CN114116976A - Genetic algorithm-based intention identification method and system - Google Patents

Genetic algorithm-based intention identification method and system Download PDF

Info

Publication number
CN114116976A
CN114116976A CN202111402425.1A CN202111402425A CN114116976A CN 114116976 A CN114116976 A CN 114116976A CN 202111402425 A CN202111402425 A CN 202111402425A CN 114116976 A CN114116976 A CN 114116976A
Authority
CN
China
Prior art keywords
intention
result
information
scoring
element information
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.)
Pending
Application number
CN202111402425.1A
Other languages
Chinese (zh)
Inventor
杨方兴
卢秋如
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.)
Individual
Original Assignee
Individual
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 Individual filed Critical Individual
Priority to CN202111402425.1A priority Critical patent/CN114116976A/en
Publication of CN114116976A publication Critical patent/CN114116976A/en
Pending legal-status Critical Current

Links

Images

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
    • G06F16/3331—Query processing
    • G06F16/334—Query execution
    • G06F16/3344—Query execution using natural language analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/12—Computing arrangements based on biological models using genetic models
    • G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Biophysics (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computational Linguistics (AREA)
  • General Engineering & Computer Science (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • General Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Data Mining & Analysis (AREA)
  • Biomedical Technology (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Evolutionary Computation (AREA)
  • Genetics & Genomics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Physiology (AREA)
  • Databases & Information Systems (AREA)
  • Machine Translation (AREA)

Abstract

The application discloses an intention identification method and system based on a genetic algorithm, wherein the method comprises the following steps: obtaining first intention element information, wherein the first intention element information comprises first vocabulary element information; constructing a first intention database, and obtaining a first intention list set, wherein the first intention list set and the first vocabulary element information have a first mapping relation to encode the first vocabulary element information, and a first encoding result is obtained; inputting the first coding result into a first fitness scoring model to obtain a first scoring result, wherein the first scoring result is used for representing semantic quality of the first intention list; and if the first scoring result meets a first preset scoring threshold, obtaining a first intention identification result, wherein the first intention identification result is the first intention with the highest first scoring result. The method solves the technical problem that the prior art cannot adapt to the intention recognition of complex and special contexts, so that the accuracy of the intention recognition is not enough.

