WO2021068490A1

WO2021068490A1 - Reply message generation method and apparatus, computer device and storage medium

Info

Publication number: WO2021068490A1
Application number: PCT/CN2020/087818
Authority: WO
Inventors: 满康瑞
Original assignee: 深圳壹账通智能科技有限公司
Priority date: 2019-10-12
Filing date: 2020-04-29
Publication date: 2021-04-15
Also published as: CN110931012A

Abstract

The present application relates to the technical field of artificial intelligence, is applied to the financial industry, and relates in particular to a reply message generation method and apparatus, a computer device and a storage medium. The method in one embodiment comprises: receiving in real time a voice message inputted by a target user, and acquiring an identity identification of the target user according to the voice message; obtaining a personality classification target result according to a preset relationship between identity identifications and personality classification results; extracting a target keyword contained in the voice message, and generating a reply candidate message according to the target keyword and the personality classification target result; acquiring the sentence structure of the reply candidate message, splitting the reply candidate message according to the sentence structure to obtain a plurality of phrases, and acquiring from a preset synonym library a synonym corresponding to each phrase so as to obtain a synonymous phrase set corresponding to each phrase; and according to the sentence structure of the reply candidate message, selecting one phrase from each synonymous phrase set and combining the various selected phrases to generate a reply message.

Description

Reply message generating method, device, computer equipment and storage medium

This application claims the priority of a Chinese patent application filed with the Chinese Patent Office on October 12, 2019, the application number is 201910968475.2, and the invention title is "response message generation method, device, computer equipment and storage medium", the entire content of which is incorporated by reference Incorporated in this application.

Technical field

This application relates to the field of artificial intelligence technology, in particular to a method, device, computer equipment, and storage medium for generating a reply message.

Background technique

With the development of Internet technology, more and more users obtain required information through the Internet. Therefore, customer service methods have gradually developed into online online customer service methods. Through the customer service system, communication and exchanges between merchants and customers can be facilitated.

The response mode of the customer service system can be divided into manual response mode and automatic response mode. The inventor realized that the traditional automatic response mode is a templated response, which cannot accurately identify the user's consulting information, so as to adapt to the user's consulting needs. Sexual response, that is, the traditional automatic response mode has the problem of low information recognition accuracy.

Summary of the invention

Based on this, it is necessary to provide a response message generation method, device, computer equipment, and storage medium that can improve the accuracy of information recognition in response to the above technical problems.

A method for generating a reply message, the method comprising:

Receiving the voice message input by the target user in real time, and obtaining the identity of the target user according to the voice information;

Acquiring the relationship between the preset identity identifier and the personality classification result, and obtaining the personality classification target result corresponding to the identity identifier of the target user according to the relationship between the preset identity identifier and the personality classification result;

Extracting target keywords in the voice message, and generating response candidate messages according to the target keywords and the personality classification target results;

Acquiring the sentence structure of the candidate reply message, and splitting the candidate reply message according to the sentence structure to obtain multiple phrases;

Obtain the synonyms corresponding to each phrase from the preset thesaurus, and obtain the synonymous phrase set corresponding to each phrase;

According to the sentence structure of the candidate reply message, a phrase is selected from each synonymous phrase set, and the selected phrases are combined to generate a reply message.

A reply message generating device, the device comprising:

A user information obtaining module, configured to receive voice messages input by a target user in real time, and obtain the identity of the target user according to the voice information;

The personality classification result acquisition module is used to acquire the relationship between the preset identity identifier and the personality classification result, and obtain the personality classification target corresponding to the identity identifier of the target user according to the relationship between the preset identity identifier and the personality classification result result;

Candidate message generating module for extracting target keywords contained in the voice message, and generating a response candidate message according to the target keywords and the personality classification target result;

A candidate message processing module, configured to obtain the sentence structure of the reply candidate message, and split the reply candidate message according to the sentence structure to obtain multiple phrases;

The synonym acquisition module is used to acquire the synonym corresponding to each phrase from the preset thesaurus, and obtain the synonymous phrase set corresponding to each phrase;

The reply message generating module is used to select a phrase from each synonymous phrase set according to the sentence structure of the reply candidate message, and combine the selected phrases to generate a reply message.

A computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the following steps when the processor executes the computer program:

Extracting the target keywords contained in the voice message, and generating a response candidate message according to the target keywords and the personality classification target result;

A computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the following steps are implemented:

The foregoing reply message generation method, device, computer equipment and storage medium, on the one hand, integrate the personality classification target results and the dimensions of the keywords in the voice message to generate reply candidate messages; on the other hand, the reply candidate messages are split and processed to obtain Multiple phrases, get the synonyms corresponding to each phrase, get the synonymous phrase set corresponding to each phrase, combine the phrases in each synonymous phrase set to generate a reply message, so that the final reply message obtained integrates multiple dimensional considerations. The accuracy of information recognition can be improved, and the form of reply messages can be enriched, and the flexibility of reply can be improved.

