CN112115244B - Dialogue interaction method, device, storage medium and electronic device - Google Patents

Abstract

The embodiment of the present application discloses a dialogue interaction method, device, storage medium and electronic device, which belongs to the field of computer technology. The method is applied to an electronic device with a preset dialogue model built in, and the electronic device analyzes the interaction satisfaction of the first user according to the first interaction result, and the first interaction result is the interaction result output by the preset dialogue model for the voice data input by the first user. When the interaction satisfaction is less than or equal to the preset threshold, the voice data is sent to the server, and the interaction instruction corresponding to the voice data sent by the server is received, and the interaction instruction and the voice data are used as sample data to optimize the preset dialogue model. The server searches for the answer data corresponding to the voice data to obtain new sample data for optimizing the training of the preset dialogue model, thereby reducing the workload of manual participation and making the process of optimizing the training of the preset dialogue model more intelligent.

Description

Dialogue interaction method, device, storage medium and electronic device

Technical Field

The present application relates to the field of computer technology, and in particular to a conversation interaction method, device, storage medium and electronic device.

Background technique

With the development of artificial intelligence technology, human-computer dialogue has also been developed in the fields of smart home, smart navigation, smart assistant, etc. The human-computer dialogue process mainly includes: the user interacts with the dialogue model using natural language, and the dialogue model parses the natural language input from the user and provides corresponding output. In the related technology, the dialogue model of human-computer dialogue can first be trained through a large amount of sample data, but in the subsequent use of the dialogue model, there may be a situation where the voice data input by the user cannot be recognized. Usually, the content of the dialogue interaction between the user and the dialogue model can be processed by manual labeling to obtain new sample data, and then the dialogue model is optimized and trained using the new sample data. However, obtaining new sample data by manual labeling will result in high labor costs.

Summary of the invention

The embodiments of the present application provide a method, device, storage medium and electronic device for dialogue interaction, which can solve the problem of high labor cost in obtaining sample data for optimizing and training dialogue models in human-computer dialogue in related technologies. The technical solution is as follows:

In a first aspect, an embodiment of the present application provides a dialog interaction method, the method comprising:

Analyzing the interaction satisfaction of the first user according to the first interaction result; wherein the first interaction result is the interaction result output by a preset dialogue model for the voice data input by the first user;

When the interaction satisfaction is less than or equal to a preset threshold, sending the voice data to a server, and receiving an interaction instruction corresponding to the voice data sent by the server;

The interaction instruction and the voice data are used as sample data to optimize and train the preset dialogue model.

In a second aspect, an embodiment of the present application provides a dialog interaction device, the device comprising:

An analysis module, configured to analyze the interaction satisfaction of the first user according to a first interaction result; wherein the first interaction result is an interaction result output by a preset dialogue model for voice data input by the first user;

A processing module, configured to send the voice data to a server when the interaction satisfaction is less than or equal to a preset threshold, and receive an interaction instruction corresponding to the voice data sent by the server;

A training module is used to optimize and train the preset dialogue model by using the interaction instruction and the voice data as sample data.

In a third aspect, an embodiment of the present application provides a computer storage medium, wherein the computer storage medium stores a plurality of instructions, wherein the instructions are suitable for being loaded by a processor and executing the above-mentioned method steps.

In a fourth aspect, an embodiment of the present application provides an electronic device, comprising: a processor, a memory, and a display screen; wherein the memory stores a computer program, and the computer program is suitable for being loaded by the processor and executing the above-mentioned method steps.

The beneficial effects brought about by the technical solutions provided by some embodiments of the present application include at least:

When the scheme of the embodiment of the present application is executed, the electronic device analyzes the interaction satisfaction of the first user based on the first interaction result, where the first interaction result is the interaction result output by the preset dialogue model for the voice data input by the first user. When the interaction satisfaction is less than or equal to the preset threshold, the electronic device sends the voice data to the server, receives the interaction instructions corresponding to the voice data sent by the server, uses the interaction instructions and the voice data as sample data to optimize and train the preset dialogue model, and obtains new sample data for optimizing and training the preset dialogue model by searching the server for answer data corresponding to the voice data, thereby reducing the workload of manual participation and making the process of optimizing and training the preset dialogue model more intelligent.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

FIG1 is a schematic diagram of a system architecture of a dialog interaction provided in an embodiment of the present application;

FIG2 is a flow chart of a conversation interaction method provided in an embodiment of the present application;

FIG3 is another flow chart of a conversation interaction method provided in an embodiment of the present application;

FIG4 is a schematic diagram of the structure of a conversation interaction device provided in an embodiment of the present application;

FIG5 is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application.

Detailed ways

In order to make the objectives, technical solutions and advantages of the present application more clear, the embodiments of the present application will be further described in detail below with reference to the accompanying drawings.

When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. Instead, they are only examples of devices and methods consistent with some aspects of the present application as detailed in the attached claims.

In the description of this application, it should be understood that the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to specific circumstances. "And/or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and/or B can represent: A exists alone, A and B exist at the same time, and B exists alone. The character "/" generally indicates that the previously associated objects are in an "or" relationship.

In order to make the objectives, technical solutions and advantages of the present application more clear, the embodiments of the present application will be further described in detail below with reference to the accompanying drawings.

Please refer to Figure 1, which shows a schematic diagram of a system architecture of a dialogue interaction provided by an embodiment of the present application, including a user 101, a terminal device 102 and a server 103. The terminal device 102 has a built-in preset dialogue model that can be used for human-computer dialogue. The terminal device 102 can conduct a dialogue interaction with the user 101 based on the preset dialogue model. The terminal device 102 can also send query data related to the user dialogue interaction to the server 103. Correspondingly, the server 103 can send the query result to the terminal device 102 after querying/searching the query data sent by the terminal device 102.

