CN116227504A - Communication method, system, equipment and storage medium for simultaneous translation - Google Patents

Communication method, system, equipment and storage medium for simultaneous translation Download PDF

Info

Publication number
CN116227504A
CN116227504A CN202310101682.4A CN202310101682A CN116227504A CN 116227504 A CN116227504 A CN 116227504A CN 202310101682 A CN202310101682 A CN 202310101682A CN 116227504 A CN116227504 A CN 116227504A
Authority
CN
China
Prior art keywords
message
information
language
translation
initial
Prior art date
Legal status (The legal status 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 status listed.)
Granted
Application number
CN202310101682.4A
Other languages
Chinese (zh)
Other versions
CN116227504B (en
Inventor
刘春梅
陈晴
陆澄霖
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guangzhou Digital Future Culture Technology Co ltd
Original Assignee
Guangzhou Digital Future Culture Technology Co ltd
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 Guangzhou Digital Future Culture Technology Co ltd filed Critical Guangzhou Digital Future Culture Technology Co ltd
Priority to CN202310101682.4A priority Critical patent/CN116227504B/en
Publication of CN116227504A publication Critical patent/CN116227504A/en
Application granted granted Critical
Publication of CN116227504B publication Critical patent/CN116227504B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/40—Processing or translation of natural language
    • G06F40/58—Use of machine translation, e.g. for multi-lingual retrieval, for server-side translation for client devices or for real-time translation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/20—Natural language analysis
    • G06F40/263—Language identification
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L13/00—Speech synthesis; Text to speech systems
    • G10L13/08—Text analysis or generation of parameters for speech synthesis out of text, e.g. grapheme to phoneme translation, prosody generation or stress or intonation determination
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L15/00—Speech recognition
    • G10L15/26—Speech to text systems
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Computational Linguistics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • General Health & Medical Sciences (AREA)
  • Human Computer Interaction (AREA)
  • Acoustics & Sound (AREA)
  • Multimedia (AREA)
  • Machine Translation (AREA)
  • Information Transfer Between Computers (AREA)

Abstract

The application relates to the technical field of instant translation and discloses a concurrent translation communication method, a system, equipment and a storage medium, wherein the concurrent translation communication method comprises the following steps: receiving language information of a first client, inputting the language information into an API, creating a corresponding initial message and sending the initial message to an information translation model; sequentially carrying out message transcription, initial language detection, target language detection and message translation on the initial message based on the communication scene to generate a target message; backing up the target message to a database, and sending the target message to a second client through an API; acquiring feedback information of the client, and generating scene correction information based on the feedback information to update an information translation model; the method and the device have the effect of improving language translation efficiency in network communication.