Description of the drawings

Figure 1 is an application environment diagram of a method for generating a reply message in an embodiment;

Figure 2 is a schematic flowchart of a method for generating a reply message in an embodiment;

FIG. 3 is a schematic flowchart of a step of generating a reply message in an embodiment;

Figure 4 is a structural block diagram of an apparatus for generating a reply message in an embodiment;

Fig. 5 is an internal structure diagram of a computer device in an embodiment.

Detailed ways

The method for generating a reply message provided in this application can be applied to the application environment as shown in FIG. 1. Among them, the user terminal 102 communicates with the server 104 through the network. The server 104 receives the voice message input by the target user from the user terminal 102 in real time, and obtains the identity of the target user according to the voice message. Obtain the relationship between the preset ID and the personality classification result, and obtain the personality classification target result corresponding to the target user's ID according to the relationship between the preset ID and the personality classification result; extract the target keywords contained in the voice message according to Target keywords and personality classification target results, generate response candidate messages; obtain the sentence structure of the response candidate message, split the response candidate message according to the sentence structure, obtain multiple phrases, and obtain each phrase from the preset thesaurus Corresponding synonyms, the corresponding synonymous phrase set of each phrase is obtained; according to the sentence structure of the reply candidate message, a phrase is selected from each synonymous phrase set, and the selected phrases are combined to generate a reply message. The user terminal 102 may be, but is not limited to, various personal computers, notebook computers, smart phones, and tablet computers. The server 104 may be implemented by an independent server or a server cluster composed of multiple servers.

In one embodiment, as shown in FIG. 2, a method for generating a reply message is provided. Taking the method applied to the server in FIG. 1 as an example for description, the method includes the following steps:

Step 202: Receive the voice message input by the target user in real time, and obtain the identity of the target user according to the voice information.

The voice message may be sent by a user through a user terminal for consulting on a business of interest. The target user refers to a user who has a business consulting requirement and sends a voice message. For example, a user can download a financial application program on the user terminal, and then enter a financial voice message in the chat box loaded by the application program. User identification is used to identify different users.

Step 204: Obtain the relationship between the preset identity identifier and the personality classification result, and obtain the personality classification target result corresponding to the identity identifier of the target user according to the relationship between the preset identity identifier and the personality classification result.

According to the preset relationship between the identity identifier and the personality classification result, before obtaining the personality classification target result corresponding to the identity identifier of the target user, it also includes: sending a personality test question to the user terminal, and receiving feedback from different user terminals on the personality test question And the corresponding identity identification; according to the feedback on the personality test questions and the predetermined personality classification rules, the personality classification results corresponding to different identification identifications are obtained; according to the personality classification results corresponding to the different identification identifications, the preset identification identification and personality classification results are generated Relationship. For example, users can be divided into talkative and brief, extroverted and introverted, professional and non-professional, and so on. Associate each personality classification result with a specific personality chat robot module, and the associated chat robot module will talk to the user in a way that the user prefers. By distinguishing different types of robot modules, not only the accuracy of the analysis of voice messages can be improved, but also the response speed to user inquiries can be accelerated.

Step 206: Extract the target keywords contained in the voice message, and generate response candidate messages according to the target keywords and the personality classification target results.

The voice message can be converted into a text message through automatic voice recognition technology, and the key words of the text message can be obtained through the trained natural language understanding model. ASR (Automatic Speech Recognition, automatic speech recognition) technology is to convert the vocabulary content of human speech into computer-readable input, such as keystrokes, binary codes, or character sequences. NLU (Natural Language Understanding) models, such as neural network models, use previous text data for training to classify input text data into various intents. If the voice message is a voice message, such as data from a phone or microphone, the original voice data is converted into text through automatic voice recognition. In addition, by analyzing the voice message, the emotional state of the user can be inferred. If it is a text message, the keywords are directly obtained through the trained natural language understanding model.

The preset database stores keywords and answers corresponding to the keywords. Search in the database according to the extracted keywords to obtain the answers corresponding to the extracted keywords; the comprehensive personality classification target results and the answers corresponding to the extracted keywords are generated Reply to candidate messages. For example, the answer candidate message includes the answer corresponding to the extracted keyword plus some emoticons, or the answer corresponding to the keyword is organized in a formal sentence as the answer candidate message.

Step 208: Obtain the sentence structure of the reply candidate message, and split the reply candidate message according to the sentence structure to obtain multiple phrases.

Take "what is the current number of money owed to banks you are required to pay back?" as the reply candidate message, and obtain the sentence structure of the reply candidate message, such as subject-predicate structure, subject-predicate-object structure, subject-predicate-object structure, subject-predicate structure, subject The predicate-object-complement structure, etc., split the response candidate message according to the sentence structure, and obtain multiple different phrases such as what is the current number of, money owed to banks, you, are required to pay back, etc.