The terminal device 102 may be any electronic device with a speaker, and the terminal device 102 may be equipped with an original trained preset dialogue model, so that the user 102 may interact with the terminal device 102 through dialogue; the terminal device 102 may include, but is not limited to, electronic devices with intelligent voice functions such as smart speakers, smart phones, tablet computers, portable computers, and desktop computers. The terminal device 102 may be hardware or software. When the terminal device 102 is software, it may be installed in the electronic devices listed above, and it may be implemented as multiple software or software modules, or as a single software or software module, which is not specifically limited here. When the terminal device 102 is hardware, a display device may also be installed thereon, and the display device may be any device capable of realizing a display function, such as a cathode ray tube display (Cathoderay tube display, referred to as CR), a light-emitting diode display (Light-emitting diode display, referred to as LED), an electronic ink screen, a liquid crystal display (Liquid crystal display, referred to as LCD), a plasma display panel (Plasma display panel, referred to as PDP), etc. The terminal device 102 can also perform data communication interaction with the server 103. The terminal device 102 can query data from the server 103 and receive query result data sent by the server 103. The server 103 can also send query result data to the terminal device 102 and receive query data sent by the terminal device 102.

The server 103 may be hardware or software. When the server 103 is hardware, it may be implemented as a distributed server cluster consisting of multiple servers, or as a single server. When the server 103 is software, it may be implemented as multiple software or software modules, or as a single software or software module, which is not specifically limited here.

It should be understood that the number of users, terminal devices and servers in Figure 1 is only illustrative and can be any number of users, terminal devices and servers according to actual needs.

The following will describe in detail the conversation interaction method provided in the embodiment of the present application in conjunction with Figures 2 to 3.

Please refer to Figure 2, which is a flowchart of a conversation interaction method provided in an embodiment of the present application. This embodiment is illustrated by applying a conversation interaction method to an electronic device (terminal). The conversation interaction method may include the following steps:

S201, analyzing the interaction satisfaction of a first user according to a first interaction result.

Among them, the first interaction result is the interaction result output by the preset dialogue model for the voice data input by the first user, that is, the dialogue content data generated by the preset dialogue model. The first user refers to the user who inputs the voice data to the preset dialogue model for the first time. "First" is to distinguish users who subsequently input the same voice data to the preset dialogue model. Correspondingly, the second user refers to the user who inputs voice data similar to or the same as the voice data to the preset dialogue model again. Interaction satisfaction refers to the user's satisfaction with the answer content of the preset dialogue model to the user's voice data. Interaction satisfaction is the terminal's predicted estimate of the answer content of the preset dialogue model, not the user's actual satisfaction.

Generally, the terminal can use the preset sample data to train the preset dialogue model, and can perform real-time dialogue interaction with the first user based on the trained preset dialogue model. The preset sample data is the sample data that can be trained to obtain the existing preset dialogue model. The sample data is the text data of the known dialogue input and output, and the larger the data volume of the sample data, the higher the accuracy of the preset dialogue model that can be trained to perform human-computer dialogue interaction. After the first user inputs voice data to the terminal, the terminal can use the existing preset dialogue model to make a corresponding answer to the voice data input by the first user. That is, the first interaction result corresponding to the voice data input by the first user is obtained. At the same time, the terminal will obtain the first interaction result output by the preset dialogue model for the voice data input by the first user, and perform semantic analysis on the first interaction result to obtain the semantic information corresponding to the first interaction result, determine the similarity between the semantic information and the preset semantics, and based on the similarity, the interaction satisfaction of the first user can be further determined, that is, predict whether the user is satisfied with the dialogue interaction process.

For example: the first user interacts with a terminal with a preset dialogue model, and the first user inputs voice data "Help me open 'Xiaodu'" (voice data input by the first user) to the terminal. After the terminal analyzes the voice data based on the existing preset dialogue model, it cannot recognize the instruction corresponding to the voice data input by the first user, and then returns to the first user voice data containing "I'm sorry, I didn't understand what the host meant" (that is, the first interaction result). At the same time, the terminal will perform semantic analysis on the voice data returned to the first user, and obtain the semantic information corresponding to the first interaction result, which contains keywords such as "sorry, no" and/or negative semantics of keywords, while the preset semantics contains keywords such as "no, can't, no, didn't find" and/or negative semantics of keywords. Based on the analysis that the similarity between the semantic information corresponding to the first interaction result and the preset semantics is 80%, and then based on the similarity, it is determined that the interaction satisfaction of the first user is also 20%, and the preset threshold of interaction satisfaction is 60%, it can be determined that the first user is dissatisfied with the first interaction result.

S202: When the interaction satisfaction level is less than or equal to a preset threshold, voice data is sent to a server, and an interaction instruction corresponding to the voice data and sent by the server is received.

The preset threshold is a standard for determining whether the user is satisfied with the interaction result, that is, the minimum lower limit of the user's interaction satisfaction that needs to be met when the user is determined to be satisfied with the interaction result. Voice data refers to the conversation voice data input by the first user to the terminal. Interaction instructions refer to the interaction data used to analyze and/or answer the voice data.

Generally, when the terminal determines that the interaction satisfaction of the first user is less than or equal to the preset threshold, it indicates that the first user is not satisfied with the current interaction result. Then, the terminal will analyze the voice data input by the first user to obtain the keywords and/or keywords corresponding to the voice data, and based on the keywords and/or keywords, the first query data corresponding to the voice data can be generated, and the first query data can be sent to the server. The first query data is text data containing key information of the voice data generated based on the voice data input by the first user. The first query data can instruct the server to search for answer data related to the first query data on a preset website through the Internet. The preset website refers to a website that can provide search services or community services. The preset website can be set arbitrarily by the user, or it can be a default website of the server. After receiving the first query data sent by the terminal, the server can search for answer data corresponding to the first query data on a preset website through the Internet, and select the answer data with the highest matching degree or the highest confidence with the first query data from the searched multiple answer data as the interaction instruction corresponding to the voice data input by the first user, and send the interaction instruction to the terminal so that the terminal can obtain the interaction instruction.