Description

Genetic algorithm-based intention identification method and system
Technical Field
The application relates to the technical field of human-computer interaction, in particular to an intention identification method and system based on a genetic algorithm.
Background
The intention identification is an important module in the human-computer interaction process, appeal and demand information of a user can be determined through accurate processing of natural language, corresponding services or responses can be matched, the satisfaction degree of the user service can be improved through good human-computer interaction service, the convenience of intelligent service is reflected, and therefore the improvement of the accuracy of the intention identification is necessary.
The current intention recognition is mainly divided into two types, one is that the user intention is determined by searching keywords from a dictionary and sequencing the keywords, the efficiency of the method is low, and the construction data of the dictionary is not necessarily comprehensive; the other is that the neural network model is used for analyzing natural sentences to determine the user intention, but the user intention is difficult to recognize when the user is in a complex or special context, such as the situation that the language of the user is distorted, the quantifier word is used wrongly, the word order is wrong, and the like.
In the process of implementing the technical solution in the embodiment of the present application, the inventor of the present application finds that the above-mentioned technology has at least the following technical problems:
the prior art cannot adapt to the intention recognition of complex and special contexts, so that the technical problem of insufficient intention recognition accuracy exists.
Disclosure of Invention
The application aims to provide an intention identification method and system based on a genetic algorithm, which are used for solving the technical problem that the intention identification accuracy is not enough due to the fact that the intention identification method and system cannot be suitable for complicated and special contexts in the prior art. The method comprises the steps of collecting intention element information in a human-computer interaction process in real time, extracting vocabulary elements from the intention element information, constructing an intention combination database which appears in historical data formed by the vocabulary elements, generating a mapping relation between the vocabulary elements and intention combination data, coding the intention elements according to the mapping relation, grading intention semantic quality represented by a coding result by using a fitness model, if the grading meets a preset value, using intention combination data corresponding to the highest grading meeting the preset value as a recognition result, grading a plurality of intention combination data by the fitness model constructed based on the context, improving the complex context adaptability and the intention retrieval capability, and achieving the technical effect of improving the intention recognition accuracy and the intention adaptability.
In view of the foregoing problems, embodiments of the present application provide an intention identification method and system based on a genetic algorithm.
In a first aspect, the present application provides a genetic algorithm-based intention recognition method, which is implemented by a genetic algorithm-based intention recognition system, wherein the method includes: obtaining first intention element information, wherein the first intention element information comprises first vocabulary element information; constructing a first intention database, and obtaining a first intention list set, wherein the first intention list set and the first vocabulary element information have a first mapping relation; coding the first vocabulary element information according to the first mapping relation to obtain a first coding result; constructing a first fitness scoring model, inputting a first coding result into the first fitness scoring model, and obtaining a first scoring result, wherein the first scoring result is used for representing the semantic quality of the first intention list; judging whether the first grading result meets a first preset grading threshold value or not; and if the first scoring result meets a first preset scoring threshold, obtaining a first intention identification result, wherein the first intention identification result is the first intention with the highest first scoring result.
In another aspect, the present application further provides a genetic algorithm-based intention identifying system for performing the genetic algorithm-based intention identifying method according to the first aspect, wherein the system comprises: a first obtaining unit configured to obtain first intention element information, wherein the first intention element information includes first vocabulary element information; the second obtaining unit is used for constructing a first intention database and obtaining a first intention list set, wherein the first intention list set and the first vocabulary element information have a first mapping relation; a third obtaining unit, configured to encode the first vocabulary element information according to the first mapping relationship, and obtain a first encoding result; a fourth obtaining unit, configured to construct a first fitness scoring model, input the first coding result into the first fitness scoring model, and obtain a first scoring result, where the first scoring result is used to represent semantic quality of the first intention list; the first judging unit is used for judging whether the first grading result meets a first preset grading threshold value or not; a fifth obtaining unit, configured to, if the first scoring result meets a first preset scoring threshold, obtain a first intention identification result, where the first intention identification result is the first intention with the highest first scoring result.
In a third aspect, the present application further provides an intention identification system based on a genetic algorithm, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method according to the first aspect when executing the program.
One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:
due to the adoption of the method, the first intention element information is obtained, wherein the first intention element information comprises first vocabulary element information; constructing a first intention database, and obtaining a first intention list set, wherein the first intention list set and the first vocabulary element information have a first mapping relation; coding the first vocabulary element information according to the first mapping relation to obtain a first coding result; constructing a first fitness scoring model, inputting a first coding result into the first fitness scoring model, and obtaining a first scoring result, wherein the first scoring result is used for representing the semantic quality of the first intention list; judging whether the first grading result meets a first preset grading threshold value or not; and if the first scoring result meets a first preset scoring threshold, obtaining a first intention identification result, wherein the first intention identification result is the first intention with the highest first scoring result. The method comprises the steps of collecting intention element information in a human-computer interaction process in real time, extracting vocabulary elements from the intention element information, constructing an intention combination database which appears in historical data formed by the vocabulary elements, generating a mapping relation between the vocabulary elements and intention combination data, coding the intention elements according to the mapping relation, grading intention semantic quality represented by a coding result by using a fitness model, if the grading meets a preset value, using intention combination data corresponding to the highest grading meeting the preset value as a recognition result, grading a plurality of intention combination data by the fitness model constructed based on the context, improving the complex context adaptability and the intention retrieval capability, and achieving the technical effect of improving the intention recognition accuracy and the intention adaptability.
The foregoing description is only an overview of the technical solutions of the present application, and the present application can be implemented according to the content of the description in order to make the technical means of the present application more clearly understood, and the following detailed description of the present application is given in order to make the above and other objects, features, and advantages of the present application more clearly understandable.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only exemplary, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.
FIG. 1 is a schematic flow chart of an intent recognition method based on genetic algorithm according to an embodiment of the present application;
FIG. 2 is a schematic flow chart of a random modification method for a first encoding result in an intent recognition method based on a genetic algorithm according to an embodiment of the present application;
FIG. 3 is a schematic structural diagram of an intention recognition system based on a genetic algorithm according to an embodiment of the present application;
fig. 4 is a schematic structural diagram of an exemplary electronic device according to an embodiment of the present application.
Description of reference numerals: a first obtaining unit 11, a second obtaining unit 12, a third obtaining unit 13, a fourth obtaining unit 14, a first judging unit 15, a fifth obtaining unit 16, a bus 300, a receiver 301, a processor 302, a transmitter 303, a memory 304, and a bus interface 305.
Detailed Description
The embodiment of the application provides an intention identification method and system based on a genetic algorithm, and solves the technical problem that the intention identification accuracy is not enough due to the fact that the intention identification method and system cannot be suitable for complicated and special contexts in the prior art. The method comprises the steps of collecting intention element information in a human-computer interaction process in real time, extracting vocabulary elements from the intention element information, constructing an intention combination database which appears in historical data formed by the vocabulary elements, generating a mapping relation between the vocabulary elements and intention combination data, coding the intention elements according to the mapping relation, grading intention semantic quality represented by a coding result by using a fitness model, if the grading meets a preset value, using intention combination data corresponding to the highest grading meeting the preset value as a recognition result, grading a plurality of intention combination data by the fitness model constructed based on the context, improving the complex context adaptability and the intention retrieval capability, and achieving the technical effect of improving the intention recognition accuracy and the intention adaptability.