Step 210: Obtain synonyms corresponding to each phrase from a preset thesaurus, and obtain a synonymous phrase set corresponding to each phrase.

Continuing from the above, search for replacement phrases/synonym phrases for each phrase from the preset corpus, and get the synonymous phrase set corresponding to each phrase as {how many|what is the current amount of|what is the total value of} {outstanding loans|money owed to banks|debts}you {current have|currentlyowe|are required to pay back}.

Step 212: According to the sentence structure of the reply candidate message, a phrase is selected from each synonymous phrase set, and the selected phrases are merged to generate a reply message.

By expanding the processing of the reply candidate message, the problem of a single reply message form can be avoided. For example, the reply candidate message is: "what is the current number of money owed to banks you are required to pay back?" to obtain the sentence structure of the reply candidate message, and split the reply candidate message according to the sentence structure, Obtain multiple different phrases such as what is the current number of, money owed to banks, you, are required to pay back, etc., and search for the replacement phrase/synonym phrase of each phrase from the preset corpus to obtain the synonymous phrase corresponding to each phrase Collection {how many|what is the current amount of|what is the total value of} {outstanding loans|money owed to banks|debts}you{current have|currentlyowe|are required to payback}. According to the sentence structure of the reply candidate message, a phrase is selected from each synonymous phrase set, and the selected phrases are merged to generate a processed reply message with the same content as the reply candidate message but different wording. Specifically, it can be:

-"How many outstanding loans you are currently required to pay back?"

-"What is the total value of money owed to banks you currently have?"

The above method for generating reply messages, on the one hand, integrates multiple dimensions such as the user’s language model, personality classification target results, and keywords in the voice message to generate reply candidate messages; on the other hand, the reply candidate messages are split and processed to obtain multiple Get the synonym corresponding to each phrase, get the synonymous phrase set corresponding to each phrase, combine the phrases in each synonymous phrase set to generate a reply message, so that the final reply message not only integrates multiple dimensional considerations, The accuracy of information recognition can be improved, and the form of reply messages can be enriched, and the flexibility of reply can be improved.

In one embodiment, as shown in FIG. 3, according to the sentence structure of the reply candidate message, a phrase is selected from each synonymous phrase set, and the selected phrases are merged to generate the reply message before further including: 302. Number each synonymous phrase set in sequence, and determine the synonymous grammatical structure according to the current grammatical structure of the reply candidate message; step 304, determine the synonymous grammatical structure according to the sequence of each synonymous phrase set in the synonymous grammatical structure Corresponding sentence structure; according to the sentence structure of the reply candidate message, select a phrase from each synonymous phrase set separately, merge the selected phrases to generate a reply message, including: step 306, corresponding from the synonymous grammatical structure Select a sentence structure in the sentence structure, select a phrase from each synonymous phrase set to combine according to the selected sentence structure, perform grammatical correction on the combined sentence, and generate a reply message. By performing the above-mentioned expansion processing on the reply candidate message, the problem of a single reply message form can be avoided. For example: "Withrespect to the financial institutions you currently deal and the {outstanding loans|money owed to banks|debts}you{currently have|currentlyowe|are required to payback},could be you tell me{howmanthey| current amount of|what is the total amount of}these"? In this case, 3*3*3*2=54 sentences can be generated, and each sentence is different. Using grammatical changes to obtain sentences with the same meaning can be achieved in the following ways. First, search for the replacement phrase/synonym phrase of each phrase from the preset corpus, obtain the synonymous phrase set corresponding to each phrase, and number the synonymous phrase set in order to obtain: [1] {how many|what is the current amount of|what is the total value of}; [2] {outstanding loans|money owed to banks|debts}; [3] {current have|currentlyowe|are required to pay back}. Then obtain the current grammatical structure of the reply candidate message, search in the grammar library according to the current grammatical structure of the reply candidate message, and obtain all the selectable synonymous grammatical structures of the reply candidate message, and determine the number of a single grammatical structure and synonymous phrase set The sentence structure of the reply candidate message. The details can be as follows: 1): "[1][2]you[3]"?

2): "With respect to the financial institutions you currently deal with and the[2]you[3],could you tell me[1]these you have"?

3): "if we consider the"[2]you[3],what is the current amount of[2]and it’s value?

4): "You"[3]"how many[2]?

5):"At this point in time",[1][2]"that you"[3]

Choose any of the above sentence structures, and then select the components of the sentence structure from the synonymous phrase set. Taking sentence structure 5 as an example, by selecting the synonymous phrase set 1_1, 2_2, 3_3, the sentence obtained is:

"At this point in time, how many money owed to banks that you are required to pay back"

Correct the grammar of the combined sentence and get the reply message:

"At this point in time, how much money owed to banks are you required to pay back"?