For example: the first user interacts with a terminal with a preset dialogue model, and the first user inputs voice data "Help me open 'Xiaodu'" (voice data input by the first user) to the terminal. After the terminal analyzes the voice data based on the existing preset dialogue model, it cannot recognize the instruction corresponding to the voice data input by the first user, and then returns to the first user voice data containing "I'm sorry, I didn't understand what the host meant" (that is, the first interaction result). At the same time, the terminal will perform semantic analysis on the voice data returned to the first user, and obtain the semantic information corresponding to the first interaction result, which contains keywords such as "sorry, no" and/or negative semantics of keywords, while the preset semantics contains keywords such as "no, can't, no, didn't find" and/or negative semantics of keywords. Based on the analysis that the similarity between the semantic information corresponding to the first interaction result and the preset semantics is 80%, and then based on the similarity, it is determined that the interaction satisfaction of the first user is also 20%, and the preset threshold of interaction satisfaction is 60%, it can be determined that the first user is dissatisfied with the first interaction result. Furthermore, when the terminal determines that the first user is dissatisfied with the first interaction result, the terminal can analyze the voice data "Help me open 'Xiaodu'" input by the first user to obtain keywords such as "open, Xiaodu", and based on the keywords, can generate text data "What does 'open Xiaodu' mean" (first query data) for querying the voice data input by the first user. The first query data is sent to the server, and the server can search for answer data corresponding to the first query data on a preset website based on the first query data, select the answer data "Open Baidu application" with the highest confidence based on the confidence corresponding to each searched answer data, and use the answer data as the interaction instruction for the voice data "Help me open 'Xiaodu'" input by the first user.

S203: Optimize and train a preset dialogue model by using the interaction instructions and voice data as sample data.

The sample data is data used to train a dialogue model that can realize human-computer dialogue. It can be sample data in text and/or voice form, and the sample data includes dialogue data from the user's perspective and the dialogue model's perspective. The preset dialogue model refers to a model that can interact with the user through human-computer dialogue. It can be trained from a large amount of sample data based on a preset dialogue algorithm. The user can interact with the terminal with the preset dialogue model in real time.

Generally, after searching for answer data matching the voice data input by the first user on a preset website, the server will use the answer data as an interaction instruction corresponding to the voice data and send the interaction instruction to the terminal. After receiving the interaction instruction, the terminal will associate the interaction instruction with the voice data and store it in the database as sample data to expand and update the sample data used to train the preset dialogue model. Subsequently, the preset dialogue model can be optimized and trained based on the sample data to obtain an optimized preset dialogue model. The next time the optimized preset dialogue model encounters voice data similar to the voice data input by the first user, it will make a corresponding answer and/or perform a corresponding processing operation based on the interaction instruction corresponding to the voice data input by the first user that the server searched for last time.

From the above content, it can be seen that in the dialogue interaction method provided by the present scheme, the electronic device analyzes the interaction satisfaction of the first user based on the first interaction result, the first interaction result is the interaction result output by the preset dialogue model for the voice data input by the first user, and when the interaction satisfaction is less than or equal to the preset threshold, the voice data is sent to the server, and the interaction instructions corresponding to the voice data sent by the server are received, and the interaction instructions and the voice data are used as sample data to optimize and train the preset dialogue model, and the new sample data for optimizing and training the preset dialogue model is obtained by searching the answer data corresponding to the voice data by the server, thereby reducing the workload of manual participation and making the process of optimizing and training the preset dialogue model more intelligent.

Please refer to Figure 3, which is another flow chart of a conversation interaction method provided in an embodiment of the present application. This embodiment is illustrated by applying a conversation interaction method to an electronic device (terminal). The conversation interaction method may include the following steps:

S301: Obtain a first interaction result.

The first interaction result is the interaction result output by the preset dialogue model for the voice data input by the first user, that is, the dialogue content data generated by the preset dialogue model. The first interaction result can be text data or voice data.

Generally, the terminal can use the preset sample data to train a preset dialogue model, and can perform real-time dialogue interaction with the first user based on the trained preset dialogue model. After the first user inputs voice data to the terminal, the terminal can use the existing preset dialogue model to make a corresponding response to the voice data input by the first user. That is, the first interaction result corresponding to the voice data input by the first user is obtained. The terminal can analyze whether the user is satisfied with the dialogue interaction process based on the first interaction result.

S302: Perform semantic analysis on the first interaction result to obtain first semantic information corresponding to the first interaction result.

The first semantic information refers to the semantics contained in the first interaction result, that is, the semantic information of the corresponding answer given by the terminal to the voice data input by the first user when the terminal interacts with the first user based on the existing preset dialogue model.

Generally, semantic analysis can be used to derive a formalized representation that can reflect the meaning of a sentence based on the syntactic structure of the sentence and the meaning of each content word in the sentence, and convert the natural language that humans can understand into a formal language that computers can understand. The first interaction result can be voice data or text data. If the first interaction result is voice data, the first interaction result is converted into text to obtain corresponding text data, and the text data is semantically analyzed to obtain semantic information corresponding to the first interaction result.

S303: Determine the interaction satisfaction of the first user based on the first semantic information.