Description

Communication method, system, equipment and storage medium for simultaneous translation
Technical Field
The present application relates to the field of instant translation technologies, and in particular, to a method, a system, an apparatus, and a storage medium for concurrent translation communication.
Background
With the development of economic globalization and nationwide network communication technology, the more common the language and text communication among users in different languages is, the network communication greatly improves the communication efficiency among people, however, the language barrier becomes a new constraint factor for the communication efficiency among users in different languages.
At present, people commonly use software or products with TTS (text to speech), STT (speech to text), ASR (speech recognition technology) or Translate (translation technology) and other technologies to realize conversion between speech and text and multiple languages so as to improve the communication efficiency between people.
Aiming at the related technology, the inventor considers that the existing language translation software or equipment has the problems of low real-time and single technical capabilities.
Disclosure of Invention
In order to improve the efficiency of language translation in network communication, the application provides a communication method, a system, equipment and a storage medium for concurrent translation.
The first technical scheme adopted by the invention of the application is as follows:
a method of concurrent translation communication, comprising:
receiving language information of a first client, inputting the language information into an API, creating a corresponding initial message and sending the initial message to an information translation model;
sequentially carrying out message transcription, initial language detection, target language detection and message translation on the initial message based on the communication scene to generate a target message;
backing up the target message to a database, and sending the target message to a second client through an API;
and acquiring feedback information of the client, and generating scene correction information based on the feedback information to update the information translation model.
By adopting the technical scheme, the language message sent by the first client is received and input into the API so as to be input into the concurrent translation program, the corresponding initial message is created based on the received language message, and the initial message is sent into the information translation model so as to be translated subsequently; based on the current language communication scene, the initial message is subjected to text audio transcription processing, initial language detection processing, target language detection processing and message translation processing in sequence, and then the function of translating the initial message is realized, so that the target message is generated; backing up the target message to a database, facilitating the user to inquire the history information, analyzing high-frequency words and sentences in each scene, and sending the target message to a second client through an API (application program interface) so as to timely send the translated information to the target user; and acquiring feedback information of each client so as to judge the translation accuracy of the current information translation model, and generating scene correction information according to the information fed back by the user so as to improve the translation accuracy of the information translation model in a specific scene, thereby improving the language translation efficiency in network communication.
In a preferred example, the present application: the method comprises the steps of sequentially carrying out message transcription, initial language detection, target language detection and message translation on the initial message based on the communication scene, and generating a target message, wherein the steps comprise:
receiving an initial message, and judging a corresponding message type according to attribute information of the initial message;
if the initial message is the first type of information, the initial message is sent to a message queue;
if the initial message is the second type information, the initial message is transcribed into the first type information through the volcanic engine based on the communication scene, the transcribed initial message is returned to the first client, and the initial message is sent to the message queue.
By adopting the technical scheme, the initial message is received, the message type corresponding to the initial message is judged from the attribute information of the initial message, the subsequent text audio transcription processing of the initial message is facilitated, the message type comprises first type information and second type information, and the first type information is the target message type of the text audio transcription processing; if the initial message is the first type of information, the initial message is directly sent to a message queue to wait for further processing; if the initial message is the second type information, the initial message is transcribed into the first type information through the volcanic engine based on the characteristics of the communication scene so as to conduct type conversion on the initial messages of different types, so as to adapt to an algorithm of an information translation model, the initial message after text audio transcription is returned to the first client, a user can conveniently judge whether a text audio transcription result is accurate or not, errors can be found timely, feedback can be achieved timely, and the initial message after text audio transcription is sent to a message queue to wait for further processing.
In a preferred example, the present application: the method comprises the steps of sequentially carrying out message transcription, initial language detection, target language detection and message translation on the initial message, and generating a target message, and further comprises the following steps:
judging communication scene information, and selecting a corresponding scene word stock;
the method comprises the steps of obtaining language type information of a first client, carrying out language detection processing on an initial message, generating a message to be translated, and determining the initial language information;
obtaining language type information of a second client and determining target language information;
and carrying out message translation on the message to be translated based on the scene word stock, the initial language information and the target language information to generate the target message.
By adopting the technical scheme, the current scene of language communication between the first client and the second client personnel is judged, communication scene information is generated, and a corresponding scene word stock is selected, so that message translation can be conveniently and pertinently carried out according to the scene word stock, and the accuracy of the message translation is improved; because words with the same/similar pronunciation and the same/similar characters and completely different meanings possibly exist in different languages, language type information of the first client is acquired, and language detection processing is carried out on the initial message by taking the language type information of the first client as a reference, so that the accuracy of language detection of the initial message is improved; generating a message to be translated, determining initial language information, and facilitating subsequent further message translation; obtaining language type information of a second client to determine target language information as target language of message translation; and carrying out message translation on the message to be translated based on the scene word stock, the initial language information and the target language information to generate a target message so as to complete the message translation work and facilitate the subsequent transmission of the target message to the second client.
In a preferred example, the present application: based on the scene word stock, the initial language information and the target language information, carrying out message translation on the message to be translated, and generating the target message, wherein the step of generating the target message comprises the following steps:
according to the initial language information, selecting a corresponding natural language algorithm to perform semantic recognition on the message to be translated, and generating a semantic recognition result;
carrying out relevance assessment on the semantic recognition result of the message to be translated and the semantic recognition result of the relevant information through a natural language algorithm to generate a relevance assessment result;
and if the relevance evaluation result is qualified, carrying out message translation on the message to be translated based on the semantic recognition result and the target language information, and generating a target message.
By adopting the technical scheme, as different countries/regions and different languages have different characteristics on the expression of the same thing, the corresponding natural language processing algorithm is selected to carry out semantic recognition on the message to be translated according to the initial language information, and a semantic recognition result is generated so as to improve the accuracy of the semantic recognition; carrying out relevance evaluation on the semantic recognition result of the message to be translated and the semantic recognition result of the relevant information through a natural language algorithm so as to judge whether the semantic recognition result of the message to be translated is suitable for the current communication scene according to the semantic of the relevant information; and if the relevance evaluation result is qualified, translating the message to be translated into a target language according to the semantic recognition result so as to generate the target message.
In a preferred example, the present application: based on the scene word stock, the initial language information and the target language information, carrying out message translation on the message to be translated, and generating the target message, wherein the step of generating the target message comprises the following steps:
if the relevance evaluation result is unqualified, determining the violations and words in the message to be translated through a natural language algorithm;
matching the replacement words from the scene word stock based on the violations and the words, and regenerating semantic recognition results and relevance evaluation results;
and carrying out message translation on the message to be translated based on the semantic recognition result and the target language information corresponding to the relevance evaluation result with the highest relevance, and generating the target message.
By adopting the technical scheme, if the relevance evaluation result is unqualified, the violating words which cause the unqualified relevance evaluation result are determined from the message to be translated through a natural language algorithm, so that the semantic recognition structure of the message to be translated can be corrected conveniently for the violating words; matching corresponding replacement words from a scene word stock based on the violating words, and carrying out semantic recognition and relevance evaluation again on the message to be translated after the replacement words are replaced, so as to generate new semantic recognition results and relevance evaluation results, and comparing relevance evaluation results corresponding to the semantic recognition results before and after the replacement words are replaced; the semantic recognition result corresponding to the correlation evaluation result with the highest correlation is determined, and the message to be translated is translated based on the semantic recognition result and the target language information, so that the target message is generated, and the consistency of the semantic of the target message and the semantic actually expected to be expressed by the first client user is improved.
In a preferred example, the present application: the step of backing up the target message in the database and sending the target message to the second client through the API includes:
uploading the target message to a MySQL database for storage, wherein the MySQL database comprises a history message library and a scene message feature library;
and sending the stored target message to the API, and sending the target message to the second client through the API.