Taking sentence structure 4 as an example, by selecting the synonymous phrase set 3_2, 2_1, the sentence obtained is:

"You currently owe how many outstanding loans?", the grammar of the sentence is correct, so the grammatical correction is still the same sentence.

Taking sentence structure 2 as an example, by selecting the synonymous phrase set 2_1, 3_1, 1_2, the sentence obtained is:

"With respect to the financial institutions you currently deal with and the outstanding loans you currently have, could you tell me what is the current amount of these you have?"

The sentence after grammatical correction is:

In one embodiment, the reply message generation method further includes: performing language pattern recognition on the voice message based on the written language vocabulary to obtain the language pattern of the target user; generating the reply candidate message according to the target keywords and personality classification target results, including: The target user’s language model, personality classification target result, and preset demand classification model are used to obtain the user’s demand type. Among them, the demand classification model is a neural network model that has been trained to solve the user’s demand type; Set the keyword-answer correspondence relationship to search to obtain the target answer corresponding to the target keyword; generate a response candidate message according to the user's demand type and the target answer. In this way, the response candidate message is generated from multiple dimensions to ensure its comprehensiveness. For example, the robot module in the server can be called to detect the written vocabulary contained in the voice message; when the written vocabulary contained in the voice message is less than a preset threshold, the user’s language mode is obtained as an informal language mode; when the voice message contains When the written language vocabulary of is greater than or equal to the preset threshold, the user's language mode is obtained as a formal language mode.

Pre-set rules. If the number of written words in the voice message is less than the threshold, it is determined that the user language mode belongs to the informal language mode L1; if the number of written words in the voice message is greater than or equal to the threshold, it is determined that the user language mode belongs to the formal language mode L2. By detecting the number of written words contained in the user's voice message, the user's language mode can be determined. If the user uses spoken short sentences (such as "yo", "hi", "yup"), it is judged as an informal language mode. If the user uses the sentence structure of the written language, such as "good morning", "good afternoon" and longer written sentence substructures, it is judged as a formal language mode.

According to the user's demand type and target answer, a candidate reply message is generated. For example, a user habitually checks the bank balance at a certain time of the day. After learning this mode, the robot R can actively send data without the user's request. In more complex automatic learning scenarios, the chatbot module of a formal role can suggest that the user set aside some money for long-term savings after receiving a monthly salary; chatbots in an informal role may also make similar suggestions. Over time, you can evaluate which role can more effectively achieve the user's financial goals. In another example, if a user requests to view their monthly consumption, the consumption can be presented as a pie chart, list or bar chart. If users prefer informal chatbots, these charts may be animated and cartoonized with "cute" artwork. However, if users prefer formal chatbots, then this representation is more likely to be concise, static, and commercial.

The training of the neural network model includes: obtaining sample data. The sample data includes the language pattern of the sample user, the corresponding personality classification result and the corresponding demand type; the language pattern of the sample user and the corresponding personality classification result are input into the neural network model for processing Train to obtain the training result; compare the training result with the corresponding demand type of the sample user, adjust the parameters of the neural network model until the training result meets the preset conditions, and obtain the trained neural network model for solving the user demand type.

The preset database stores keywords and the answers corresponding to the keywords. According to the extracted target keywords, search in the database to obtain the answers corresponding to the extracted target keywords; comprehensive user’s needs types and the extracted target keywords correspond to Answer, generate reply candidate message. If the financial robot module is a chat robot model corresponding to the "official" role, it will respond with a lengthy formal sentence. If the financial robot module is a chat robot model corresponding to an "informal" role, a shorter colloquial sentence is used to respond, and emojis can also be used.

such as:

(Official chatbot)

User A: "Good morning"

Robot R_f: "Good morning Jacob, what can I do for you today?"

(Informal chatbot)

User A: "Hey Paco"

Robot R_i: "How is it?"

In the example, the robot module uses "formal" and "informal" chatbot roles, and there may be more and more personas, such as "happy", "business", "childish", "scolding", "old man" , "Wise", "carefree" and so on.

It should be understood that, although the various steps in the flowchart of FIGS. 2-3 are displayed in sequence as indicated by the arrows, these steps are not necessarily performed in sequence in the order indicated by the arrows. Unless there is a clear description in this article, there is no strict order for the execution of these steps, and these steps can be executed in other orders. Moreover, at least some of the steps in Figure 2-3 may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. These sub-steps or stages The execution order of is not necessarily performed sequentially, but may be performed alternately or alternately with at least a part of other steps or sub-steps or stages of other steps.