The first user refers to the user who inputs the voice data to the preset dialogue model for the first time. "First" is used to distinguish users who subsequently input the same voice data to the preset dialogue model. Correspondingly, the second user refers to the user who inputs voice data similar to or the same as the voice data to the preset dialogue model again. Interaction satisfaction refers to the user's satisfaction with the answer content of the preset dialogue model to the user's voice data. Interaction satisfaction is the terminal's prediction of the user's satisfaction based on the answer content of the preset dialogue model, not the user's actual satisfaction.

Generally, after the terminal performs semantic analysis on the first interaction result, it can obtain the first semantic information corresponding to the first interaction result. Further analysis of the first semantic information can determine the interaction satisfaction of the first user. The terminal is provided with preset semantics, which are usually semantics containing negative meanings. By analyzing the similarity between the first semantic information and the preset semantics, it is possible to predict the interaction satisfaction corresponding to the similarity, that is, the interaction satisfaction of the first user. Generally, the user's interaction satisfaction is negatively correlated with the semantic similarity. The higher the semantic similarity, the lower the user's interaction satisfaction.

For example: when the similarity between the first semantic information and the preset semantics is 0% to 20%, the corresponding interaction satisfaction is 80% to 100%; when the similarity between the first semantic information and the preset semantics is 21% to 50%, the corresponding interaction satisfaction is 60% to 79%; when the similarity between the first semantic information and the preset semantics is 51% to 80%, the corresponding interaction satisfaction is 20% to 59%; when the similarity between the first semantic information and the preset semantics is 81% to 100%, the corresponding interaction satisfaction is 0% to 19%.

S304: When the interaction satisfaction is less than or equal to a preset threshold, the voice data is analyzed to obtain keywords and/or key words corresponding to the voice data.

The preset threshold is a standard for determining whether the user is satisfied with the interaction result. That is, when it is determined that the user is satisfied with the interaction result, the user's interaction satisfaction needs to meet the minimum lower limit of the interaction satisfaction. The voice data here refers to the conversation voice data input by the first user into the terminal, that is, the voice data input by the first user when the first user's interaction satisfaction is less than or equal to the preset threshold.

Generally, when the interaction satisfaction of the first user is less than or equal to a preset threshold, it can be determined that the first user is dissatisfied with the conversation interaction process, and the terminal can analyze the voice data input by the first user to obtain corresponding keywords and/or key words, which can extract and reflect the key content of the voice data input by the first user, and based on the keywords and/or key words, query data corresponding to the voice data input by the first user can be generated.

S305, generating first query data corresponding to the voice data based on the keywords and/or key words, sending the first query data to the server, and receiving an interaction instruction corresponding to the voice data sent by the server.

Among them, voice data refers to the conversation voice data input by the first user into the terminal, that is, the voice data input by the first user when the interaction satisfaction of the first user is less than or equal to a preset threshold; first query data refers to the query data generated by the terminal based on the voice data input by the first user, and the first query data can instruct the server to search for answer data related to the first query data on a preset website through the Internet; interaction instructions refer to interaction data used to analyze and/or answer the voice data.

Generally, when the terminal determines that the interaction satisfaction of the first user is less than or equal to a preset threshold, it indicates that the first user is not satisfied with the current interaction result, and the terminal will analyze the voice data input by the first user to obtain keywords and/or keywords corresponding to the voice data. Based on the keywords and/or keywords, first query data corresponding to the voice data can be generated, and the first query data can be sent to the server. The first query data is text data containing key information of the voice data generated based on the voice data input by the first user. The first query data can instruct the server to search for answer data related to the first query data on a preset website through the Internet. The preset website refers to a website that can provide search services or community services. The preset website can be set arbitrarily by the user, or it can be a default website of the server. After receiving the first query data sent by the terminal, the server can search for answer data corresponding to the first query data on a preset website through the Internet, and select the answer data with the highest matching degree or the highest confidence with the first query data from the searched multiple answer data as the interaction instruction corresponding to the voice data input by the first user, and send the interaction instruction to the terminal so that the terminal can obtain the interaction instruction.

For example: the first user interacts with a terminal with a preset dialogue model, and the first user inputs voice data "Help me open 'Xiaodu'" (voice data input by the first user) to the terminal. After the terminal analyzes the voice data based on the existing preset dialogue model, it cannot recognize the instruction corresponding to the voice data input by the first user, and then returns to the first user voice data containing "I'm sorry, I didn't understand what the host meant" (that is, the first interaction result). At the same time, the terminal will perform semantic analysis on the voice data returned to the first user, and obtain the semantic information corresponding to the first interaction result as containing keywords such as "sorry, no" and/or negative semantics of keywords, while the preset semantics are containing keywords such as "no, can't, no, didn't find" and/or negative semantics of keywords. Based on the analysis that the similarity between the semantic information corresponding to the first interaction result and the preset semantics is 80%, and then based on the similarity, it is determined that the interaction satisfaction of the first user is also 20%, and the preset threshold of interaction satisfaction is 60%, it can be determined that the first user is dissatisfied with the first interaction result. Furthermore, when the terminal determines that the first user is dissatisfied with the first interaction result, the terminal can analyze the voice data "Help me open 'Xiaodu'" input by the first user to obtain keywords such as "open, Xiaodu", and based on the keywords, can generate text data "What does 'open Xiaodu' mean" (first query data) for querying the voice data input by the first user. The first query data is sent to the server, and the server can search for answer data corresponding to the first query data on a preset website based on the first query data, select the answer data "Open Baidu application" with the highest confidence based on the confidence corresponding to each searched answer data, and use the answer data as the interaction instruction for the voice data "Help me open 'Xiaodu'" input by the first user.

S306: Use the interaction instructions and voice data as sample data to optimize and train the preset dialogue model.