In the following, the technical solutions in the embodiments of the present application will be clearly and completely described with reference to the accompanying drawings, and it is to be understood that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments of the present application, and it should be understood that the present application is not limited by the example embodiments described herein. All other embodiments, which can be derived by a person skilled in the art from the embodiments of the present application without making any creative effort, shall fall within the protection scope of the present application. It should be further noted that, for the convenience of description, only some but not all of the elements relevant to the present application are shown in the drawings.
Summary of the application
The intention identification is an important module in the human-computer interaction process, appeal and demand information of a user can be determined through accurate processing of natural language, corresponding services or responses can be matched, the satisfaction degree of the user service can be improved through good human-computer interaction service, the convenience of intelligent service is reflected, and therefore the improvement of the accuracy of the intention identification is necessary.
The current intention recognition is mainly divided into two types, one is that the user intention is determined by searching keywords from a dictionary and sequencing the keywords, the efficiency of the method is low, and the construction data of the dictionary is not necessarily comprehensive; the other is that the neural network model is used for analyzing natural sentences to determine the user intention, but the user intention is difficult to recognize when the user is in a complex or special context, such as the situation that the language of the user is distorted, the quantifier word is used wrongly, the word order is wrong, and the like.
The prior art cannot adapt to the intention recognition of complex and special contexts, so that the technical problem of insufficient intention recognition accuracy exists.
In view of the above technical problems, the technical solution provided by the present application has the following general idea:
the application provides an intention identification method based on a genetic algorithm, which is applied to an intention identification system based on a genetic algorithm, wherein the method comprises the following steps: obtaining first intention element information, wherein the first intention element information comprises first vocabulary element information; constructing a first intention database, and obtaining a first intention list set, wherein the first intention list set and the first vocabulary element information have a first mapping relation; coding the first vocabulary element information according to the first mapping relation to obtain a first coding result; constructing a first fitness scoring model, inputting a first coding result into the first fitness scoring model, and obtaining a first scoring result, wherein the first scoring result is used for representing the semantic quality of the first intention list; judging whether the first grading result meets a first preset grading threshold value or not; and if the first scoring result meets a first preset scoring threshold, obtaining a first intention identification result, wherein the first intention identification result is the first intention with the highest first scoring result.
Having thus described the general principles of the present application, various non-limiting embodiments thereof will now be described in detail with reference to the accompanying drawings.
Example one
Referring to fig. 1, an intention identification method based on a genetic algorithm is provided in an embodiment of the present application, wherein the method includes:
s100: obtaining first intention element information, wherein the first intention element information comprises first vocabulary element information;
specifically, the first intention element refers to elements such as natural language words and body actions generated by a user in a multi-round man-machine interaction process, and the elements can express user requirement information. Storing each round of interaction process into a group of data according to the sequence of interaction, and marking element information of interaction end as a service end state; and marking the corresponding intention element of the interactive process as the service state. Further, in the current round of interaction in the service state, the intention elements corresponding to the user sentences which are replied by interaction are marked as intention identification ending states; and marking the intention elements corresponding to the user sentences to be replied as intention to-be-recognized states.
The first vocabulary element refers to natural language vocabulary generated by a user in a plurality of interactive processes of an intention to be recognized state extracted from the first intention element, and comprises single words and fixed collocation vocabulary, such as: individual, sheet, etc. of the individual semantic meaning of the element; fixed and collocated vocabulary intention elements such as apples, durian and the like. Further, if the man-machine interaction process is through character interaction, the punctuation marks, punctuation marks and the like of the sentence input by the user are all the elements of the user intention; if the human-computer interaction process is input through voice, the sentence break and pause of the input voice of the user are all the elements of the user intention. The method for extracting the intention element is as follows: after an input sentence of a user is read, a corpus constructed based on big data is used for counting the occurrence probability of words formed by adjacent words, the occurrence probability is high when the occurrence frequency of the adjacent words is high, word segmentation is carried out according to the probability value, and the words are split into a plurality of intention elements.
By splitting the input sentences of the user into the intention elements, the corpus constructed based on the big data has strong recognition capability on the vocabularies, provides comprehensive basic information for intention combination in the next step, and ensures the accuracy of the intention recognition result.
S200: constructing a first intention database, and obtaining a first intention list set, wherein the first intention list set and the first vocabulary element information have a first mapping relation;
specifically, the first intention database refers to a database constructed by inputting and searching a first vocabulary element into a corpus to obtain multiple used permutation sequence combinations stored in a historical service process, and stores permutation sequence sentence-breaking combination information of the first vocabulary element information and intention information obtained after combination in the historical service process in a one-to-one corresponding list form to generate a first intention list set; the first mapping relation refers to combination information of corresponding intentions obtained by sentence breaking and permutation and combination of the first vocabulary element information. In the first intention database, each group of first mapping relation and first intention are stored in a one-to-one correspondence mode in a list form, corresponding first vocabulary element information can be called through the first intention and the first mapping relation, and the first intention information can be characterized through the first mapping relation and the first vocabulary element information.
Through calling the historical service data with higher contact ratio with the first vocabulary element information, the contact ratio calculation mode is as follows: degree of coincidence ═First, the Heavy load A Combination of Chinese herbs Word Word Sink (C) Sink (C) To be administered To be administered Vegetable extract Vegetable extract Letter Letter Information processing device Information processing device General assembly Number of Number of Measurement of Measurement of X 100%, when the degree of coincidence is equal to or greater than a preset degree of coincidence, exemplified are: and 95%, the corresponding historical data can be called to construct a first intention database, so that whether the intention identification result in the historical service data meets the user requirement or not can be conveniently judged by combining the context in the later step, if yes, the later step analysis process is not needed, the intention identification result in the historical service data is directly output as the identification result, and the intention identification efficiency is improved.
S300: coding the first vocabulary element information according to the first mapping relation to obtain a first coding result;
further, based on the encoding of the first vocabulary element information according to the first mapping relationship, a first encoding result is obtained, and step S300 includes:
s310: sequencing the first vocabulary element information through the first mapping relation to obtain a first sequencing result;
s320: and obtaining a first preset coding rule, and coding the first vocabulary element information in sequence according to the first sequencing result through the first preset coding rule to obtain a first coding result.
Specifically, the first encoding result refers to a result obtained by encoding first vocabulary element information based on a first mapping relation by using a preset encoding rule, and the first vocabulary element information is processed in an encoding form in order to convert specific first vocabulary element information into data which can be identified and processed by a computer; the first ordering result refers to an ordering sequence sentence-break combination relation of first vocabulary element information represented by a first mapping relation; the first preset encoding rule refers to a preset unified encoding standard, so that encoding and decoding processes of encoding results are the same reference. The encoding process is illustratively: the single word or punctuation of the first vocabulary element information is used as a special symbol: numbers, upper and lower case letters and other symbols are identified, further, binary codes are used as a first preset coding rule to represent the numbers, the upper and lower case letters and the other symbols to obtain corresponding coding sequences, and different numbers, the upper and lower case letters and other symbols can be linked to a first intention list set through a first mapping relation, so that the first coding result is determined. The first vocabulary element information and the first sequencing result can be represented through the first coding result, the multiple groups of first intents correspond to the multiple groups of first coding results, and after the coding is finished, the first coding results are set to be in a state of waiting for response, so that the information feedback processing is facilitated.
S400: constructing a first fitness scoring model, inputting the first coding result into the first fitness scoring model, and obtaining a first scoring result, wherein the first scoring result is used for representing the semantic quality of the first intention list;