In one embodiment, as shown in FIG. 4, a reply message generation device is provided, including: user information acquisition module 402, personality classification result acquisition module 404, candidate message generation module 406, candidate message processing module 408, synonym acquisition Module 410 and reply message generating module 412. Among them, the user information obtaining module is used to receive the voice message input by the target user in real time, and obtain the identity of the target user according to the voice message. The personality classification result acquisition module is used to acquire the relationship between the preset identity identifier and the personality classification result, and obtain the personality classification target result corresponding to the identity identifier of the target user according to the relationship between the preset identity identifier and the personality classification result. The candidate message generating module is used to extract the target keywords contained in the voice message, and generate the reply candidate message according to the target keywords and the character classification target result. The candidate message processing module is used to obtain the sentence structure of the reply candidate message, and split the reply candidate message according to the sentence structure to obtain multiple phrases. The synonym acquisition module is used to acquire the synonyms corresponding to each phrase from the preset thesaurus, and obtain the synonymous phrase set corresponding to each phrase. The reply message generation module is used to select a phrase from each synonymous phrase set according to the sentence structure of the reply candidate message, and combine the selected phrases to generate a reply message.

In one embodiment, the reply message generation module further includes a sentence pattern determination module, which is used to sequentially number each synonymous phrase set, and determine the synonymous grammatical structure of the reply candidate message according to the current grammatical structure of the reply candidate message; The sequence of the synonymous phrase set in the synonymous grammatical structure determines the sentence structure corresponding to the synonymous grammatical structure; the reply message generation module is used to select a sentence structure from the syntactic structure corresponding to the synonymous grammatical structure. Select a phrase from each synonymous phrase set to combine the sentence structure of, and make grammatical corrections to the combined sentence to generate a reply message.

In one embodiment, the personality classification result acquisition module also includes a corresponding relationship establishment module, which is used to send personality test questions to the user terminal, and receive feedback on the personality test questions from different user terminals and the corresponding identification; according to the personality test Question feedback and predetermined personality classification rules to obtain the personality classification results of different identities; according to the personality classification results corresponding to the different identities, the relationship between the preset identity identification and the personality classification results is generated.

In one embodiment, the reply message generating device further includes a language pattern acquisition module, which is used to perform language pattern recognition on the voice message based on the written language vocabulary to obtain the language pattern of the target user; the candidate message generation module is used to generate the language pattern of the target user according to the language pattern of the target user. , Personality classification target results and preset demand classification model to obtain the user's demand type, where the demand classification model is a neural network model that has been trained to solve the user's demand type; according to the target keyword in the preset keyword- Search in the answer correspondence relationship to obtain the target answer corresponding to the target keyword; generate a response candidate message according to the user's demand type and target answer.

In one embodiment, the language mode acquisition module is also used to obtain the user’s language mode as an informal language mode when the written language vocabulary contained in the voice message is less than a preset threshold; when the written language vocabulary contained in the voice message is greater than or equal to the predetermined threshold, When the threshold is set, the user's language mode is obtained as the official language mode.

In one embodiment, the candidate message generation module also includes a model building module before it is used to obtain sample data. The sample data includes the language mode of the sample user, the corresponding personality classification result, and the corresponding demand type; the language mode of the sample user and The corresponding personality classification results are input to the neural network model for training, and the training results are obtained; the training results are compared with the corresponding types of needs of the sample users, and the parameters of the neural network model are adjusted until the training results meet the preset conditions, and the trained users are obtained. Neural network model for solving the types of user needs.

Regarding the specific limitation of the reply message generating device, please refer to the above limitation on the reply message generating method, which will not be repeated here. Each module in the above reply message generating device can be implemented in whole or in part by software, hardware or a combination thereof. The above modules can be embedded in the form of hardware or independent of the processor in the computer device, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as shown in FIG. 5. The computer equipment includes a processor, a memory, a network interface, and a database connected through a system bus. Among them, the processor of the computer device is used to provide calculation 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, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database of the computer equipment is used to store data such as the relationship between the identity identifier and the personality classification result, and reply messages. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for generating a reply message is realized.

Those skilled in the art can understand that the structure shown in FIG. 5 is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may Including more or fewer parts than shown in the figure, or combining some parts, or having a different arrangement of parts.

In one embodiment, a computer device is provided, including a memory and a processor, the memory stores a computer program, and the processor implements the steps of the reply message generation method in any embodiment when the computer program is executed. The method for generating a reply message mainly includes: receiving voice information input by a target user in real time, and obtaining the identity of the target user according to the voice information; obtaining the relationship between the preset identity and the result of personality classification, and obtaining the relationship between the preset identity and the personality classification result according to the prediction. Set the relationship between the identity identifier and the personality classification result to obtain the personality classification target result corresponding to the identity identifier of the target user; extract the target keywords contained in the voice information, and according to the target keywords and the personality classification targets As a result, a candidate reply message is generated; the sentence structure of the candidate reply message is obtained, and the candidate reply message is split according to the sentence structure to obtain a plurality of phrases; each of the above-mentioned phrases is obtained from the preset thesaurus Synonyms corresponding to the phrase, obtain the synonymous phrase set corresponding to each phrase; according to the sentence structure of the reply candidate message, select a phrase from each synonymous phrase set, and combine the selected phrases To generate a reply message.