For details, please refer to the above step S203, which will not be described again here.

S307: Use the optimized preset dialogue model to analyze the voice data input by the second user and output a second interaction result.

Among them, the second user refers to a user who inputs the same or similar voice data as the voice data input by the first user to the optimized preset dialogue model again. The second user can be the first user who inputs the same or similar voice data as the first time to the optimized preset dialogue model for the second time, or a user who inputs voice data similar to or the same as the voice data input by the first user to the optimized preset dialogue model and is different from the first user. The voice data input by the second user refers to voice data that is the same or similar to the voice data input by the first user to the optimized preset dialogue model. The second interaction result refers to the interaction result output by the optimized preset dialogue model for the voice data input by the second user. That is, the dialogue content data generated by the optimized preset dialogue model corresponding to the voice data input by the second user. The second interaction result can be text data or voice data.

Generally, after optimizing and training the preset dialogue model based on the above steps S301 to S306, the terminal can interact with a new user or a first user based on the optimized preset dialogue model. The user interacting with the optimized preset dialogue model is called the second user. When interacting with the optimized preset dialogue model, voice data that is the same or similar to the voice data input by the first user may be encountered. The optimized preset dialogue model will analyze the voice data input by the second user and output a second interaction result corresponding to the voice data. The second interaction result is based on the interactive instruction corresponding to the voice data input by the first user searched on the preset website by the server through the Internet, and the corresponding generated dialogue content data for responding to the voice data input by the second user. The preset dialogue model can be optimized more intelligently, and the accuracy of the preset model in recognizing user instructions can be improved without the participation of the user. If it is predicted that the user is still dissatisfied with the dialogue content data, the terminal can send the voice data currently input by the second user to the server, so that the server can publish the question data corresponding to the voice data on the preset website, and collect the answer data of other netizens for the question data in a similar way to crowdfunding answers.

S308: Perform semantic analysis on the second interaction result to obtain second semantic information corresponding to the second interaction result.

The second semantic information refers to the semantics contained in the second interaction result, that is, the semantic information of the corresponding response made by the terminal to the voice data input by the second user when the terminal interacts with the second user based on the optimized preset dialogue model.

Generally, semantic analysis can be used to derive a formalized representation that can reflect the meaning of a sentence based on the syntactic structure of the sentence and the meaning of each content word in the sentence, and convert the natural language that humans can understand into a formal language that computers can understand. The second interaction result can be voice data or text data. If the second interaction result is voice data, the second interaction result is converted into text to obtain the corresponding text data, and the text data is semantically analyzed to obtain semantic information corresponding to the second interaction result.

S309: Determine the interaction satisfaction of the second user based on the second semantic information.

Among them, the interaction satisfaction of the second user refers to the second user's satisfaction with the dialogue interaction of the optimized dialogue model. The interaction satisfaction is the terminal's predicted estimate of the user's satisfaction based on the answer content of the preset dialogue model, not the user's actual satisfaction.

Generally, after the terminal performs semantic analysis on the second interaction result, it can obtain the second semantic information corresponding to the second interaction result. Further analysis of the second semantic information can predict the interaction satisfaction of the second user. The terminal is provided with preset semantics, which are usually semantics containing negative meanings. By analyzing the similarity between the second semantic information and the preset semantics, the interaction satisfaction corresponding to the similarity can be predicted, that is, the interaction satisfaction of the second user. Generally, the user's interaction satisfaction is negatively correlated with the semantic similarity. The higher the semantic similarity, the lower the user's interaction satisfaction.

S310: When the interaction satisfaction of the second user is less than or equal to a preset threshold, second query data is sent to the server, and a target query result sent by the server is received.

Among them, the second query data refers to the query data generated based on the keywords and/or key words corresponding to the voice data input by the second user. The second query data is also used to indicate the question data published by the server on the preset website, so that the server can collect answer data for the question data from other netizens on the preset website. The target query result is the query result corresponding to the second query data queried by the server on the preset website, that is, the answer data that best matches the second query data selected/queried by the server from the published question data as the query result. The second query data corresponds to multiple query results, and different query results correspond to different confidences. The target query result is the query result with the highest confidence among the multiple query results. The confidence is input by multiple users based on the preset website. The confidence is the support rate/like rate of other website users for the question data, etc. The preset website refers to a website that can provide search services or community services. It can be set arbitrarily by the user or it can be the default website of the server.

Generally, when the terminal determines that the interaction satisfaction of the second user is less than or equal to a preset threshold, it indicates that the second user is dissatisfied with the current interaction result. Then, the terminal will analyze the voice data input by the second user to obtain keywords and/or key words corresponding to the voice data. Based on the keywords and/or key words, second query data corresponding to the voice data can be generated, i.e., question data for indicating the server to publish on a preset website. The terminal will send the second query data to the server so that the server can collect answer data for the question data from other netizens on the preset website. The preset website can be set arbitrarily by the user or can be the default website of the server. Since the server actively publishes question data on the preset website to seek corresponding The server needs a certain amount of waiting time to wait for other website users to answer the question data. The server can query the answer data corresponding to the published question data based on the preset time interval, and the question data may correspond to multiple answer data, that is, the server will query multiple query results, and different query results correspond to different confidence levels. Therefore, the server will select the query result with the highest confidence level from the query results that can be queried as the target query result each time, and send the target query result obtained by the query to the terminal. After receiving the target query result sent by the server, the terminal will use the target query result and the voice data input by the second user as new sample data to optimize and train the optimized preset dialogue model again.