By adopting the technical scheme, the target message generated after the message translation is uploaded to a MySQL database for storage, wherein the MySQL database comprises a history message library and a scene message feature library, the history message library is used for storing history messages, so that a sender or a receiver of language messages can conveniently check message records, and the scene message feature library is used for recording the characteristics of voice communication in the scene, so that high-frequency words and special words in the scene can be conveniently obtained, and the scene word library can be updated; and sending the stored target message to an API, and sending the target message to a second client through the API, so that the user receives the language message sent by the first client and translated.
In a preferred example, the present application: after the step of matching the replacement word from the scene word stock based on the violation and the word and regenerating the semantic recognition result and the relevance evaluation result, the method comprises the following steps:
Acquiring the number of times of relevance evaluation to determine evaluation number information, and acquiring a relevance evaluation result;
when the evaluation frequency information is larger than a preset evaluation frequency threshold value, if a qualified relevance evaluation result is not obtained, generating a violation and search formula based on the violation word and communication scene information;
and inputting the violations and the search terms into a search engine, acquiring search meaning information corresponding to the violations and the words, and regenerating a semantic recognition result based on the search meaning information.
By adopting the technical scheme, after matching and replacing the replacement word from the scene word stock based on the violating word of the message to be translated, carrying out relevance evaluation again, obtaining the times of relevance evaluation to generate evaluation times information, and obtaining the result of each relevance evaluation; when the evaluation frequency information is larger than a preset evaluation frequency threshold value, judging whether the result of each relevance evaluation is qualified, and if the qualified relevance evaluation result is not obtained, generating a violation and search formula based on the violation word and communication scene information; the violations and the search formula are input into a search engine, so that meanings corresponding to the violations and the words are obtained from the Internet based on the violations and the search formula, search meaning information is generated, and a semantic recognition result is generated again according to the search meaning information, so that when the current semantics of the violations and the words cannot be successfully obtained from a scene word stock, the semantics of the violations and the words in the current scene are searched from the Internet by means of the search engine, and the message translation work of the message to be translated is successfully executed subsequently.
The second object of the present application is achieved by the following technical scheme:
a concurrent translation communication system, comprising:
the initial message creation module is used for receiving the language message of the first client and inputting the language message into the API, creating a corresponding initial message and sending the initial message to the information translation model;
the message translation module is used for sequentially carrying out message transcription, initial language detection, target language detection and message translation on the initial message based on the communication scene to generate a target message;
the target message sending module is used for backing up the target message into the database and sending the target message to the second client through the API;
and the scene correction module is used for acquiring feedback information of the client, and generating scene correction information based on the feedback information so as to update the information translation model.
By adopting the technical scheme, the language message sent by the first client is received and input into the API so as to be input into the concurrent translation program, the corresponding initial message is created based on the received language message, and the initial message is sent into the information translation model so as to be translated subsequently; based on the current language communication scene, the initial message is subjected to text audio transcription processing, initial language detection processing, target language detection processing and message translation processing in sequence, and then the function of translating the initial message is realized, so that the target message is generated; backing up the target message to a database, facilitating the user to inquire the history information, analyzing high-frequency words and sentences in each scene, and sending the target message to a second client through an API (application program interface) so as to timely send the translated information to the target user; and acquiring feedback information of each client so as to judge the translation accuracy of the current information translation model, and generating scene correction information according to the information fed back by the user so as to improve the translation accuracy of the information translation model in a specific scene, thereby improving the language translation efficiency in network communication.
In a preferred example, the present application: the voice input module comprises a microphone and a high-definition acquisition module, and a recording noise reduction sub-module and an echo cancellation sub-module are arranged in the high-definition acquisition module.
By adopting the technical scheme, the communication system for simultaneous interpretation comprises an electronic terminal for acquiring audio language information and receiving audio voice information of other clients, wherein the electronic terminal comprises a voice input module, the voice input module comprises a microphone for acquiring the audio language information of a user and a high-definition acquisition module for processing the acquired audio language information, and the high-definition acquisition module sends the processed audio language information to a module for executing a communication method for simultaneous interpretation; the high-definition acquisition module comprises a recording noise reduction sub-module and an echo cancellation sub-module, wherein the recording noise reduction sub-module is used for carrying out noise reduction processing on audio language information acquired by the microphone, and the echo cancellation sub-module is used for carrying out echo cancellation processing when the microphone acquires the audio language information so as to improve the recording acquisition capacity of the electronic terminal and further improve the accuracy of subsequent functional modules such as semantic understanding, language translation and the like.
The third object of the present application is achieved by the following technical scheme:
a computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, the processor implementing the steps of the communication method of concurrent translation described above when the computer program is executed.
The fourth object of the present application is achieved by the following technical scheme:
a computer readable storage medium storing a computer program which when executed by a processor performs the steps of the communication method of concurrent translation described above.
In summary, the present application includes at least one of the following beneficial technical effects:
1. receiving a language message sent by a first client and inputting the language message into an API (application program interface) so as to input the language message into a concurrent translation program, creating a corresponding initial message based on the received language message, and sending the initial message into an information translation model so as to translate the initial message subsequently; based on the current language communication scene, the initial message is subjected to text audio transcription processing, initial language detection processing, target language detection processing and message translation processing in sequence, and then the function of translating the initial message is realized, so that the target message is generated; backing up the target message to a database, facilitating the user to inquire the history information, analyzing high-frequency words and sentences in each scene, and sending the target message to a second client through an API (application program interface) so as to timely send the translated information to the target user; and acquiring feedback information of each client so as to judge the translation accuracy of the current information translation model, and generating scene correction information according to the information fed back by the user so as to improve the translation accuracy of the information translation model in a specific scene, thereby improving the language translation efficiency in network communication.
2. Because different countries/regions and different languages have different characteristics on the expression of the same thing, according to the initial language information, a corresponding natural language processing algorithm is selected to carry out semantic recognition on the message to be translated, and a semantic recognition result is generated so as to improve the accuracy of the semantic recognition; carrying out relevance evaluation on the semantic recognition result of the initial message and the semantic recognition result of the relevant information through a natural language algorithm so as to judge whether the semantic recognition result of the initial message is suitable for the current communication scene according to the semantic of the relevant information; and if the relevance evaluation result is qualified, translating the message to be translated into a target language according to the semantic recognition result so as to generate the target message.
3. If the relevance evaluation result is unqualified, determining violations and words which cause the unqualified relevance evaluation result from the message to be translated through a natural language algorithm, and facilitating the subsequent correction of the semantic recognition structure of the message to be translated aiming at the violations and words; matching corresponding replacement words from a scene word stock based on the violating words, and carrying out semantic recognition and relevance evaluation again on the message to be translated after the replacement words are replaced, so as to generate new semantic recognition results and relevance evaluation results, and comparing relevance evaluation results corresponding to the semantic recognition results before and after the replacement words are replaced; the semantic recognition result corresponding to the correlation evaluation result with the highest correlation is determined, and the message to be translated is translated based on the semantic recognition result and the target language information, so that the target message is generated, and the consistency of the semantic of the target message and the semantic actually expected to be expressed by the first client user is improved.
Drawings
FIG. 1 is a flow chart of a method of concurrent translation communication according to an embodiment of the present application.
Fig. 2 is a flowchart of step S20 in the communication method of the concurrent translation of the present application.
Fig. 3 is another flowchart of step S20 in the communication method of the concurrent translation of the present application.
Fig. 4 is a flowchart of step S30 in the communication method of the concurrent translation of the present application.
Fig. 5 is a flowchart of step S27 in the communication method of concurrent translation in the second embodiment of the present application.
Fig. 6 is another flowchart of step S27 in the communication method of the concurrent translation of the present application.
Fig. 7 is a flowchart of step S275 in the communication method of the concurrent translation of the present application.
Fig. 8 is a schematic block diagram of a communication system for concurrent translation in the third embodiment of the present application.
Fig. 9 is a schematic view of an apparatus in a fourth embodiment of the present application.
Detailed Description
The present application is described in further detail below in conjunction with figures 1 to 9.
Example 1
The application discloses a communication method of concurrent translation, which can be used in cross-country network language communication, and can be particularly used as a functional component of related software of instant messaging, online conferences and online games, and can also be used as a built-in program of translation equipment.