In one embodiment, a computer-readable storage medium is provided. The computer-readable storage medium may be non-volatile or volatile, and a computer program is stored thereon. When the computer program is executed by a processor, Implement the steps of the reply message generation method in any embodiment. The method for generating a reply message mainly includes: receiving voice information input by a target user in real time, and obtaining the identity of the target user according to the voice information; obtaining the relationship between the preset identity and the result of personality classification, and obtaining the relationship between the preset identity and the personality classification result according to the prediction. Set the relationship between the identity identifier and the personality classification result to obtain the personality classification target result corresponding to the identity identifier of the target user; extract the target keywords contained in the voice information, and according to the target keywords and the personality classification targets As a result, a candidate reply message is generated; the sentence structure of the candidate reply message is obtained, and the candidate reply message is split according to the sentence structure to obtain a plurality of phrases; each of the above-mentioned phrases is obtained from the preset thesaurus Synonyms corresponding to the phrase, obtain the synonymous phrase set corresponding to each phrase; according to the sentence structure of the reply candidate message, select a phrase from each synonymous phrase set, and combine the selected phrases To generate a reply message.

A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage. In the medium, when the computer program is executed, it may include the procedures of the above-mentioned method embodiments. Wherein, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and/or volatile memory. Non-volatile memory may include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not a limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous chain Channel (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

The above-mentioned embodiments only express several implementation manners of the present application, and their description is relatively specific and detailed, but they should not be understood as a limitation on the scope of the invention patent. It should be pointed out that for those of ordinary skill in the art, without departing from the concept of this application, several modifications and improvements can be made, and these all fall within the protection scope of this application. Therefore, the scope of protection of the patent of this application shall be subject to the appended claims.

Claims

A method for generating a reply message, wherein the method includes:

Receiving the voice information input by the target user in real time, and obtaining the identity of the target user according to the voice information;

Acquiring the relationship between the preset identity identifier and the personality classification result, and obtaining the personality classification target result corresponding to the identity identifier of the target user according to the relationship between the preset identity identifier and the personality classification result;

Extracting the target keywords contained in the voice information, and generating a response candidate message according to the target keywords and the personality classification target result;

Acquiring the sentence structure of the candidate reply message, and splitting the candidate reply message according to the sentence structure to obtain multiple phrases;

Obtain the synonyms corresponding to each phrase from the preset thesaurus, and obtain the synonymous phrase set corresponding to each phrase;

According to the sentence structure of the candidate reply message, a phrase is selected from each synonymous phrase set, and the selected phrases are combined to generate a reply message.
2. The method according to claim 1, wherein the sentence structure of the candidate message for reply selects a phrase from each synonymous phrase set, and combines the selected phrases to generate a reply message Before, it also included:

Serially number each of the synonymous phrase sets, and determine the synonymous grammatical structure according to the current grammatical structure of the reply candidate message;

Determine the sentence structure corresponding to the synonymous grammatical structure according to the sequence of the synonymous phrase set in the synonymous grammatical structure;

According to the sentence structure of the reply candidate message, selecting a phrase from each synonymous phrase set and combining the selected phrases to generate a reply message includes:

Select a sentence structure from the sentence structure corresponding to the synonymous grammatical structure, select a phrase from each synonymous phrase set to combine according to the selected sentence structure, and perform grammatical correction on the combined sentence To generate a reply message.
The method according to claim 1, wherein before obtaining the personality classification target result corresponding to the identity of the target user according to the relationship between the preset identity identifier and the personality classification result, the method further comprises:

Sending a personality test question to the user terminal, and receiving feedback from different user terminals on the personality test question and the corresponding identification;

According to the feedback on the personality test question and the predetermined personality classification rules, the personality classification results corresponding to different identities are obtained;

According to the personality classification results corresponding to the different identities, a preset relationship between the identities and the personality classification results is generated.
The method according to claim 1, wherein the method further comprises:

Performing language pattern recognition on the voice message based on the written language vocabulary to obtain the language pattern of the target user;

The generating a response candidate message according to the target keyword and the personality classification target result includes:

Obtain the user's demand type according to the target user's language mode, personality classification target result, and a preset demand classification model, where the demand classification model is a trained neural network model for solving the user's demand type;

Searching in a preset keyword-answer correspondence according to the target keyword to obtain a target answer corresponding to the target keyword;

According to the user's demand type and the target answer, a response candidate message is generated.
The method according to claim 4, wherein the performing language pattern recognition on the voice message based on the written language vocabulary to obtain the language pattern of the user comprises:

When the written language vocabulary contained in the voice message is less than a preset threshold, the user's language mode is obtained as an informal language mode;