For example: the first user interacts with a terminal with a preset dialogue model, and the first user inputs voice data "Help me open 'Xiaodu'" (voice data input by the first user) to the terminal. After the terminal analyzes the voice data based on the existing preset dialogue model, it cannot recognize the instruction corresponding to the voice data input by the first user, and then returns to the first user voice data containing "I'm sorry, I didn't understand what the host meant" (that is, the first interaction result). At the same time, the terminal will perform semantic analysis on the voice data returned to the first user, and obtain the semantic information corresponding to the first interaction result, which contains keywords such as "sorry, no" and/or negative semantics of keywords, while the preset semantics contains keywords such as "no, can't, no, didn't find" and/or negative semantics of keywords. Based on the analysis that the similarity between the semantic information corresponding to the first interaction result and the preset semantics is 80%, and then based on the similarity, it is determined that the interaction satisfaction of the first user corresponds to 20%, and the preset threshold of interaction satisfaction is 60%, it can be determined that the first user is dissatisfied with the first interaction result. Furthermore, when the terminal determines that the first user is dissatisfied with the first interaction result, the terminal may analyze the voice data "Help me open 'Xiaodu'" input by the first user to obtain keywords such as "open, Xiaodu", and based on the keywords, the terminal may generate text data "What does 'open Xiaodu' mean" (first query data) for querying the voice data input by the first user. The first query data is sent to the server, and the server may search for answer data corresponding to the first query data on a preset website based on the first query data, and select the answer data "Open Baidu Application" with the highest confidence based on the confidence corresponding to each searched answer data, and use the answer data as the interaction instruction for the voice data "Help me open 'Xiaodu'" input by the first user. The terminal will associate the interaction instruction with the voice data input by the first user and use it as sample data to optimize and train the existing preset dialogue model to obtain a prioritized preset dialogue model.

When the second user inputs voice data "Use 'Xiaodu' to navigate" into the optimized preset dialogue model, the voice data is similar to the voice data input by the first user. However, after the terminal analyzes the voice data based on the optimized preset dialogue model, it still cannot recognize the instruction corresponding to the voice data input by the second user. Then, the corresponding voice data containing "Sorry, I can't find 'Xiaodu' for navigation" (i.e., the second interaction result) is returned to the second user. At the same time, the terminal performs semantic analysis on the second interaction result to obtain semantic information corresponding to the second interaction result, including keywords such as "sorry, no" and/or negative semantics of keywords. The similarity between the semantics and the preset semantics is calculated to be 80%. Based on the similarity, the second user's interaction satisfaction is determined to be 20%, and the preset threshold of interaction satisfaction is 60%. It can be determined that the second user is dissatisfied with the second interaction result. Furthermore, when the terminal determines that the second user is dissatisfied with the second interaction result, the terminal can analyze the voice data "Use 'Xiaodu' to navigate" input by the second user to obtain keywords such as "Xiaodu, navigation", and based on the keywords, can generate text data for querying the voice data input by the second user "What navigation is used by 'Xiaodu Navigation'? How to use Xiaodu Navigation?" (second query data), that is, for indicating the question data published by the server on the preset website. The terminal will send the second query data to the server, and the server can publish the question data "What navigation is used by 'Xiaodu Navigation'? How to use Xiaodu Navigation?" on the preset website, waiting for other website users to answer the question; the server will query the answer data corresponding to the published question data on the preset website at intervals of every three days, and the server will query 10 answer data corresponding to the question data (query results). The confidence levels of the 10 query results are different. The confidence level is the support rate/like rate of other website users for the question data. The server will select the query result with the highest confidence level from these 10 query results, "'Xiaodu Navigation' uses Baidu Maps for navigation. After opening Baidu Maps, you can enter the destination in the corresponding search bar. Baidu Maps will plan a navigation route to your destination based on your current location" as the target query result, and send the target query result to the terminal. The terminal will associate the target query result with the voice data input by the second user and use it as new sample data to optimize and train the optimized preset dialogue model again.

S311, using the target query result and the voice data as sample data to optimize and train the optimized preset dialogue model.

Generally, after the terminal receives the target query result sent by the server, it will associate the target query result with the voice data input by the second user and store it in the database as new sample data, so as to expand and update the sample data used to train the preset dialogue model; and the optimized preset dialogue model can be optimized again based on the new sample data, so that when the re-optimized preset dialogue model recognizes the voice data similar to or identical to the voice data input by the second user again, it can make a corresponding answer and/or perform a corresponding processing operation based on the target query result.

From the above content, it can be seen that in the dialogue interaction method provided by the present solution, the electronic device obtains a first interaction result, performs semantic analysis on the first interaction result to obtain first semantic information corresponding to the first interaction result, determines the interaction satisfaction of the first user based on the first semantic information, and when the interaction satisfaction is less than or equal to a preset threshold, analyzes the voice data to obtain keywords and/or keywords corresponding to the voice data, generates first query data corresponding to the voice data based on the keywords and/or keywords, sends the first query data to the server, receives interaction instructions corresponding to the voice data sent by the server, uses the interaction instructions and the voice data as sample data to optimize and train a preset dialogue model, and uses the optimized preset dialogue model to perform training on the second user input The method comprises the steps of: analyzing the input voice data, outputting a second interaction result, performing semantic analysis on the second interaction result to obtain second semantic information corresponding to the second interaction result, determining the interaction satisfaction of the second user based on the second semantic information, and when the interaction satisfaction of the second user is less than or equal to a preset threshold, sending second query data to the server, receiving the target query result sent by the server, and optimizing and training the optimized preset dialogue model by using the target query result and the voice data as sample data. In this way, the sample data used to train the preset dialogue model is updated and/or expanded, so that the preset dialogue model can continuously learn and optimize based on the dialogue interaction process with the user, thereby improving the learning ability of the preset dialogue model and the accuracy of human-computer dialogue.