specifically, the first fitness scoring model refers to an intelligent model which is constructed based on a neural network model and combined with an actual multi-turn dialogue interaction context and used for scoring and screening the applicability degree of the first intention characterized by the first coding result. The determination work of the first scoring result is divided into two steps, wherein the first step of scoring is as follows: and traversing the first intention list set to obtain fitness scores corresponding to a plurality of groups of first intentions, wherein the scoring standard is set based on the degree of fit of the first intentions and the service types of the multi-turn interactive context representations, the preferred fitness score is 100 in total, and the higher the degree of fit of the service types of the first intentions and the multi-turn interactive context representations is, the higher the score is, the lower the score is, and the lowest the score is 0.
Exemplary are: extracting tag information of a service type from a multi-party call interaction context, wherein the service type is as follows: online ticket, locomotive type: high-speed rail, seat type: second seat, time: 10 month 1, ticket purchaser: zhang III, site: xi' an to beijing, quantity: one sheet. If any three intentions are: "Zhang Sanliang buys No. 10 month No. 1 second seat high-speed railway ticket from Xian to Beijing"; "Zhang Sanlian buys No. 10 month No. 1 high-speed railway ticket of second class from Beijing to Xian"; "Zhang Sanliang buys a No. 10 month No. 1 high-speed railway ticket of second class from Xian to Beijing". Since the tag information is 7 in total, each tag accounts for 100/7, the satisfaction degree of passing tags is scored firstly, the first sentence satisfies 6 tags, the second sentence satisfies 6 tags, the third sentence satisfies 7 tags, and then if the order context is not matched, 30 points are deducted. The first sentence is not deducted, the second sentence is-30 points, the third sentence is not deducted, and the final scores are respectively: 85 min, 55 min and 100 min. Assuming that the next Tuesday is No. 10 month 1, any score which is less than or equal to 70 points and greater than 60 points is marked for the complex context which cannot be identified by the information system of 'Zhang III to buy a second-class high-speed railway ticket from Xian to Beijing on the next Tuesday'.
The second step is to perform a screening, exemplary: directly screening out the items with scores lower than 60 points, indicating that the items cannot be adapted to the current interaction context, and marking the first intention information with scores more than 60 points and less than or equal to 70 points as a complex or special context intention information type; scores greater than 70 are labeled as intent information for the general context. By scoring the intention information represented by the first coding result based on the context fitness and screening out the intention information with lower fitness, the intention information is divided into two types of complicated or special context intention information and intention information of a common context in the later step for situation processing, the interference of the low fitness information is reduced, after elimination, the data redundancy is reduced, and the processing efficiency is improved.
S500: judging whether the first grading result meets a first preset grading threshold value or not;
s600: if the first scoring result meets the first preset scoring threshold, obtaining a first intention identification result, wherein the first intention identification result is the first intention with the highest first scoring result.
Specifically, the first preset scoring threshold refers to a preset lowest scoring value meeting the intention information of the general context, and when the first scoring result is greater than the first preset scoring threshold, the first scoring result meets the first preset scoring threshold; further, after all the first scoring results which are larger than the first preset scoring threshold value are traversed, multiple groups of first intentions meeting the first preset scoring threshold value are obtained, and the intention with the highest first scoring result of the multiple groups of first intentions is output as the first intention identification result. The following are exemplary: the first preset score threshold is set to 70 points, as in the example above: and if the score of the first sentence is 85 points and the score of the third sentence is 100 points, selecting the third sentence as the first intention recognition result, and if the highest scores are the same, selecting one of the first sentence and the third sentence as the first intention recognition result. And then the corresponding service can be provided for the user through the first intention recognition result.
Further, the method further includes step S700:
s710: obtaining first action element information according to the first intention element information, wherein the time sequences of the first action element information and the first vocabulary element information are in one-to-one correspondence;
s720: obtaining a first intention adjustment parameter according to the first action element information;
s730: and adjusting the first intention list set through the first intention adjusting parameter to obtain a second intention list set, wherein the second intention list set and the first vocabulary element information have a second mapping relation.
Specifically, the first action element information is information corresponding to the time sequence of the first vocabulary element information of the user extracted from the first intention element information, and further can be used as a sign language recognition action element of the deaf-mute; the preferred intelligent camera device which is in communication connection with the intention recognition system is used for carrying out image acquisition on a user who interacts with the intention recognition system in real time, extracting an action data stream and storing the action data stream according to a time sequence.
Furthermore, when the user has characters or voice input, the first action element information and the first vocabulary element information with the same sequence are stored in a list form in a one-to-one correspondence manner; when the user only inputs the action, the first action element information is stored separately and is identified by combining big data, and the action identification is exemplarily as follows: extracting coordinate information of human body motion joint points in each frame of video, such as: joints of the neck, the chest, the head, the right shoulder, the left shoulder, the right elbow, the left elbow, the right wrist, the left wrist and the like; and storing the actions of the joint points according to a time sequence to generate an action data stream, matching historical action intentions based on the identity information of the user, and matching corresponding action intentions with big data to finish identification.
The first intention adjustment parameter is divided into two types: one is a process of representing that first action element information is used for replacing first vocabulary element information to become new first vocabulary element information when a user cannot input a special context of characters, and further replacing a first intention list set with intention data corresponding to the first action element information to obtain a second intention list set; the second is a process that the user can input the common context of the first vocabulary element, the representation uses the combination of the first action element information and the first vocabulary element information, and the intention data corresponding to the first action element information is used for perfecting the first intention list set to obtain the second intention list set. The second mapping relation refers to a permutation and combination relation between the second intention list set after being characterized and the first vocabulary element information after being characterized. Through the addition of the first action element information, the obtained second intention list set is richer and more comprehensive than the content represented by the first intention list set, and the adaptability to special contexts and the accuracy of intention identification are improved.
Further, based on the constructing of the first fitness scoring model, step S400 includes:
s410: obtaining first contextual feature information according to the first user portrait information;
s420: constructing a first preset scoring rule according to the first contextual feature information;
s430: constructing a first fitness scoring function according to the second mapping relation and the first preset scoring rule;
s440: and training the first fitness scoring model according to the first fitness scoring function.
Specifically, the first user portrait information refers to user tag information for performing human-computer interaction, including but not limited to: name, gender, service interest points, consultation records and other information; the first contextual feature information refers to first user profile information based on first user profile information for identifying user intent, such as: extracting interaction record information in a plurality of rounds of interaction processes, consulted service type information and other characteristic information from the user image information. Furthermore, the service type information can be matched with the intention element information in the current first coding result, and the service type information of the current consultation is judged; the current interaction progress can be determined through the interaction record information in the multi-round interaction process. The first preset scoring rule refers to a preset evaluation rule for the adaptation degree of the first coding result and the first contextual characteristic information, and exemplarily: listing the extracted first context characteristic information, wherein the more the intention element information represented by the first coding result is satisfied, the stronger the adaptation degree is; the first fitness scoring function refers to a function which is constructed by combining a first preset scoring rule and a second mapping relation and evaluates the fitness of the intentions in the first intention list set represented by the first preset scoring rule, and is exemplarily shown as a score calculation method in step S600.
Furthermore, a frame structure of a first fitness scoring model is constructed by combining a customized first fitness scoring function with a neural network, and then a plurality of groups of training data are used for training the model, wherein the plurality of groups of training data all comprise: and when the first fitness scoring model converges, the first coding result and the identification information for identifying the first scoring result can be scored in real time, so that accurate processing data is provided for further intention identification.
Further, as shown in fig. 2, the method further includes step S800:
s810: if the first grading result does not meet the first preset grading threshold value, obtaining a first arrangement instruction, wherein the first arrangement instruction comprises a first random ordering instruction and a first random synonymous replacement instruction;
s820: randomly ordering and modifying the first coding result according to the first random ordering instruction to obtain a first modification result;
s830: carrying out random synonymous replacement on the first coding result according to the first random synonymous replacement instruction to obtain a second modification result;
s840: and inputting the first modification result and/or the second modification result into the first fitness scoring model to obtain a second scoring result.
Specifically, when the first scoring result does not satisfy the first preset scoring threshold, the first coding result is adjusted through the first randomly generated arrangement instruction; the first orchestration instruction comprises the first randomly ordered instruction: refers to an instruction for randomly ordering and modifying the first coding result which is not screened out, and the following embodiments are provided:
randomly adjusting the coding sequence of the first coding result which is not screened out; after modification, the first modified result is obtained, and when the first coding result is the same as the first coding result which is already screened out, the first coding result is screened out again.
The first arranging instruction comprises the first random synonymous replacement instruction, and random synonymous word replacement can be performed on the un-screened first coding result through the first random synonymous replacement instruction, so that the following embodiments are provided:
and performing replacement of lexical element synonymous codes on the un-screened first coding result, and obtaining the second modification result after the replacement, wherein the screened coding result is screened immediately in the replacement process. Exemplary are as follows: the original intention information corresponding to the first encoding result is: assuming that the next Tuesday is No. 10 month 1, for the situation that Zhang Sandu is to buy a second-class seat high-speed railway ticket from Xian to Beijing, the coding content corresponding to the next Tuesday is replaced by the No. 10 month 1 by calculating the current time node, and then the coding content is input into the first fitness scoring model for scoring, so that a scoring result meeting a first preset scoring threshold value can be obtained, and the intention identification of the complex context is realized.
The first modification result and the second modification result are completely random when the first coding result is modified, and only one of the modifications may occur at the same time, so that due to random global traversal retrieval, the intention content represented by the first modification result and the second modification result obtained after the first coding result is updated, and the first modification result or the second modification result is more than that in the first intention list set, so that the intention identification result with the optimal score can be obtained, and the accuracy of intention identification of special contexts such as partial spoken language and the like is improved.
Further, based on the obtaining the first contextual feature information according to the first user portrait information, step S410 includes:
s411: acquiring first natural sentence information, wherein the first natural sentence information and the first intention element information are in one-to-one correspondence;
s412: obtaining first front and rear statement information according to the first natural statement information;
s413: acquiring first service scene information according to the first user portrait information;
s414: and obtaining the first contextual feature information according to the first front-back statement information and the first service scene information.
Specifically, the first natural sentence information refers to input sentence information of a first user for multi-round interaction, and the first intention element information is a result obtained by splitting the first natural sentence information; the first front and back statement information refers to context information of multi-round interaction of the first natural statement information, and the requirement data of the user can be conveniently identified by extracting the context information; the first service scene information refers to consulting history of the first user portrait information is extracted to obtain consulting service type information under the current situation of first front and back statement information; further, similar interactive contents in historical interactive data can be matched according to the first front and back statement information to determine requirement information with high probability of the user, the plurality of contextual characteristic information under the business type can be matched by consulting business type information, and the requirement information with high probability of the user and the plurality of contextual characteristic information are set as first contextual characteristic information. The first contextual characteristic information is extracted to provide a judgment reference for scoring of the first coding result, when the data volume is small in the early stage, namely when the first fitness scoring model is initialized, the first contextual characteristic information needs to be marked on the first coding result by depending on a worker, when the data volume is large, namely when the first fitness scoring model is updated again after convergence, the worker does not need to provide the first contextual characteristic information, and the first contextual characteristic information can be automatically generated based on historical data and big data.
Further, the method step S414 includes:
s4141: obtaining first historical interaction data according to the first user portrait information;
s4142: using front and back statement information and business scene information in a plurality of groups of first historical interactive data as input training data, using identification data for identifying first user demand information as output training data, and constructing a demand prediction model;
s4143: and inputting the first fore-and-aft statement information and the first business scenario information into the demand prediction model to obtain a first demand prediction result, wherein the first demand prediction result comprises the first contextual feature information.
Specifically, the determination of the first contextual feature information can be acquired by using an intelligent model constructed based on a neural network, the first historical interaction data refers to extracted multi-round interaction historical consultation information of the first user portrait information, and the training data amount is preferably supplemented through big data matching with human-computer interaction data of the same service scene as the multi-round interaction historical consultation information of the first user portrait information; the first user demand information refers to historical user demand information corresponding to the multiple rounds of interactive historical consultation information; and constructing the demand forecasting model by taking the front and back statement information and the business scene information in the plurality of groups of first historical interactive data as input training data and taking identification data for identifying first user demand information as output training data. The first demand prediction result refers to an output result obtained by inputting first front and rear statement information and first business scene information into the demand prediction model after the demand prediction model converges, the user demand information is predicted through the neural network model, the first demand prediction result is split to obtain a plurality of label information representing the context, the label information is recorded as first context feature information, the first coding result is further graded based on the first context feature information, and the grading is determined to be an intention recognition result according with the first demand prediction result when the grading reaches a preset value, so that the accuracy is high.
In summary, the intention identification method based on the genetic algorithm provided by the embodiment of the present application has the following technical effects:
1. the application aims to provide an intention identification method and system based on a genetic algorithm, which are used for solving the technical problem that the intention identification accuracy is not enough due to the fact that the intention identification method and system cannot be suitable for complicated and special contexts in the prior art. The method comprises the steps of collecting intention element information in a human-computer interaction process in real time, extracting vocabulary elements from the intention element information, constructing an intention combination database which appears in historical data formed by the vocabulary elements, generating a mapping relation between the vocabulary elements and intention combination data, coding the intention elements according to the mapping relation, grading intention semantic quality represented by a coding result by using a fitness model, if the grading meets a preset value, using intention combination data corresponding to the highest grading meeting the preset value as a recognition result, grading a plurality of intention combination data by the fitness model constructed based on the context, improving the complex context adaptability and the intention retrieval capability, and achieving the technical effect of improving the intention recognition accuracy and the intention adaptability.
2. Due to random global traversal retrieval, the first modification result and the second modification result obtained after the first coding result is updated, and the intention content represented by the first modification result or the second modification result is more than that in the first intention list set, so that the intention identification result with the optimal score can be obtained, and the accuracy of intention identification of special contexts such as partial spoken language and the like is improved.
Example two
Based on the intention identification method based on the genetic algorithm in the previous embodiment, the invention also provides an intention identification system based on the genetic algorithm, please refer to fig. 3, wherein the system comprises:
a first obtaining unit 11, wherein the first obtaining unit 11 is used for obtaining first intention element information, and the first intention element information comprises first vocabulary element information;
a second obtaining unit 12, where the second obtaining unit 12 is configured to construct a first intention database, and obtain a first intention list set, where the first intention list set and the first vocabulary element information have a first mapping relationship;
a third obtaining unit 13, where the third obtaining unit 13 is configured to encode the first vocabulary element information according to the first mapping relationship, and obtain a first encoding result;