When the written language vocabulary included in the voice message is greater than or equal to the preset threshold, the user's language mode is obtained as a formal language mode.
The method according to claim 4, wherein, before obtaining the user's demand type according to the target user's language mode, personality classification target result, and preset demand classification model, the method comprises:

Obtain sample data, which includes the language patterns of the sample users, the corresponding personality classification results, and the corresponding demand types;

Input the language patterns of the sample users and the corresponding personality classification results into the neural network model for training, and obtain the training results;

The training results are compared with the corresponding types of needs of the sample users, and the parameters of the neural network model are adjusted until the training results meet the preset conditions, and the trained neural network model for solving the types of user needs is obtained.
A reply message generating device, wherein the device includes:

A user information obtaining module, configured to receive voice messages input by a target user in real time, and obtain the identity of the target user according to the voice information;

The personality classification result acquisition module is used to acquire the relationship between the preset identity identifier and the personality classification result, and obtain the personality classification target corresponding to the identity identifier of the target user according to the relationship between the preset identity identifier and the personality classification result result;

Candidate message generating module for extracting target keywords contained in the voice message, and generating a response candidate message according to the target keywords and the personality classification target result;

A candidate message processing module, configured to obtain the sentence structure of the reply candidate message, and split the reply candidate message according to the sentence structure to obtain multiple phrases;

The synonym acquisition module is used to acquire the synonym corresponding to each phrase from the preset thesaurus, and obtain the synonymous phrase set corresponding to each phrase;

The reply message generating module is used to select a phrase from each synonymous phrase set according to the sentence structure of the reply candidate message, and combine the selected phrases to generate a reply message.
8. The device according to claim 7, wherein the reply message generating module further comprises:

The sentence pattern determination module is used to sequentially number each of the synonymous phrase sets, determine the synonymous grammatical structure of the reply candidate message according to the current grammatical structure of the reply candidate message; State the order in the synonymous grammatical structure, and determine the sentence structure corresponding to the synonymous grammatical structure;

The reply message generating module is used to select a sentence structure from the sentence structure corresponding to the synonymous grammatical structure, and select a phrase from each synonymous phrase set for combination according to the selected sentence structure, Correct the grammar of the combined sentence to generate a reply message.
A computer device includes a memory and a processor, the memory stores a computer program, wherein the processor implements the steps of a method for generating a reply message when the computer program is executed, and receives voice information input by a target user in real time, And obtain the identity of the target user according to the voice information;

Acquiring the relationship between the preset identity identifier and the personality classification result, and obtaining the personality classification target result corresponding to the identity identifier of the target user according to the relationship between the preset identity identifier and the personality classification result;

Extracting the target keywords contained in the voice information, and generating a response candidate message according to the target keywords and the personality classification target result;

Acquiring the sentence structure of the candidate reply message, and splitting the candidate reply message according to the sentence structure to obtain multiple phrases;

Obtain the synonyms corresponding to each phrase from the preset thesaurus, and obtain the synonymous phrase set corresponding to each phrase;

According to the sentence structure of the candidate reply message, a phrase is selected from each synonymous phrase set, and the selected phrases are combined to generate a reply message.
9. The computer device according to claim 9, wherein the sentence structure of the reply candidate message selects one phrase from each synonymous phrase set, and combines the selected phrases to generate a reply Before the message, it also includes:

Serially number each of the synonymous phrase sets, and determine the synonymous grammatical structure according to the current grammatical structure of the reply candidate message;

Determine the sentence structure corresponding to the synonymous grammatical structure according to the sequence of the synonymous phrase set in the synonymous grammatical structure;

According to the sentence structure of the reply candidate message, selecting a phrase from each synonymous phrase set and combining the selected phrases to generate a reply message includes:

Select a sentence structure from the sentence structure corresponding to the synonymous grammatical structure, select a phrase from each synonymous phrase set to combine according to the selected sentence structure, and perform grammatical correction on the combined sentence To generate a reply message.
9. The computer device according to claim 9, wherein before obtaining the personality classification target result corresponding to the identity of the target user according to the relationship between the preset identity identifier and the personality classification result, the method further comprises:

Sending a personality test question to the user terminal, and receiving feedback from different user terminals on the personality test question and the corresponding identification;

According to the feedback on the personality test question and the predetermined personality classification rules, the personality classification results corresponding to different identities are obtained;

According to the personality classification results corresponding to the different identities, a preset relationship between the identities and the personality classification results is generated.
The computer device according to claim 9, wherein the method for generating a reply message further comprises:

Performing language pattern recognition on the voice message based on the written language vocabulary to obtain the language pattern of the target user;

The generating a response candidate message according to the target keyword and the personality classification target result includes:

Obtain the user's demand type according to the target user's language mode, personality classification target result, and a preset demand classification model, where the demand classification model is a trained neural network model for solving the user's demand type;