In this embodiment, the electronic terminal with the concurrent translation program is provided with a voice input module and a voice output module for collecting voice, where the voice input module includes a microphone and a high-definition collection module, the microphone is used for collecting audio language information of a user, the high-definition collection module is used for processing the collected audio language information and inputting the processed audio language information into the concurrent translation program, and the voice output module includes a speaker for receiving audio language information after concurrent translation sent by other clients to the user.
The high-definition acquisition module is internally provided with a recording noise reduction sub-module and an echo cancellation sub-module, wherein the recording noise reduction sub-module is internally provided with a noise reduction algorithm, the noise reduction algorithm is a spectral subtraction algorithm, and the specific process for noise reduction of the audio information is as follows: acquiring non-human voice section audio in the audio information, analyzing and recording noise spectrum energy corresponding to noise; subtracting noise spectral energy from the spectrum of all audio to reduce noise in the audio information; generating a compensation signal through decoding so as to filter out external noise, and stably improving the sampling frequency of the human voice; the higher the sampling frequency is, the better the sound quality is, so that the effect of collecting human voice with high definition can be achieved.
The echo cancellation submodule is internally provided with a self-adaptive echo cancellation algorithm which is used for removing sound played by the mobile phone when the microphone records audio, and an analog echo signal is estimated by adjusting an iteration update coefficient of a filter through the self-adaptive algorithm so as to approximate to the echo signal passing through an actual echo path, and then the analog echo signal is subtracted from audio information acquired by the microphone so as to achieve the function of echo cancellation.
The built-in noise reduction algorithm and echo cancellation algorithm of the high-definition acquisition module are used for realizing the capability of optimizing the recording acquisition of the electronic terminal by the algorithm, and the linkage scheme is formed by software driving hardware, so that the recording acquisition capability of the electronic terminal is greatly improved, and the accuracy of subsequent functional modules such as semantic understanding and language translation is further improved.
As shown in fig. 1, the communication method of concurrent translation specifically includes the following steps:
s10: and receiving the language message of the first client, inputting the language message into the API, creating a corresponding initial message and sending the initial message to the information translation model.
In this embodiment, the API refers to an application programming interface, which is an interface of the concurrent translation program, so that a language message received from the first client is conveniently input into the concurrent translation program; the information translation model is a model for executing information translation work in a simultaneous translation program, and a voice recognition service module, a translation service module and a reading service module are arranged in the information translation model; the first client and the second client refer to clients that are corresponding to the concurrent translation program by using a concurrent translation communication method, and in this embodiment, a language message is sent to the second client by the first client for expansion and explanation.
Specifically, receiving a language message sent by a first client, sending the language message to an API interface, and inputting the language message into a concurrent translation program through the API interface, wherein the language message can be audio information or text information; generating attribute information according to the language message and the corresponding message type, generating an initial message based on the language message and the corresponding attribute information, and facilitating the subsequent acquisition of a language message original file from the initial message and the judgment of the corresponding message type; the initial message is sent to the information translation model for further translation processing of the initial message.
Specifically, the voice recognition service module performs voice recognition service through a Wenet engine; the translation service module carries out translation service on the voice recognition result by the Google engine; and the reading service module carries out reading service on the translation result through the Google engine.
S20: and carrying out message transcription, initial language detection, target language detection and message translation on the initial message in sequence based on the communication scene to generate a target message.
In this embodiment, the communication scenario refers to a scenario of language communication between the first client and the second client, for example, cross-country chat, cross-country online game communication, cross-country online conference, and the like; message transcription refers to the process of changing the message type of an initial message, including converting audio information into text information and converting text information into audio information; the target message refers to information which is finally sent to the second client after the language message sent by the first client is translated.
Specifically, the theme of the current language communication is determined, so that a corresponding communication scene is determined, and the consistency of the semantic meaning of the target message and the meaning actually intended to be expressed by the original language message sender can be improved according to the word for adjusting the communication scene when the message is translated later; the initial message is subjected to message transfer, so that the message types of different initial messages input into the information translation model are unified, and further translation work is conveniently executed subsequently; initial language detection is carried out on the initial message so as to carry out semantic recognition based on the language category of the initial message; detecting the language used by the user of the target message receiving end, so as to be convenient for determining the language required to be used when the message translation work is executed subsequently; and translating the initial message after the message transfer processing according to the initial language and the target language to generate a target message, so that the target message is convenient to be sent to a second client.
Referring to fig. 2, in step S20, the method includes:
s21: and receiving the initial message, and judging the corresponding message type according to the attribute information of the initial message.
Specifically, an initial message sent to the concurrent translation program through the API interface is received, attribute information corresponding to the initial message is read, so that the initial message is judged to be audio information or text information, and a corresponding message type is determined.
S22: and if the initial message is the first type of information, sending the initial message to a message queue.
In this embodiment, the message type includes a first type of information and a second type of information, where the first type of information is a target message type for text audio message transcription processing work; the first type information and the second type information can be one of audio information or text information, and the first type information and the second type information are information with different message types; the message queue may select a Redis message queue or a RabbitMQ message queue.
Specifically, if the initial message is the first type of information, the initial message is the target message type of the message transfer processing work in the application, and the initial message can be directly sent to the message queue to wait for further message translation processing.
S23: if the initial message is the second type information, the initial message is transcribed into the first type information through the volcanic engine based on the communication scene, the transcribed initial message is returned to the first client, and the initial message is sent to the message queue.
Specifically, if the initial message is the second type information, the initial message is transcribed into the first type information by the volcanic engine based on the characteristics of the communication scene so as to conduct type conversion on the different types of initial messages to adapt to the algorithm of the information translation model, and in the embodiment, the first type information is audio information; the second type of information is text information; the initial message after the message transfer processing is sent to the first client, so that a user of the first client can conveniently judge whether the message transfer processing result is accurate or not, and errors can be found timely and feedback can be achieved timely; and sending the initial message after the message transfer processing to a message queue to wait for further message translation processing.
Referring to fig. 3, in step S20, the method further includes:
s24: and judging the communication scene information, and selecting a corresponding scene word stock.
Specifically, when the message translation model is used for message translation, the current communication scene is firstly judged, so that communication scene information is determined, a corresponding scene word stock is selected according to the current communication scene information, subsequent word use in the message translation work based on the scene word stock is convenient to adjust, and the correlation between the target message generated after the message translation and the current communication scene is improved.
Specifically, the judging mode of the communication scene information includes: 1. is determined by the settings of the parties involved in the language communication; 2. semantic recognition is carried out on the historical language communication message, and then the historical language communication message is determined; 3. according to the current running program of the device running the concurrent translation program, for example, when the concurrent translation program and the online game program are running on the device at the same time, a scene word stock corresponding to the online game can be automatically selected; in this embodiment, when determining the communication scene information, any one of the above determination methods may be applied, or a combination of multiple determination methods may be used to perform the comprehensive determination.
S25: the method comprises the steps of obtaining language type information of a first client, carrying out language detection processing on an initial message, generating a message to be translated, and determining the initial language information.
In this embodiment, the language type information of the first client refers to information of a common language type of the user of the first client.
Specifically, words with the same pronunciation, the same words and the same words, but with the same meaning and the same meaning may exist in different languages, so that language type information of the first client is acquired, the language detection processing of the initial message by using the language type information of the first client as a reference to select a natural language algorithm trained by the language specification is facilitated, and the accuracy of language detection of the initial message is facilitated; the language type information of the first client is obtained, and is specifically generated after judgment is carried out according to the registration information, the positioning information, the IP address and the historical transmission information of the user.
Specifically, because the language of the language message actually sent by the user of the first client may have a difference with the language type information corresponding to the language message, for example, the chinese user may include english words in the sent language message, the *** engine is used to perform language detection processing on the initial message, thereby generating a message to be translated, and determining that the language type related in the message to be translated is the initial language information, so that further message translation is convenient to be performed subsequently; for example, if the language message sent by the user of the first client is chinese information interspersed with english words, the corresponding initial language information is chinese and english.
S26: and obtaining language type information of the second client and determining target language information.
In this embodiment, the language type information of the second client refers to information of a common language type of the user of the second client.
Specifically, language type information of the second client is acquired, so that the target language information can be conveniently determined as the target language of message translation.