a fourth obtaining unit 14, where the fourth obtaining unit 14 is configured to construct a first fitness scoring model, input the first coding result into the first fitness scoring model, and obtain a first scoring result, where the first scoring result is used to characterize semantic quality of the first intention list;
a first judging unit 15, where the first judging unit 15 is configured to judge whether the first scoring result meets a first preset scoring threshold;
a fifth obtaining unit 16, where the fifth obtaining unit 16 is configured to obtain a first intention identification result if the first scoring result meets a first preset scoring threshold, where the first intention identification result is the first intention with the highest first scoring result.
Further, the system further comprises:
a sixth obtaining unit, configured to obtain first action element information according to the first intention element information, where the first action element information and the first vocabulary element information are in one-to-one correspondence in time sequence;
a seventh obtaining unit configured to obtain a first intention adjustment parameter according to the first action element information;
an eighth obtaining unit, configured to adjust the first intention list set by the first intention adjustment parameter, and obtain a second intention list set, where the second intention list set and the first vocabulary element information have a second mapping relationship.
Further, the system further comprises:
a ninth obtaining unit, configured to obtain first contextual feature information according to the first user portrait information;
the first construction unit is used for constructing a first preset scoring rule according to the first contextual feature information;
the second construction unit is used for constructing a first fitness scoring function according to the second mapping relation and the first preset scoring rule;
and the first training unit is used for training the first fitness scoring model according to the first fitness scoring function.
Further, the system further comprises:
a tenth obtaining unit, configured to, if the first scoring result does not meet the first preset scoring threshold, obtain a first arrangement instruction, where the first arrangement instruction includes a first random ordering instruction and a first random synonymous replacement instruction;
an eleventh obtaining unit, configured to perform random ordering modification on the first encoding result according to the first random ordering instruction, so as to obtain a first modification result;
a twelfth obtaining unit, configured to perform random synonymous replacement on the first encoding result according to the first random synonymous replacement instruction, and obtain a second modification result;
a thirteenth obtaining unit, configured to input the first modification result and/or the second modification result into the first fitness scoring model, and obtain a second scoring result.
Further, the system further comprises:
a fourteenth obtaining unit, configured to obtain first natural sentence information, where the first natural sentence information and the first intention element information are in one-to-one correspondence;
a fifteenth obtaining unit configured to obtain first front-rear sentence information from the first natural sentence information;
a sixteenth obtaining unit, configured to obtain first service scene information according to the first user portrait information;
a seventeenth obtaining unit, configured to obtain the first context feature information according to the first previous and subsequent statement information and the first service scenario information.
Further, the system further comprises:
an eighteenth obtaining unit, configured to obtain first historical interaction data according to the first user portrait information;
a third construction unit, configured to use previous and subsequent statement information and business scenario information in a plurality of sets of the first historical interaction data as input training data, use identification data identifying first user demand information as output training data, and construct a demand prediction model;
a nineteenth obtaining unit, configured to input the first contextual statement information and the first business scenario information into the demand prediction model, and obtain a first demand prediction result, where the first demand prediction result includes the first contextual feature information.
Further, the system further comprises:
a twentieth obtaining unit, configured to sort the first vocabulary element information by using the first mapping relationship, and obtain a first sorting result;
a twenty-first obtaining unit, configured to obtain a first preset encoding rule, and sequentially encode the first vocabulary element information according to the first ordering result according to the first preset encoding rule, so as to obtain the first encoding result.
The embodiments in the present specification are described in a progressive manner, each embodiment focuses on differences from other embodiments, the genetic algorithm based intention identification method and the specific example in the first embodiment of fig. 1 are also applicable to the genetic algorithm based intention identification system in the present embodiment, and a genetic algorithm based intention identification system in the present embodiment is clearly known to those skilled in the art through the foregoing detailed description of the genetic algorithm based intention identification method, so that details are not described herein for brevity of the description. The device disclosed by the embodiment corresponds to the method disclosed by the embodiment, so that the description is simple, and the relevant points can be referred to the method part for description.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Exemplary electronic device
The electronic device of the embodiment of the present application is described below with reference to fig. 4.
Fig. 4 illustrates a schematic structural diagram of an electronic device according to an embodiment of the present application.
Based on the inventive concept of a genetic algorithm based intention identification method as in the previous embodiments, the present invention further provides a genetic algorithm based intention identification system, on which a computer program is stored, which when executed by a processor implements the steps of any one of the above-described genetic algorithm based intention identification methods.
Where in fig. 4 a bus architecture (represented by bus 300), bus 300 may include any number of interconnected buses and bridges, bus 300 linking together various circuits including one or more processors, represented by processor 302, and memory, represented by memory 304. The bus 300 may also link together various other circuits such as peripherals, voltage regulators, power management circuits, and the like, which are well known in the art, and therefore, will not be described any further herein. A bus interface 305 provides an interface between the bus 300 and the receiver 301 and transmitter 303. The receiver 301 and the transmitter 303 may be the same element, i.e., a transceiver, providing a means for communicating with various other apparatus over a transmission medium.
The processor 302 is responsible for managing the bus 300 and general processing, and the memory 304 may be used for storing data used by the processor 302 in performing operations.
The application provides an intention identification method based on a genetic algorithm, which is applied to an intention identification system based on a genetic algorithm, wherein the method comprises the following steps: obtaining first intention element information, wherein the first intention element information comprises first vocabulary element information; constructing a first intention database, and obtaining a first intention list set, wherein the first intention list set and the first vocabulary element information have a first mapping relation; coding the first vocabulary element information according to the first mapping relation to obtain a first coding result; constructing a first fitness scoring model, inputting a first coding result into the first fitness scoring model, and obtaining a first scoring result, wherein the first scoring result is used for representing the semantic quality of the first intention list; judging whether the first grading result meets a first preset grading threshold value or not; and if the first scoring result meets a first preset scoring threshold, obtaining a first intention identification result, wherein the first intention identification result is the first intention with the highest first scoring result. The method solves the technical problem that the prior art cannot adapt to the intention recognition of complex and special contexts, so that the accuracy of the intention recognition is not enough. The method comprises the steps of collecting intention element information in a human-computer interaction process in real time, extracting vocabulary elements from the intention element information, constructing an intention combination database which appears in historical data formed by the vocabulary elements, generating a mapping relation between the vocabulary elements and intention combination data, coding the intention elements according to the mapping relation, grading intention semantic quality represented by a coding result by using a fitness model, if the grading meets a preset value, using intention combination data corresponding to the highest grading meeting the preset value as a recognition result, grading a plurality of intention combination data by the fitness model constructed based on the context, improving the complex context adaptability and the intention retrieval capability, and achieving the technical effect of improving the intention recognition accuracy and the intention adaptability.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, apparatus, or computer program product. Accordingly, the present application may take the form of an entirely software embodiment, an entirely hardware embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application is in the form of a computer program product that may be embodied on one or more computer-usable storage media having computer-usable program code embodied therewith. And such computer-usable storage media include, but are not limited to: various media capable of storing program codes, such as a usb disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk Memory, a Compact Disc Read-Only Memory (CD-ROM), and an optical Memory.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a system for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including an instruction system which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks. While preferred embodiments of the present invention have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including preferred embodiments and all such alterations and modifications as fall within the scope of the invention.
It will be apparent to those skilled in the art that various changes and modifications may be made in the present invention without departing from the spirit and scope of the invention. Thus, if such modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include such modifications and variations.