Searching in a preset keyword-answer correspondence according to the target keyword to obtain a target answer corresponding to the target keyword;

According to the user's demand type and the target answer, a response candidate message is generated.
The computer device according to claim 12, wherein the performing language pattern recognition on the voice message based on the written language vocabulary to obtain the language pattern of the user comprises:

When the written language vocabulary contained in the voice message is less than the preset threshold, the user's language mode is obtained as an informal language mode;

When the written language vocabulary included in the voice message is greater than or equal to the preset threshold, the user's language mode is obtained as a formal language mode.
11. The computer device according to claim 12, wherein, before obtaining the user's demand type according to the target user's language model, personality classification target result, and preset demand classification model, the method comprises:

Obtain sample data, which includes the language patterns of the sample users, the corresponding personality classification results, and the corresponding demand types;

Input the language patterns of the sample users and the corresponding personality classification results into the neural network model for training, and obtain the training results;

The training results are compared with the needs types corresponding to the sample users, and the parameters of the neural network model are adjusted until the training results meet the preset conditions, and a trained neural network model for solving user needs types is obtained.
A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the steps of a method for generating a reply message are realized, and the method for generating a reply message includes:

Receiving the voice information input by the target user in real time, and obtaining the identity of the target user according to the voice information;

Acquiring the relationship between the preset identity identifier and the personality classification result, and obtaining the personality classification target result corresponding to the identity identifier of the target user according to the relationship between the preset identity identifier and the personality classification result;

Extracting the target keywords contained in the voice information, and generating a response candidate message according to the target keywords and the personality classification target result;

Acquiring the sentence structure of the candidate reply message, and splitting the candidate reply message according to the sentence structure to obtain multiple phrases;

Obtain the synonyms corresponding to each phrase from the preset thesaurus, and obtain the synonymous phrase set corresponding to each phrase;

According to the sentence structure of the candidate reply message, a phrase is selected from each synonymous phrase set, and the selected phrases are combined to generate a reply message.
15. The computer-readable storage medium according to claim 15, wherein the phrase is selected from each of the synonymous phrase sets according to the sentence structure of the reply candidate message, and the selected phrases are combined Before generating the reply message, it also includes:

Serially number each of the synonymous phrase sets, and determine the synonymous grammatical structure according to the current grammatical structure of the reply candidate message;

Determine the sentence structure corresponding to the synonymous grammatical structure according to the sequence of the synonymous phrase set in the synonymous grammatical structure;

According to the sentence structure of the reply candidate message, selecting a phrase from each synonymous phrase set and combining the selected phrases to generate a reply message includes:

Select a sentence structure from the sentence structure corresponding to the synonymous grammatical structure, select a phrase from each synonymous phrase set to combine according to the selected sentence structure, and perform grammatical correction on the combined sentence To generate a reply message.
The computer-readable storage medium according to claim 15, wherein, before obtaining the personality classification target result corresponding to the identity of the target user according to the preset relationship between the identity identifier and the personality classification result, further include:

Sending a personality test question to the user terminal, and receiving feedback from different user terminals on the personality test question and the corresponding identification;

According to the feedback on the personality test question and the predetermined personality classification rules, the personality classification results corresponding to different identities are obtained;

According to the personality classification results corresponding to the different identities, a preset relationship between the identities and the personality classification results is generated.
The computer-readable storage medium according to claim 15, wherein the method for generating a reply message further comprises:

Performing language pattern recognition on the voice message based on the written language vocabulary to obtain the language pattern of the target user;

The generating a response candidate message according to the target keyword and the personality classification target result includes:

Obtain the user's demand type according to the target user's language mode, personality classification target result, and a preset demand classification model, where the demand classification model is a trained neural network model for solving the user's demand type;

Searching in a preset keyword-answer correspondence according to the target keyword to obtain a target answer corresponding to the target keyword;

According to the user's demand type and the target answer, a response candidate message is generated.
18. The computer-readable storage medium of claim 18, wherein the performing language pattern recognition on the voice message based on the written language vocabulary to obtain the user's language pattern comprises:

When the written language vocabulary contained in the voice message is less than the preset threshold, the user's language mode is obtained as an informal language mode;

When the written language vocabulary included in the voice message is greater than or equal to the preset threshold, the user's language mode is obtained as a formal language mode.
18. The computer-readable storage medium according to claim 18, wherein, before obtaining the user's demand type according to the target user's language model, personality classification target result, and preset demand classification model, the method comprises:

Obtain sample data, which includes the language patterns of the sample users, the corresponding personality classification results, and the corresponding demand types;

Input the language patterns of the sample users and the corresponding personality classification results into the neural network model for training, and obtain the training results;

The training results are compared with the corresponding types of needs of the sample users, and the parameters of the neural network model are adjusted until the training results meet the preset conditions, and the trained neural network model for solving the types of user needs is obtained.