S27: and carrying out message translation on the message to be translated based on the scene word stock, the initial language information and the target language information to generate the target message.
Specifically, according to the initial language information, judging which words and sentences in the message to be translated are consistent with the language type information of the first client, and which words and sentences are inconsistent with the language type information of the first client, wherein the words and sentences in the part inconsistent with the language type information of the first client can be regarded as words and sentences which are deliberately emphasized by the user of the first client, and the translation is not performed.
Specifically, according to the target language information, message translation is carried out on the information to be translated, and then the result of the message translation is corrected according to the scene word stock so as to generate the target message, thereby being convenient for improving the relevance between the target message and the current communication scene and further completing the message translation work.
S30: and backing up the target message to a database, and sending the target message to the second client through the API.
Specifically, the target message is backed up to the database, so that the user can conveniently inquire the historical information, and the natural language recognition algorithm can conveniently analyze high-frequency words and sentences in each scene according to the historical information, and can conduct targeted training, and the target message is sent to the second client through the API, so that the translated information can be sent to the target user in time.
Referring to fig. 4, in step S30, the method includes:
s31: and uploading the target message to a MySQL database for storage, wherein the MySQL database comprises a history message library and a scene message feature library.
In this embodiment, the MySQL database may be selected from the Redis MySQL database or the rabkitmq MySQL database; the historical message library is used for storing all historical messages processed by the information translation model; the scene message feature library is used for storing high-frequency words and other feature words and sentences of each communication scene extracted from the historical messages.
Specifically, all target messages are uploaded to a history message library in a MySQL database for storage, all history messages in the history message library are analyzed according to a set period, high-frequency words and special words appearing in different communication scenes are judged, and the high-frequency words and the special words are added to a scene message feature library so as to be used for updating a scene word library; and after each analysis is completed, clearing the historical messages in the historical message library so as to clear the storage space of the memory.
S32: and sending the stored target message to the API, and sending the target message to the second client through the API.
Specifically, the stored target message is sent to the API, and the target message is sent to the second client through the API, so that the user receives the language message sent by the first client and translated.
S40: and acquiring feedback information of the client, and generating scene correction information based on the feedback information to update the information translation model.
In this embodiment, the feedback information refers to information that is fed back when each user participating in language communication has a question or cannot understand a message generated after message transcription or message translation.
Specifically, feedback information sent by each client is obtained in real time, wherein the feedback information comprises unclear feedback information sent by a user when the user cannot understand the received information, and further comprises correction feedback information which occurs when the user considers that translation is incorrect; generating scene correction information based on the feedback information so as to delete or reduce weight of inaccurate words and sentences in a scene word stock; the information translation model is updated, and the translation accuracy of the information translation model is improved.
Example two
In many specific scenarios, the vocabulary used in user language communication differs from the original meaning of the vocabulary; for example, in the language exchange of an electronic game, players often give out numbers for game characters and game props, which are easy to understand for players using the same language, but in the language exchange of an international network game, users in other languages are easy to misunderstand.
On the basis of the first embodiment, referring to fig. 5, in step S27, it includes:
s271: and according to the initial language information, selecting a corresponding natural language algorithm to perform semantic recognition on the message to be translated, and generating a semantic recognition result.
Specifically, because different countries/regions and different languages have different characteristics on the expression of the same thing, according to the initial language information corresponding to the initial message, a natural language algorithm trained by the specialization of the corresponding language type is selected to perform semantic recognition on the message to be translated, so that a semantic recognition result is generated, and the accuracy of the semantic recognition is improved.
S272: and carrying out relevance assessment on the semantic identification result of the message to be translated and the semantic identification result of the relevant information through a natural language algorithm, and generating a relevance assessment result.
In this embodiment, the associated information refers to a language message sent by each party earlier than the current time in the current language communication process, so as to be used for comparing with the initial message currently being translated.
Specifically, the semantic recognition result of the message to be translated is compared with the semantic recognition result of the associated information through a natural language algorithm, so that the relevance of the semantic recognition result of the message to be translated to the semantic recognition result of the associated information is evaluated, a relevance evaluation result is generated, and the relevance of the message to be translated currently in translation and the associated information is conveniently judged according to the relevance evaluation result, so that whether the semantic of the message to be translated currently in translation accords with the current communication scene or not is judged in an auxiliary mode.
S273: and if the relevance evaluation result is qualified, carrying out message translation on the message to be translated based on the semantic recognition result and the target language information, and generating a target message.
Specifically, whether the relevance evaluation result is qualified or not is judged, if so, the message translation is carried out on the message to be translated based on the semantic recognition result of the message to be translated and the target language information, so that the target message is generated.
Referring to fig. 6, in step S27, the method further includes:
s274: if the relevance evaluation result is unqualified, determining the violations and words in the message to be translated through a natural language algorithm.
Specifically, if the relevance evaluation result is unqualified, the situation that the message to be translated is difficult to understand or is not consistent with the current communication scene is considered to be possibly caused when the message to be translated is directly translated; for example, a player of a network game will usually take a foreign character for a game character according to the appearance and characteristics of the game character, the pinyin of one game character name in the "X game" is "x.xi.an" (X, A in this embodiment refers to single or multiple characters, X, A may also be used to refer to pinyin of single or multiple characters, and is used for word separation in pinyin), and since one game character in the "X game" has an appearance similar to a well-known character "XA" in a country, and the "xi.an" in the character name pinyin is difficult to accurately separate words in a common input method and is easy to be input into homophones of "xian", the character is often called "XA" or homophones of "x.xian" by domestic players, and in the scene of an online game between domestic players and foreign players, if the homophones of "XA" x.xian "are directly translated, the foreign players are easy to confuse information sent by domestic players.
Specifically, after the semantic recognition is performed on the message to be translated according to the scene, the network game scene of which the current communication scene is an X game can be judged according to the semantic recognition result of the natural language algorithm on the associated information, and the XA non-game characters in the message to be translated have poor semantic association with the associated information, so that the association evaluation result is judged to be unqualified; the offence word "XA" is identified from the message to be translated by natural language algorithms for subsequent substitution of the offence word.
S275: and matching the replacement words from the scene word stock based on the violations and regenerating the semantic recognition result and the relevance evaluation result.
In this embodiment, the scene word library is a word library for storing high-frequency words and special words in the corresponding communication scene.
Specifically, since the foreign number of the game character whose violation term "XA" is "x.xi.an" is known by the domestic player of the "X game", the scene thesaurus can pre-record such synonyms or related terms for subsequent replacement when the violation term is detected to be present in the information to be translated.
Specifically, the violating words are input into the corresponding scene word stock, the corresponding replacement words are matched, for example, the violating words XA are input into the scene word stock, the official role names of the XA in the game are matched to be the Chinese character names of the X.xi.an, the character names corresponding to the X.xi.an are used as the replacement words of the violating words XA, and the semantic recognition and the relevance assessment are carried out again to generate new semantic recognition results and new relevance assessment results.
Further, if there are multiple synonyms or related words in the scene word library for one violation word, for example, "XA" has homonyms of related words "x.xi.an" and "x.xian", multiple substitutions may be performed on the violation word in the message to be translated, and semantic recognition and relevance evaluation are performed again according to the result of each substitution, so as to determine the most suitable substitution word from the multiple synonyms.
Wherein, referring to fig. 7, in step S275, it includes:
s2751: and determining evaluation frequency information based on the frequency of the relevance evaluation, and acquiring a relevance evaluation result.
Specifically, after matching and replacing the replacement word from the scene word stock based on the violating word of the message to be translated, carrying out relevance evaluation again, obtaining the times of relevance evaluation to generate evaluation times information, obtaining the result of each relevance evaluation, and facilitating the subsequent comparison of relevance of semantic recognition results of each time so as to select the semantic recognition result with highest relevance.
S2752: when the evaluation frequency information is larger than a preset evaluation frequency threshold value, if no qualified relevance evaluation result is obtained, generating a violation and search formula based on the violation word and communication scene information.
In this embodiment, a threshold number of times is set to reduce the delay of the generation of the target message due to the time taken to repeatedly match the replacement word, regenerate the semantic recognition result, and regenerate the relevance evaluation result; the frequency threshold is used for limiting the frequency of regenerating the semantic recognition result and the relevance evaluation result; the violations and retrievals refer to retrievals used in searching for actual meanings of violations and words from the internet.