The following are device embodiments of the present application, which can be used to execute the method embodiments of the present application. For details not disclosed in the device embodiments of the present application, please refer to the method embodiments of the present application.

Please refer to FIG. 4, which shows a schematic diagram of the structure of a conversation interaction device provided by an exemplary embodiment of the present application, hereinafter referred to as device 4. Device 4 can be implemented as all or part of an electronic device through software, hardware, or a combination of both. Device 4 includes:

An analysis module 401 is used to analyze the interaction satisfaction of the first user according to a first interaction result; wherein the first interaction result is an interaction result output by a preset dialogue model for the voice data input by the first user;

The processing module 402 is used to send the voice data to the server when the interaction satisfaction is less than or equal to a preset threshold, and receive an interaction instruction corresponding to the voice data sent by the server;

The training module 403 is used to optimize and train the preset dialogue model by using the interaction instruction and the voice data as sample data.

Optionally, the analysis module 401 includes:

An acquiring unit, configured to acquire the first interaction result;

A first analysis unit, configured to perform semantic analysis on the first interaction result to obtain first semantic information corresponding to the first interaction result;

A first determining unit is configured to determine the interaction satisfaction of the first user based on the first semantic information.

Optionally, the processing module 402 includes:

A second analysis unit, configured to analyze the voice data to obtain keywords and/or key words corresponding to the voice data;

The first processing unit is configured to generate first query data corresponding to the voice data based on the keyword and/or key word, and send the first query data to the server.

Optionally, the training module 403 further includes:

A second processing unit, configured to analyze the voice data input by the second user by using the optimized preset dialogue model, and output a second interaction result;

a third analyzing unit, configured to analyze the interaction satisfaction of the second user according to the second interaction result;

a third processing unit, configured to send second query data to a server and receive a target query result sent by the server when the interaction satisfaction of the second user is less than or equal to the preset threshold; wherein the target query result is a query result corresponding to the second query data queried by the server on a preset website; and the second query data is obtained according to the keywords and/or key words corresponding to the voice data;

A training unit is used to optimize and train the optimized preset dialogue model by using the target query result and the voice data as sample data.

Optionally, the training module 403 further includes:

a fourth analyzing unit, configured to perform semantic analysis on the second interaction result to obtain second semantic information corresponding to the second interaction result;

A second determining unit is configured to determine the interaction satisfaction of the second user based on the second semantic information.

It should be noted that the conversation interaction device provided in the above embodiment only uses the division of the above functional modules as an example when executing the conversation interaction method. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the conversation interaction device provided in the above embodiment and the conversation interaction method embodiment belong to the same concept, and the implementation process thereof is detailed in the method embodiment, which will not be repeated here.

The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

The embodiment of the present application also provides a computer storage medium, which can store multiple instructions, and the instructions are suitable for being loaded by a processor and executing the method steps as described above. The specific execution process can be found in the specific description of the embodiments shown in Figures 2 to 3, and will not be repeated here.

The present application also provides an electronic device, including a processor, a memory and a display screen; wherein the memory stores a computer program, and the computer program is suitable for being loaded by the processor and executing the above-mentioned method steps.

Please refer to FIG5 , which is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. As shown in FIG5 , the electronic device 500 may include: at least one processor 501 , at least one network interface 504 , a user interface 503 , a memory 505 , and at least one communication bus 502 .

The communication bus 502 is used to realize the connection and communication between these components.

The user interface 503 may include a display screen (Display) and a camera (Camera), and the optional user interface 503 may also include a standard wired interface and a wireless interface.

The network interface 504 may optionally include a standard wired interface or a wireless interface (such as a WI-FI interface).

Among them, the processor 501 may include one or more processing cores. The processor 501 uses various interfaces and lines to connect various parts of the entire electronic device 500, and executes various functions and processes data of the electronic device 500 by running or executing instructions, programs, code sets or instruction sets stored in the memory 505, and calling data stored in the memory 505. Optionally, the processor 501 can be implemented in at least one hardware form of digital signal processing (Digital Signal Processing, DSP), field programmable gate array (Field-Programmable Gate Array, FPGA), and programmable logic array (Programmable Logic Array, PLA). The processor 501 can integrate one or a combination of a central processing unit (Central Processing Unit, CPU), a graphics processing unit (Graphics Processing Unit, GPU) and a modem. Among them, the CPU mainly processes the operating system, user interface and application programs; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; the modem is used to process wireless communication. It can be understood that the above-mentioned modem may not be integrated into the processor 501, and it can be implemented separately through a chip.

Among them, the memory 505 may include a random access memory (RAM) or a read-only memory (Read-Only Memory). Optionally, the memory 505 includes a non-transitory computer-readable storage medium. The memory 505 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 505 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area may store data involved in the above-mentioned various method embodiments, etc. The memory 505 may also be optionally at least one storage device located away from the aforementioned processor 501. As shown in Figure 5, the memory 505 as a computer storage medium may include an operating system, a network communication module, a user interface module, and a dialogue interaction application.

In the electronic device 500 shown in FIG. 5 , the user interface 503 is mainly used to provide an input interface for the user and obtain data input by the user; and the processor 501 can be used to call the dialogue interaction application stored in the memory 505 and specifically perform the following operations:

Analyzing the interaction satisfaction of the first user according to the first interaction result; wherein the first interaction result is the interaction result output by a preset dialogue model for the voice data input by the first user;

When the interaction satisfaction is less than or equal to a preset threshold, sending the voice data to a server, and receiving an interaction instruction corresponding to the voice data sent by the server;

The interaction instruction and the voice data are used as sample data to optimize and train the preset dialogue model.