Claims (9)

1. An intent recognition method based on a genetic algorithm, wherein the method comprises:
obtaining first intention element information, wherein the first intention element information comprises first vocabulary element information;
constructing a first intention database, and obtaining a first intention list set, wherein the first intention list set and the first vocabulary element information have a first mapping relation;
coding the first vocabulary element information according to the first mapping relation to obtain a first coding result;
constructing a first fitness scoring model, inputting the first coding result into the first fitness scoring model, and obtaining a first scoring result, wherein the first scoring result is used for representing the semantic quality of the first intention list;
judging whether the first grading result meets a first preset grading threshold value or not;
if the first scoring result meets the first preset scoring threshold, obtaining a first intention identification result, wherein the first intention identification result is the first intention with the highest first scoring result.
2. The method of claim 1, wherein the method further comprises:
obtaining first action element information according to the first intention element information, wherein the time sequences of the first action element information and the first vocabulary element information are in one-to-one correspondence;
obtaining a first intention adjustment parameter according to the first action element information;
and adjusting the first intention list set through the first intention adjusting parameter to obtain a second intention list set, wherein the second intention list set and the first vocabulary element information have a second mapping relation.
3. The method of claim 2, wherein the constructing a first fitness scoring model comprises:
obtaining first contextual feature information according to the first user portrait information;
constructing a first preset scoring rule according to the first contextual feature information;
constructing a first fitness scoring function according to the second mapping relation and the first preset scoring rule;
and training the first fitness scoring model according to the first fitness scoring function.
4. The method of claim 3, wherein the method further comprises:
if the first grading result does not meet the first preset grading threshold value, obtaining a first arrangement instruction, wherein the first arrangement instruction comprises a first random ordering instruction and a first random synonymous replacement instruction;
randomly ordering and modifying the first coding result according to the first random ordering instruction to obtain a first modification result;
carrying out random synonymous replacement on the first coding result according to the first random synonymous replacement instruction to obtain a second modification result;
and inputting the first modification result and/or the second modification result into the first fitness scoring model to obtain a second scoring result.
5. The method of claim 3, wherein obtaining first contextual feature information based on the first user profile information comprises:
acquiring first natural sentence information, wherein the first natural sentence information and the first intention element information are in one-to-one correspondence;
obtaining first front and rear statement information according to the first natural statement information;
acquiring first service scene information according to the first user portrait information;
and obtaining the first contextual feature information according to the first front-back statement information and the first service scene information.
6. The method of claim 5, wherein the method comprises:
obtaining first historical interaction data according to the first user portrait information;
using front and back statement information and business scene information in a plurality of groups of first historical interactive data as input training data, using identification data for identifying first user demand information as output training data, and constructing a demand prediction model;
and inputting the first fore-and-aft statement information and the first business scenario information into the demand prediction model to obtain a first demand prediction result, wherein the first demand prediction result comprises the first contextual feature information.
7. The method as claimed in claim 1, wherein said encoding said first vocabulary element information according to said first mapping relationship to obtain a first encoding result comprises:
sequencing the first vocabulary element information through the first mapping relation to obtain a first sequencing result;
and obtaining a first preset coding rule, and coding the first vocabulary element information in sequence according to the first sequencing result through the first preset coding rule to obtain a first coding result.
8. An intent recognition system based on genetic algorithms, wherein the system comprises:
a first obtaining unit configured to obtain first intention element information, wherein the first intention element information includes first vocabulary element information;
the second obtaining unit is used for constructing a first intention database and obtaining a first intention list set, wherein the first intention list set and the first vocabulary element information have a first mapping relation;
a third obtaining unit, configured to encode the first vocabulary element information according to the first mapping relationship, and obtain a first encoding result;
a fourth obtaining unit, configured to construct a first fitness scoring model, input the first coding result into the first fitness scoring model, and obtain a first scoring result, where the first scoring result is used to represent semantic quality of the first intention list;
the first judging unit is used for judging whether the first grading result meets a first preset grading threshold value or not;
a fifth obtaining unit, configured to, if the first scoring result meets the first preset scoring threshold, obtain a first intention identification result, where the first intention identification result is the first intention with the highest first scoring result.
9. A genetic algorithm based intention recognition system comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor when executing the program implements the steps of the method of any one of claims 1 to 7.
CN202111402425.1A 2021-11-19 2021-11-19 Genetic algorithm-based intention identification method and system Pending CN114116976A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111402425.1A CN114116976A (en) 2021-11-19 2021-11-19 Genetic algorithm-based intention identification method and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111402425.1A CN114116976A (en) 2021-11-19 2021-11-19 Genetic algorithm-based intention identification method and system