Specifically, comparing the evaluation frequency information with a preset evaluation frequency threshold, judging whether the result of each relevance evaluation is qualified or not when the evaluation frequency information is larger than the preset evaluation frequency threshold, and if the qualified relevance evaluation result is not obtained, generating violations and search based on the violations and words and communication scene information so as to search the actual meaning of the violations and words from the Internet; for example, when the violation sum word "XA" cannot be matched to an appropriate replacement word from the scene word stock, a violation sum search "X game XA" is generated based on the violation sum word "XA" and the communication scene information "X game".
S2753: and inputting the violations and the search terms into a search engine, acquiring search meaning information corresponding to the violations and the words, and regenerating a semantic recognition result based on the search meaning information.
In the present embodiment, the retrieval meaning information refers to information on the actual meaning of the violation and word searched from the internet according to the violation and retrieval formula.
Specifically, the violations and the search results are input into a search engine, the search results are processed through a natural language algorithm to obtain meanings corresponding to the violations and the words, search meaning information is generated, and a semantic recognition result is generated again according to the search meaning information, so that when the current semantics of the violations and the words cannot be successfully obtained by a scene word library, the semantics of the violations and the words in the current scene are searched by the search engine from the Internet, and the message translation work of the message to be translated is successfully executed subsequently.
S276: and carrying out message translation on the message to be translated based on the semantic recognition result and the target language information corresponding to the relevance evaluation result with the highest relevance, and generating the target message.
Specifically, after the semantic recognition and the relevance evaluation are carried out on the message to be translated again, the relevance evaluation result with the highest relevance in multiple relevance evaluations is determined, and the message to be translated is translated according to the corresponding semantic recognition result and the target language information, so that the target message is generated, and the consistency of the semantic of the target message and the semantic actually intended to be expressed by the first client user is improved.
It should be understood that the sequence number of each step in the foregoing embodiment does not mean that the execution sequence of each process should be determined by the function and the internal logic of each process, and should not limit the implementation process of the embodiment of the present application in any way.
Example III
As shown in fig. 8, the present application discloses a concurrent translation communication system for executing the steps of the concurrent translation communication method, where the concurrent translation communication system corresponds to the concurrent translation communication method in the above embodiment.
The communication system for simultaneous translation comprises an initial message creation module, a message translation module, a target message sending module and a scene correction module. The detailed description of each functional module is as follows:
the initial message creation module is used for receiving the language message of the first client and inputting the language message into the API, creating a corresponding initial message and sending the initial message to the information translation model;
the message translation module is used for sequentially carrying out message transcription, initial language detection, target language detection and message translation on the initial message based on the communication scene to generate a target message;
the target message sending module is used for backing up the target message into the database and sending the target message to the second client through the API;
And the scene correction module is used for acquiring feedback information of the client, and generating scene correction information based on the feedback information so as to update the information translation model.
Wherein the message translation module comprises:
the message type judging sub-module is used for receiving the initial message and judging the corresponding message type according to the attribute information of the initial message;
the initial message sending sub-module is used for sending the initial message to the message queue if the initial message is the first type of information;
the initial message transfer and sending sub-module is used for transferring the initial message into the first type of information through the volcanic engine based on the communication scene if the initial message is the second type of information, returning the transferred initial message to the first client, and sending the initial message to the message queue;
the scene word stock selecting sub-module is used for judging communication scene information and selecting a corresponding scene word stock;
the initial language determining sub-module is used for acquiring language type information of the first client, carrying out language detection processing on the initial message, generating a message to be translated, and determining the initial language information;
the target language determining submodule is used for acquiring language type information of the second client and determining target language information;
And the target message generation sub-module is used for translating the message to be translated based on the scene word stock, the initial language information and the target language information to generate the target message.
Wherein the target message sending module comprises:
the target message storage sub-module is used for uploading the target message to a MySQL database for storage, wherein the MySQL database comprises a history message library and a scene message feature library;
and the target message output sub-module is used for sending the stored target message to the API and sending the target message to the second client through the API.
Wherein the target message generation submodule includes:
the semantic recognition sub-module is used for selecting a corresponding natural language algorithm to carry out semantic recognition on the message to be translated according to the initial language information to generate a semantic recognition result;
the relevance evaluation sub-module is used for carrying out relevance evaluation on the semantic recognition result of the initial message and the semantic recognition result of the relevant information through a natural language algorithm to generate a relevance evaluation result;
the evaluation qualification translation submodule is used for translating the message to be translated based on the semantic recognition result and the target language information to generate a target message if the relevance evaluation result is qualified;
The violation word determining sub-module is used for determining the violation word in the message to be translated through a natural language algorithm if the relevance evaluation result is unqualified;
the replacement word matching sub-module is used for matching the replacement word from the scene word stock based on the violation and word and regenerating a semantic recognition result and a relevance evaluation result;
and the evaluation and selection translation submodule is used for translating the message to be translated based on the semantic recognition result and the target language information corresponding to the correlation evaluation result with the highest correlation, and generating the target message.
Wherein, the replacement word matching submodule includes:
the evaluation frequency acquisition sub-module is used for acquiring frequency determination evaluation frequency information of relevance evaluation and acquiring a relevance evaluation result;
the violating and searching type generating sub-module is used for generating violating and searching type based on violating words and communication scene information if a qualified relevance evaluating result is not obtained when the evaluating frequency information is larger than a preset evaluating frequency threshold value;
and the search semantic recognition sub-module is used for inputting the violations and the search terms into a search engine, acquiring search meaning information corresponding to the violations and the terms, and regenerating a semantic recognition result based on the search meaning information.
For specific limitations of the communication system of concurrent translation, reference may be made to the above limitation of the communication method of concurrent translation, and no further description is given here; all or part of each module in the communication system of the simultaneous translation can be realized by software, hardware and a combination thereof; the above modules may be embedded in hardware or may be independent of a processor in the computer device, or may be stored in software in a memory in the computer device, so that the processor may call and execute operations corresponding to the above modules.
Example IV
A computer device, which may be a server, may have an internal structure as shown in fig. 9. The computer device includes a processor, a memory, a network interface, and a database connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database of the computer equipment is used for storing data such as initial information, an information translation model, target information, feedback information, scene correction information, first type information, second type information, communication scene information, scene word stock, information to be translated, initial language information, target language information, a historical information stock, a scene information feature stock and the like. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by the processor to implement a communication method of concurrent translation.
In one embodiment, a computer device is provided comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, the processor implementing the following steps when executing the computer program:
s10: receiving language information of a first client, inputting the language information into an API, creating a corresponding initial message and sending the initial message to an information translation model;
s20: sequentially carrying out message transcription, initial language detection, target language detection and message translation on the initial message based on the communication scene to generate a target message;
s30: backing up the target message to a database, and sending the target message to a second client through an API;
s40: and acquiring feedback information of the client, and generating scene correction information based on the feedback information to update the information translation model.
In one embodiment, a computer readable storage medium is provided having a computer program stored thereon, which when executed by a processor, performs the steps of:
s10: receiving language information of a first client, inputting the language information into an API, creating a corresponding initial message and sending the initial message to an information translation model;
s20: sequentially carrying out message transcription, initial language detection, target language detection and message translation on the initial message based on the communication scene to generate a target message;
S30: backing up the target message to a database, and sending the target message to a second client through an API;
s40: and acquiring feedback information of the client, and generating scene correction information based on the feedback information to update the information translation model.
Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the various embodiments provided herein may include non-volatile and/or volatile memory. The nonvolatile memory can include Read Only Memory (ROM), programmable ROM (PROM), electrically Programmable ROM (EPROM), electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double Data Rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink), DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), among others.
It will be apparent to those skilled in the art that, for convenience and brevity of description, only the above-described division of the functional units and modules is illustrated, and in practical application, the above-described functional distribution may be performed by different functional units and modules according to needs, i.e. the internal structure of the apparatus is divided into different functional units or modules to perform all or part of the above-described functions.
The above embodiments are only for illustrating the technical solution of the present application, and are not limiting; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art will understand; the technical scheme described in the foregoing embodiments can be modified or some of the features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present application, and are intended to be included in the scope of the present application.