In one embodiment, when analyzing the interaction satisfaction of the first user according to the first interaction result, the processor 501 specifically performs the following operations:

Obtaining the first interaction result;

Performing semantic analysis on the first interaction result to obtain first semantic information corresponding to the first interaction result;

The interaction satisfaction of the first user is determined based on the first semantic information.

In one embodiment, when executing sending the voice data to the server, the processor 501 further performs the following operations:

Analyzing the voice data to obtain keywords and/or key words corresponding to the voice data;

First query data corresponding to the voice data is generated based on the keyword and/or key word, and the first query data is sent to the server.

In one embodiment, after executing the optimization training of the preset dialogue model by using the interaction instruction and the voice data as sample data, the processor 501 further performs the following operations:

Analyze the voice data input by the second user using the optimized preset dialogue model, and output a second interaction result;

analyzing the interaction satisfaction of the second user according to the second interaction result;

When the interaction satisfaction of the second user is less than or equal to the preset threshold, sending second query data to the server, and receiving a target query result sent by the server; wherein the target query result is a query result corresponding to the second query data queried by the server on a preset website; the second query data is obtained according to the keywords and/or key words corresponding to the voice data;

The target query result and the voice data are used as sample data to perform optimization training on the optimized preset dialogue model.

In an embodiment, when analyzing the interaction satisfaction of the second user according to the second interaction result, the processor 501 further performs the following operations:

Performing semantic analysis on the second interaction result to obtain second semantic information corresponding to the second interaction result;

The interaction satisfaction of the second user is determined based on the second semantic information.

In an embodiment of the present application, the electronic device analyzes the interaction satisfaction of the first user based on a first interaction result, where the first interaction result is an interaction result output by a preset dialogue model for voice data input by the first user. When the interaction satisfaction is less than or equal to a preset threshold, the electronic device sends voice data to a server, receives interaction instructions corresponding to the voice data sent by the server, uses the interaction instructions and the voice data as sample data to optimize and train the preset dialogue model, and obtains new sample data for optimizing and training the preset dialogue model by searching the server for answer data corresponding to the voice data, thereby reducing the workload of manual participation and making the process of optimizing and training the preset dialogue model more intelligent.

Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

The above description is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims (9)

1. A conversation interaction method, characterized in that the method comprises: Obtaining a first interaction result; wherein the first interaction result is an interaction result output by a preset dialogue model for voice data input by a first user; Performing semantic analysis on the first interaction result to obtain first semantic information corresponding to the first interaction result; Determining the interaction satisfaction of the first user based on the similarity between the first semantic information and the preset semantics; wherein the interaction satisfaction is the predicted satisfaction of the first user with the first interaction result, the preset semantics includes semantics with negative meanings, and the interaction satisfaction is negatively correlated with the preset semantics; When the interaction satisfaction is less than or equal to a preset threshold, sending the voice data to a server, and receiving an interaction instruction corresponding to the voice data sent by the server; The interaction instruction and the voice data are used as sample data to optimize and train the preset dialogue model. 2. The method according to claim 1, wherein sending the voice data to the server comprises: Analyzing the voice data to obtain keywords and/or key words corresponding to the voice data; First query data corresponding to the voice data is generated based on the keyword and/or key word, and the first query data is sent to the server. 3. The method according to claim 1, characterized in that after optimizing and training the preset dialogue model using the interaction instruction and the voice data as sample data, it further comprises: Analyze the voice data input by the second user using the optimized preset dialogue model, and output a second interaction result; analyzing the interaction satisfaction of the second user according to the second interaction result; When the interaction satisfaction of the second user is less than or equal to the preset threshold, sending second query data to the server, and receiving a target query result sent by the server; wherein the target query result is a query result corresponding to the second query data queried by the server on a preset website; the second query data is obtained according to the keywords and/or key words corresponding to the voice data; The target query result and the voice data are used as sample data to perform optimization training on the optimized preset dialogue model. 4. The method according to claim 3, characterized in that analyzing the interaction satisfaction of the second user according to the second interaction result comprises: Performing semantic analysis on the second interaction result to obtain second semantic information corresponding to the second interaction result; The interaction satisfaction of the second user is determined based on the second semantic information. 5. The method according to claim 3 is characterized in that the second query data corresponds to multiple query results, different query results correspond to different confidence levels, and the target query result is the query result with the highest confidence level among the multiple query results. 6 . The method according to claim 5 , wherein the confidence level is input by multiple users based on the preset website. 7. A conversation interaction device, characterized in that the device comprises: An acquisition module, configured to acquire a first interaction result; wherein the first interaction result is an interaction result output by a preset dialogue model for voice data input by a first user; An analysis module, configured to perform semantic analysis on the first interaction result to obtain first semantic information corresponding to the first interaction result; a determination module, configured to determine the interaction satisfaction of the first user based on the similarity between the first semantic information and the preset semantics; wherein the interaction satisfaction is the predicted satisfaction of the first user with the first interaction result, the preset semantics includes semantics with negative meanings, and the interaction satisfaction is negatively correlated with the preset semantics; A processing module, configured to send the voice data to a server when the interaction satisfaction is less than or equal to a preset threshold, and receive an interaction instruction corresponding to the voice data sent by the server; A training module is used to optimize and train the preset dialogue model by using the interaction instruction and the voice data as sample data. 8. A computer storage medium, characterized in that the computer storage medium stores a plurality of instructions, wherein the instructions are suitable for being loaded by a processor and executing the method steps as claimed in any one of claims 1 to 6. 9. An electronic device, characterized in that it comprises: a processor, a memory and a display screen; wherein the memory stores a computer program, and the computer program is suitable for being loaded by the processor and executing the method steps as claimed in any one of claims 1 to 6.