Publications (1)

Publication Number Publication Date
CN114116976A true CN114116976A (en) 2022-03-01

Family

ID=80371730

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111402425.1A Pending CN114116976A (en) 2021-11-19 2021-11-19 Genetic algorithm-based intention identification method and system

Country Status (1)

Country Link
CN (1) CN114116976A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116089586A (en) * 2023-02-10 2023-05-09 百度在线网络技术（北京）有限公司 Question generation method based on text and training method of question generation model

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116089586A (en) * 2023-02-10 2023-05-09 百度在线网络技术（北京）有限公司 Question generation method based on text and training method of question generation model
CN116089586B (en) * 2023-02-10 2023-11-14 百度在线网络技术（北京）有限公司 Question generation method based on text and training method of question generation model

Similar Documents

Publication Publication Date Title
CN109189901B (en) Method for automatically discovering new classification and corresponding corpus in intelligent customer service system
CN110399609B (en) Intention recognition method, device, equipment and computer readable storage medium
CN112800170A (en) Question matching method and device and question reply method and device
CN112346567A (en) Virtual interaction model generation method and device based on AI (Artificial Intelligence) and computer equipment
CN104951433A (en) Method and system for intention recognition based on context
CN111177310A (en) Intelligent scene conversation method and device for power service robot
CN110222145A (en) A kind of intelligence law appraisal procedure and system
CN111897930A (en) Automatic question answering method and system, intelligent device and storage medium
CN114238607B (en) Deep interactive AI intelligent job-searching consultant method, system and storage medium
CN111831801B (en) Man-machine conversation method and system
CN114757176A (en) Method for obtaining target intention recognition model and intention recognition method
CN115470338B (en) Multi-scenario intelligent question answering method and system based on multi-path recall
CN111125457A (en) Deep cross-modal Hash retrieval method and device
CN113821605A (en) Event extraction method
CN111651606B (en) Text processing method and device and electronic equipment
CN111429157A (en) Method, device and equipment for evaluating and processing complaint work order and storage medium
CN113064980A (en) Intelligent question and answer method and device, computer equipment and storage medium
CN115293817A (en) Advertisement text generation method and device, equipment, medium and product thereof
CN113312468A (en) Conversation mode-based conversation recommendation method, device, equipment and medium
CN115146124A (en) Question-answering system response method and device, equipment, medium and product thereof
CN114116976A (en) Genetic algorithm-based intention identification method and system
TW202034207A (en) Dialogue system using intention detection ensemble learning and method thereof
CN114281972A (en) Dialog control method, system storage medium and server based on subject object tracking and cognitive inference
CN111460114A (en) Retrieval method, device, equipment and computer readable storage medium
CN112256864A (en) Multi-intention recognition method and device, electronic equipment and readable storage medium

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