Claims (10)

1. A method of concurrent translation communication, comprising:
receiving language information of a first client, inputting the language information into an API, creating a corresponding initial message and sending the initial message to an information translation model;
Sequentially carrying out message transcription, initial language detection, target language detection and message translation on the initial message based on the communication scene to generate a target message;
backing up the target message to a database, and sending the target message to a second client through an API;
and acquiring feedback information of the client, and generating scene correction information based on the feedback information to update the information translation model.
2. The method for concurrent translation communication according to claim 1, wherein: the method comprises the steps of sequentially carrying out message transcription, initial language detection, target language detection and message translation on the initial message based on the communication scene, and generating a target message, wherein the steps comprise:
receiving an initial message, and judging a corresponding message type according to attribute information of the initial message;
if the initial message is the first type of information, the initial message is sent to a message queue;
if the initial message is the second type information, the initial message is transcribed into the first type information through the volcanic engine based on the communication scene, the transcribed initial message is returned to the first client, and the initial message is sent to the message queue.
3. The method for concurrent translation communication according to claim 1, wherein: the method comprises the steps of sequentially carrying out message transcription, initial language detection, target language detection and message translation on the initial message, and generating a target message, and further comprises the following steps:
Judging communication scene information, and selecting a corresponding scene word stock;
the method comprises the steps of obtaining language type information of a first client, carrying out language detection processing on an initial message, generating a message to be translated, and determining the initial language information;
obtaining language type information of a second client and determining target language information;
and carrying out message translation on the message to be translated based on the scene word stock, the initial language information and the target language information to generate the target message.
4. A method of concurrent translation communication according to claim 3, wherein: based on the scene word stock, the initial language information and the target language information, carrying out message translation on the message to be translated, and generating the target message, wherein the step of generating the target message comprises the following steps:
according to the initial language information, selecting a corresponding natural language algorithm to perform semantic recognition on the message to be translated, and generating a semantic recognition result;
carrying out relevance assessment on the semantic recognition result of the message to be translated and the semantic recognition result of the relevant information through a natural language algorithm to generate a relevance assessment result;
and if the relevance evaluation result is qualified, carrying out message translation on the message to be translated based on the semantic recognition result and the target language information, and generating a target message.
5. The method for concurrent translation communication according to claim 4, wherein: based on the scene word stock, the initial language information and the target language information, carrying out message translation on the message to be translated, and generating the target message, wherein the step of generating the target message comprises the following steps:
if the relevance evaluation result is unqualified, determining the violations and words in the message to be translated through a natural language algorithm;
matching the replacement words from the scene word stock based on the violations and the words, and regenerating semantic recognition results and relevance evaluation results;
and carrying out message translation on the message to be translated based on the semantic recognition result and the target language information corresponding to the relevance evaluation result with the highest relevance, and generating the target message.
6. The method for concurrent translation communication according to claim 5, wherein: after the step of matching the replacement word from the scene word stock based on the violation and the word and regenerating the semantic recognition result and the relevance evaluation result, the method comprises the following steps:
acquiring the number of times of relevance evaluation to determine evaluation number information, and acquiring a relevance evaluation result;
when the evaluation frequency information is larger than a preset evaluation frequency threshold value, if a qualified relevance evaluation result is not obtained, generating a violation and search formula based on the violation word and communication scene information;
And inputting the violations and the search terms into a search engine, acquiring search meaning information corresponding to the violations and the words, and regenerating a semantic recognition result based on the search meaning information.
7. A co-transmission translation communication system, comprising:
the initial message creation module is used for receiving the language message of the first client and inputting the language message into the API, creating a corresponding initial message and sending the initial message to the information translation model;
the message translation module is used for sequentially carrying out message transcription, initial language detection, target language detection and message translation on the initial message based on the communication scene to generate a target message;
the target message sending module is used for backing up the target message into the database and sending the target message to the second client through the API;
and the scene correction module is used for acquiring feedback information of the client, and generating scene correction information based on the feedback information so as to update the information translation model.
8. The concurrent translation communication system according to claim 7, wherein: the voice input module comprises a microphone and a high-definition acquisition module, and a recording noise reduction sub-module and an echo cancellation sub-module are arranged in the high-definition acquisition module.
9. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor, when executing the computer program, carries out the steps of the communication method of simultaneous interpretation according to any one of claims 1 to 7.
10. A computer readable storage medium storing a computer program, wherein the computer program when executed by a processor performs the steps of the concurrent translated communication method according to any one of claims 1 to 7.
CN202310101682.4A 2023-02-08 2023-02-08 Communication method, system, equipment and storage medium for simultaneous translation Active CN116227504B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310101682.4A CN116227504B (en) 2023-02-08 2023-02-08 Communication method, system, equipment and storage medium for simultaneous translation

