCT/CN2020/093597 2019-09-17 2020-05-30 Method for obtaining input text in artificial intelligence interview, and related apparatus WO2021051877A1 (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910877092.4A CN110717021B (en) 2019-09-17 2019-09-17 Input text acquisition and related device in artificial intelligence interview
CN201910877092.4 2019-09-17

Publications (1)

Publication Number Publication Date
WO2021051877A1 true WO2021051877A1 (en) 2021-03-25

Family

ID=69209876

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/093597 WO2021051877A1 (en) 2019-09-17 2020-05-30 Method for obtaining input text in artificial intelligence interview, and related apparatus

Country Status (2)

Country Link
CN (1) CN110717021B (en)
WO (1) WO2021051877A1 (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110717021B (en) * 2019-09-17 2023-08-29 平安科技（深圳）有限公司 Input text acquisition and related device in artificial intelligence interview
CN111694936B (en) * 2020-04-26 2023-06-06 平安科技（深圳）有限公司 Method, device, computer equipment and storage medium for identification of AI intelligent interview
CN112164403A (en) * 2020-09-27 2021-01-01 江苏四象软件有限公司 Natural language processing system based on artificial intelligence
CN112684913B (en) * 2020-12-30 2023-07-14 维沃移动通信有限公司 Information correction method and device and electronic equipment

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20070038452A1 (en) * 2005-08-12 2007-02-15 Avaya Technology Corp. Tonal correction of speech
CN105045778A (en) * 2015-06-24 2015-11-11 江苏科技大学 Chinese homonym error auto-proofreading method
CN106528532A (en) * 2016-11-07 2017-03-22 上海智臻智能网络科技股份有限公司 Text error correction method and device and terminal
CN107451121A (en) * 2017-08-03 2017-12-08 京东方科技集团股份有限公司 A kind of audio recognition method and its device
CN107608963A (en) * 2017-09-12 2018-01-19 马上消费金融股份有限公司 A kind of Chinese error correction based on mutual information, device, equipment and storage medium
CN110210028A (en) * 2019-05-30 2019-09-06 杭州远传新业科技有限公司 For domain feature words extracting method, device, equipment and the medium of speech translation text
CN110717021A (en) * 2019-09-17 2020-01-21 平安科技（深圳）有限公司 Input text and related device for obtaining artificial intelligence interview

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR100639931B1 (en) * 2004-11-24 2006-11-01 한국전자통신연구원 Recognition error correction apparatus for interactive voice recognition system and method therefof
CN106534548B (en) * 2016-11-17 2020-06-12 科大讯飞股份有限公司 Voice error correction method and device
US10382138B2 (en) * 2017-03-06 2019-08-13 Nec Corporation Constellation optimization based on generalized mutual information over a nonlinear optical channel
CN108874921A (en) * 2018-05-30 2018-11-23 广州杰赛科技股份有限公司 Extract method, apparatus, terminal device and the storage medium of text feature word
CN109800414B (en) * 2018-12-13 2023-04-18 科大讯飞股份有限公司 Method and system for recommending language correction

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20070038452A1 (en) * 2005-08-12 2007-02-15 Avaya Technology Corp. Tonal correction of speech
CN105045778A (en) * 2015-06-24 2015-11-11 江苏科技大学 Chinese homonym error auto-proofreading method
CN106528532A (en) * 2016-11-07 2017-03-22 上海智臻智能网络科技股份有限公司 Text error correction method and device and terminal
CN107451121A (en) * 2017-08-03 2017-12-08 京东方科技集团股份有限公司 A kind of audio recognition method and its device
CN107608963A (en) * 2017-09-12 2018-01-19 马上消费金融股份有限公司 A kind of Chinese error correction based on mutual information, device, equipment and storage medium
CN110210028A (en) * 2019-05-30 2019-09-06 杭州远传新业科技有限公司 For domain feature words extracting method, device, equipment and the medium of speech translation text
CN110717021A (en) * 2019-09-17 2020-01-21 平安科技（深圳）有限公司 Input text and related device for obtaining artificial intelligence interview

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
FENG XIN: "Phone-level Based Mispronunciation Automatic Detection and Its Application", CHINESE MASTER'S THESES FULL-TEXT DATABASE, 1 January 2009 (2009-01-01), pages 1 - 63, XP055792589, ISSN: 1674-0246 *
YU TAO: "Study on Automatic Error-Correction in Chinese Speech-to-Text Recognition", MASTER THESIS, 1 February 2003 (2003-02-01), CN, pages 1 - 52, XP009529672 *

Also Published As

Publication number Publication date
CN110717021A (en) 2020-01-21
CN110717021B (en) 2023-08-29

Similar Documents

Publication Publication Date Title
CN110148416B (en) Speech recognition method, device, equipment and storage medium
US10176804B2 (en) Analyzing textual data
US11182568B2 (en) Sentence evaluation apparatus and sentence evaluation method
WO2021051877A1 (en) Method for obtaining input text in artificial intelligence interview, and related apparatus
CN110163181B (en) Sign language identification method and device
KR20230040951A (en) Speech recognition method, apparatus and device, and storage medium
CN114580382A (en) Text error correction method and device
US20190095447A1 (en) Method, apparatus, device and storage medium for establishing error correction model based on error correction platform
CN111144102B (en) Method and device for identifying entity in statement and electronic equipment
WO2021208444A1 (en) Method and apparatus for automatically generating electronic cases, a device, and a storage medium
CN113094478B (en) Expression reply method, device, equipment and storage medium
CN112632248A (en) Question answering method, device, computer equipment and storage medium
CN111723870A (en) Data set acquisition method, device, equipment and medium based on artificial intelligence
CN111126084B (en) Data processing method, device, electronic equipment and storage medium
JP2021081713A (en) Method, device, apparatus, and media for processing voice signal
US20230065965A1 (en) Text processing method and apparatus
WO2021217866A1 (en) Method and apparatus for ai interview recognition, computer device and storage medium
CN110413750B (en) Method and device for recalling standard questions according to user questions
CN113536784A (en) Text processing method and device, computer equipment and storage medium
CN115858776B (en) Variant text classification recognition method, system, storage medium and electronic equipment
CN110991155B (en) Text correction method, device and medium
CN111401069A (en) Intention recognition method and intention recognition device for conversation text and terminal
WO2021129410A1 (en) Method and device for text processing
CN114528851A (en) Reply statement determination method and device, electronic equipment and storage medium
CN114186020A (en) Semantic association method

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20865803

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20865803

Country of ref document: EP

Kind code of ref document: A1