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310101682.4A CN116227504B (en) 2023-02-08 2023-02-08 Communication method, system, equipment and storage medium for simultaneous translation

Publications (2)

Publication Number Publication Date
CN116227504A true CN116227504A (en) 2023-06-06
CN116227504B CN116227504B (en) 2024-01-23

Family

ID=86570846

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310101682.4A Active CN116227504B (en) 2023-02-08 2023-02-08 Communication method, system, equipment and storage medium for simultaneous translation

Country Status (1)

Country Link
CN (1) CN116227504B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117094329A (en) * 2023-10-13 2023-11-21 哈尔滨工业大学(深圳)(哈尔滨工业大学深圳科技创新研究院) Voice translation method and device for solving voice ambiguity

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110119514A (en) * 2019-04-02 2019-08-13 杭州灵沃盛智能科技有限公司 The instant translation method of information, device and system
CN111709431A (en) * 2020-06-15 2020-09-25 厦门大学 Instant translation method and device, computer equipment and storage medium
CN111814492A (en) * 2020-06-15 2020-10-23 上海传英信息技术有限公司 Translation method, terminal and computer storage medium
WO2021077333A1 (en) * 2019-10-23 2021-04-29 深圳市欢太科技有限公司 Simultaneous interpretation method and device, and storage medium
CN113591495A (en) * 2021-06-30 2021-11-02 北京搜狗智能科技有限公司 Speech translation method, device and storage medium

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110119514A (en) * 2019-04-02 2019-08-13 杭州灵沃盛智能科技有限公司 The instant translation method of information, device and system
WO2021077333A1 (en) * 2019-10-23 2021-04-29 深圳市欢太科技有限公司 Simultaneous interpretation method and device, and storage medium
CN111709431A (en) * 2020-06-15 2020-09-25 厦门大学 Instant translation method and device, computer equipment and storage medium
CN111814492A (en) * 2020-06-15 2020-10-23 上海传英信息技术有限公司 Translation method, terminal and computer storage medium
CN113591495A (en) * 2021-06-30 2021-11-02 北京搜狗智能科技有限公司 Speech translation method, device and storage medium

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117094329A (en) * 2023-10-13 2023-11-21 哈尔滨工业大学(深圳)(哈尔滨工业大学深圳科技创新研究院) Voice translation method and device for solving voice ambiguity
CN117094329B (en) * 2023-10-13 2024-02-02 哈尔滨工业大学(深圳)(哈尔滨工业大学深圳科技创新研究院) Voice translation method and device for solving voice ambiguity

Also Published As

Publication number Publication date
CN116227504B (en) 2024-01-23

Similar Documents

Publication Publication Date Title
CN110765244B (en) Method, device, computer equipment and storage medium for obtaining answering operation
US20210027788A1 (en) Conversation interaction method, apparatus and computer readable storage medium
US9564127B2 (en) Speech recognition method and system based on user personalized information
CN103635963B (en) Language model across languages initialization
CN109325091B (en) Method, device, equipment and medium for updating attribute information of interest points
US20070203708A1 (en) System and method for providing transcription services using a speech server in an interactive voice response system
CN110322872A (en) Conference voice data processing method, device, computer equipment and storage medium
CN111145737A (en) Voice test method and device and electronic equipment
US11553085B2 (en) Method and apparatus for predicting customer satisfaction from a conversation
CN110970021A (en) Question-answering control method, device and system
CN116227504B (en) Communication method, system, equipment and storage medium for simultaneous translation
CN112925945A (en) Conference summary generation method, device, equipment and storage medium
CN110377745B (en) Information processing method, information retrieval device and server
CN107239547A (en) Voice error correction method, terminal and storage medium for ordering song by voice
CN110287364B (en) Voice search method, system, device and computer readable storage medium
CN112562659B (en) Speech recognition method, device, electronic equipment and storage medium
CN111831795A (en) Multi-turn conversation processing method and device, electronic equipment and storage medium
CN111159334A (en) Method and system for house source follow-up information processing
CN111179939A (en) Voice transcription method, voice transcription device and computer storage medium
CN111416728A (en) Method, system, device and medium for predicting session end and online customer service
KR100911301B1 (en) Method and Apparatus for Determining Typographical Error
CN110209804B (en) Target corpus determining method and device, storage medium and electronic device
US20220309413A1 (en) Method and apparatus for automated workflow guidance to an agent in a call center environment
CN111477212A (en) Content recognition, model training and data processing method, system and equipment
CN111881253B (en) Method and device for determining synonymous entity pair

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant