Human-computer interaction method, device, equipment, medium and product based on large model

Abstract

The present application provides a human-computer interaction method, device, equipment, medium and product based on a large model, the method comprising: obtaining a target question input by a user; inputting the target question into a preset intent recognition model to generate a corresponding user interaction intention; if the user interaction intention is a business-related intention, using a small business model to generate corresponding target information based on the target question; the small business model is generated based on a large model and training samples; the training samples are related to preset business data; executing a processing flow related to the target information according to the target information, generating a first processing result, and feeding back the first processing result to the user. The human-computer interaction method based on a large model of the present application can realize the function of the large model to parse business-related issues, so that users can get answers to business-related questions through the large model, thereby improving the user's interactive experience.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a large model-based human-computer interaction method, device, equipment, medium and product. Background Art

[0002] Large Language Model (LLM), also known as a large model, is a natural language processing (NLP) model based on deep learning technology, usually with billions or even tens of billions of parameters. These models are trained on large amounts of text data to generate and understand natural language texts and perform various language tasks such as text generation, translation, question answering, summarization, sentiment analysis, etc.

[0003] Currently, during human-computer conversations, large language models lack the ability to use industry data or business data to analyze problems.

[0004] Therefore, there is currently a lack of a human-computer interaction method that can parse business-related issues. Summary of the invention

[0005] The present application provides a large model-based human-computer interaction method, device, equipment, medium and product to solve the problem of the current lack of a human-computer interaction method that can analyze business-related issues.

[0006] The first aspect of the present application provides a human-computer interaction method based on a large model, comprising:

[0007] Get the target question input by the user;

[0008] Input the target question into a preset intention recognition model to generate a corresponding user interaction intention;

[0009] If the user interaction intention is a business-related intention, a small business model is used to generate corresponding target information based on the target problem; the small business model is generated based on the large model and training samples; the training samples are related to preset business data;

[0010] A processing flow related to the target information is executed according to the target information, a first processing result is generated, and the first processing result is fed back to a user.

[0011] Furthermore, in the above method, the step of using the small business model to generate corresponding target information based on the target problem includes:

[0012] If the target problem is an operating system-related problem, a small business model is used to identify the operating system-related problem, generate corresponding device information, and use the device information as the target information;

[0013] If the target problem is a problem related to business data, a small business model is used to identify the problem related to the business data, generate corresponding business information and corresponding parameter information, and use the business information and the parameter information as the target information.

[0014] Furthermore, in the above method, executing a processing flow related to the target information according to the target information to generate a first processing result includes:

[0015] A preset function calling tool is used to call a corresponding function in the large model based on the target information to execute a processing flow corresponding to the target information.

[0016] Furthermore, in the above method, if the user interaction intention is a professional knowledge query intention, the method further includes:

[0017] A small knowledge model is used to search for matching knowledge information from a preset professional knowledge database based on the target problem; the small knowledge model is generated based on the large model and training samples related to the preset professional knowledge; the preset professional knowledge database stores relevant professional knowledge of the coal industry;

[0018] Using a large model to perform data analysis and data sorting on the target problem and the knowledge information to generate a second processing result;

[0019] Determine whether it is necessary to call data in the knowledge graph database according to the second processing result;

[0020] If it is determined that there is no need to call the data in the knowledge graph database, the second processing result is fed back to the user;

[0021] If it is determined that data in the knowledge graph database needs to be called, entity extraction processing is performed on the target question to generate entity data to be queried;

[0022] Acquire corresponding knowledge graph data from the knowledge graph database according to the entity data;

[0023] The knowledge graph data and the target question are input into a large model, an interaction result corresponding to the target question is generated, and the interaction result is fed back to the user.

[0024] Furthermore, in the above method, if the user interaction intention is a general query intention, the method further includes:

[0025] Get the historical conversation messages of this round;

[0026] Input the target question and the historical conversation messages of this round into the big model to generate corresponding query result data;

[0027] Calling a preset data parsing tool to convert the data format of the query result data into a standard data format required for display;

[0028] The converted query result data is displayed in the preset display area.

[0029] Furthermore, the method as described above further comprises:

[0030] When data or process fails, corresponding fault information is generated;

[0031] Send the fault information and the recorded process execution information to the big model to generate corresponding retry related information; the retry related information is to re-execute the entire process or to re-execute the process from a certain process in the middle;

[0032] Execute the corresponding process according to the retry related information.

[0033] The second aspect of the present application provides a human-computer interaction device based on a large model, comprising:

[0034] An acquisition module is used to obtain the target question input by the user;

[0035] A first generating module, used to input the target question into a preset intention recognition model to generate a corresponding user interaction intention;

[0036] A second generation module is used to generate corresponding target information based on the target problem using a small business model if the user interaction intention is a business-related intention; the small business model is generated based on the large model and training samples; the training samples are related to preset business data;

[0037] The feedback module is used to execute a processing flow related to the target information according to the target information, generate a first processing result, and feed back the first processing result to a user.

[0038] Furthermore, in the above-mentioned device, when the second generating module uses the small business model to generate corresponding target information based on the target problem, it is specifically used to:

[0039] If the target problem is an operating system-related problem, a small business model is used to identify the operating system-related problem, generate corresponding device information, and use the device information as the target information; if the target problem is a business data-related problem, a small business model is used to identify the business data-related problem, generate corresponding business information and corresponding parameter information, and use the business information and the parameter information as the target information.

[0040] Further, in the above-mentioned device, when the feedback module executes the processing flow related to the target information according to the target information to generate the first processing result, it is specifically used to:

[0041] A preset function calling tool is used to call a corresponding function in the large model based on the target information to execute a processing flow corresponding to the target information.

[0042] Furthermore, in the above device, if the user interaction intention is a professional knowledge query intention, the device further includes:

[0043] A professional knowledge processing module is used to use a small knowledge model to search for matching knowledge information from a preset professional knowledge database based on the target problem; the small knowledge model is generated based on the large model and training samples related to the preset professional knowledge; the preset professional knowledge database stores relevant professional knowledge of the coal industry; the large model is used to perform data analysis and data organization on the target problem and the knowledge information to generate a second processing result; based on the second processing result, it is determined whether it is necessary to call the data in the knowledge graph database; if it is determined that it is not necessary to call the data in the knowledge graph database, the second processing result is fed back to the user; if it is determined that the data in the knowledge graph database needs to be called, entity extraction processing is performed on the target problem to generate entity data to be queried; corresponding knowledge graph data is obtained from the knowledge graph database based on the entity data; the knowledge graph data and the target problem are input into the large model to generate an interactive result corresponding to the target problem, and the interactive result is fed back to the user.

[0044] Furthermore, in the above device, if the user interaction intention is a general query intention, the device further includes:

[0045] The general processing module is used to obtain the historical conversation messages of this round; input the target question and the historical conversation messages of this round into the big model to generate corresponding query result data; call the preset data parsing tool to convert the data format of the query result data into the standard data format required for display; and display the converted query result data in the preset display area.

[0046] Furthermore, the device as described above further comprises:

[0047] The retry module is used to generate corresponding fault information when a data or process fails; send the fault information and the recorded process execution information to the large model to generate corresponding retry related information; the retry related information is to re-execute the entire process or re-execute the process from a certain middle process; and execute the corresponding process according to the retry related information.

[0048] A third aspect of the present application provides an electronic device, including: a memory and a processor;

[0049] The memory stores computer-executable instructions;

[0050] The processor executes the computer-executable instructions stored in the memory to implement the large model-based human-computer interaction method as described in any one of the first aspects.

[0051] The fourth aspect of the present application provides a computer-readable storage medium, in which computer execution instructions are stored. When the computer execution instructions are executed by a processor, they are used to implement the large model-based human-computer interaction method described in any one of the first aspects.

[0052] The fifth aspect of the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the large model-based human-computer interaction method described in any one of the first aspects.

[0053] The present application provides a human-computer interaction method, device, equipment, medium and product based on a large model, the method comprising: obtaining a target question input by a user; inputting the target question into a preset intention recognition model to generate a corresponding user interaction intention; if the user interaction intention is a business-related intention, using a small business model to generate corresponding target information based on the target question; the small business model is generated based on the large model and training samples; the training samples are related to preset business data; executing a processing flow related to the target information according to the target information, generating a first processing result, and feeding back the first processing result to the user. The human-computer interaction method based on a large model of the present application identifies the target question input by the user through a preset intention recognition model to generate a corresponding user interaction intention. If the user interaction intention is a business-related intention, using a small business model to generate corresponding target information based on the target question. At the same time, executing a processing flow related to the target information according to the target information, generating a first processing result, and feeding back the first processing result to the user. Thus, the function of the large model to parse business-related questions is realized, so that users can get answers to business-related questions through the large model, and improve the user's interactive experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0055] Figure 1 A scene diagram for realizing the large model-based human-computer interaction method of the embodiment of the present application;

[0056] Figure 2Schematic diagram of the process of the human-computer interaction method based on the large model provided in this application Figure 1 ;

[0057] Figure 3 Schematic diagram of the process of the human-computer interaction method based on the large model provided in this application Figure 2 ;

[0058] Figure 4 A schematic diagram of the overall process of the human-computer interaction method based on a large model provided in this application;

[0059] Figure 5 A schematic diagram of the structure of a human-computer interaction device based on a large model provided in this application;

[0060] Figure 6 A schematic diagram of the structure of the electronic device provided in this application.

[0061] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0062] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0063] It should be noted that the human-computer interaction method, device, equipment, medium and product based on the large model disclosed in the present invention can be used in the field of artificial intelligence technology. It can also be used in any field other than the field of artificial intelligence technology. The application field of the human-computer interaction method, device, equipment, medium and product based on the large model disclosed in the present invention is not limited.

[0064] The technical solution of the present application is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0065] In order to clearly understand the technical solution of the present application, the concept of the present solution is first introduced in detail. The question-answering system is an important means of obtaining information from large-scale data. The question-answering system based on natural language processing technology enables users to ask questions in an intuitive and natural way to obtain the required information. At present, human-computer question-answering conversations generally use language large models, which can also be referred to as large models, to interact with users. These models are trained on large amounts of text data to generate and understand natural language texts and perform various language tasks, such as text generation, translation, question-answering, summarization, sentiment analysis, etc.

[0066] However, large language models generally lack the ability to use industry data or business data to parse problems. Therefore, there is currently a lack of a human-computer interaction method that can parse business-related problems. Therefore, in response to the problem that there is a lack of a human-computer interaction method that can parse business-related problems in the prior art, the inventors found in their research that in order to solve this problem, an intent recognition model can be pre-set to identify the target question input by the user and generate the corresponding user interaction intention. If the user interaction intention is a business-related intention, a pre-trained small business model is used to perform corresponding processing based on the target question to complete the interaction with the user. Among them, the small business model is generated based on the large language model, which can also be called a large model and training samples.

[0067] Specifically, obtain the target question input by the user. Input the target question into the preset intent recognition model to generate the corresponding user interaction intention. If the user interaction intention is a business-related intention, use the small business model to generate the corresponding target information based on the target question. The small business model is generated based on the large model and training samples. The training samples are related to the preset business data. Execute the processing flow related to the target information according to the target information, generate a first processing result, and feedback the first processing result to the user.

[0068] The human-computer interaction method based on the big model of the present application identifies the target question input by the user through a preset intention recognition model, and generates the corresponding user interaction intention. If the user interaction intention is a business-related intention, the small business model is used to generate the corresponding target information based on the target question. At the same time, the processing flow related to the target information is executed according to the target information, a first processing result is generated, and the first processing result is fed back to the user. Thereby realizing the function of the big model to parse business-related problems, so that users can get answers to business-related questions through the big model, and improve the user's interactive experience.

[0069] Based on the above creative findings, the inventor proposed the technical solution of the present application.

[0070] The following is an introduction to the application scenarios of the human-computer interaction method based on a large model provided in the embodiment of the present application. Figure 1As shown, 1 is an electronic device and 2 is a user. The electronic device 1 may be a computer, a tablet computer, etc.

[0071] For example, when user 2 needs to query data, he / she inputs the question to be queried in electronic device 1, that is, ① inputs the target question, and electronic device 1 performs the following processing:

[0072] ② Input the target question into the preset intention recognition model to generate the corresponding user interaction intention.

[0073] ③ If the user's interaction intention is business-related, the small business model is used to generate the corresponding target information based on the target problem. The small business model is generated based on the large model and training samples. The training samples are related to the preset business data.

[0074] ④ Execute the processing flow related to the target information according to the target information, generate a first processing result, and feed back the first processing result to the user. In this embodiment, the feedback can be in the form of voice, direct display, etc.

[0075] The embodiments of the present application are introduced below in conjunction with the drawings in the specification.

[0076] Figure 2 Schematic diagram of the process of the human-computer interaction method based on the large model provided in this application Figure 1 ,like Figure 2 As shown, in this embodiment, the execution subject of the embodiment of the present application is a human-computer interaction device based on a large model, and the human-computer interaction device based on a large model can be integrated into an electronic device, such as a tablet computer, a computer, etc. Then the human-computer interaction method based on a large model provided in this embodiment includes the following steps:

[0077] S101, obtaining a target question input by a user.

[0078] In this embodiment, the user input method can be that the user inputs the target question to be queried by voice, or directly inputs the text of the target question by keyboard or touch light. At the same time, the method of obtaining the target question can be obtained by recognizing the voice input by the user, or directly obtained based on the text input by the user.

[0079] At the same time, when obtaining the target question, the historical conversation messages of this round can also be obtained for use in large-scale model analysis to determine the purpose of the user's input of the target question and the target query content that is more suitable for the current scenario.

[0080] S102, input the target question into a preset intention recognition model to generate a corresponding user interaction intention.

[0081] In this embodiment, user interaction intentions are divided into multiple types in advance, such as general query intentions, that is, only the big model itself is used to interact with the user, professional knowledge query intentions, that is, the big model is combined with the data of the professional knowledge database and / or the knowledge graph database to interact with the user, and business-related intentions, that is, the big model is combined with business data to interact with the user.

[0082] At the same time, the preset intent recognition model can be trained in advance, such as using questions that users may enter and the user interaction intentions corresponding to the questions as training samples, and performing continuous iterative training to improve the accuracy of the intent recognition model in identifying the user interaction intentions corresponding to the target questions.

[0083] In addition, when the preset intent recognition model is actually applied, the user's actual query questions and the actual recognition effect of the preset intent recognition model can be collected at intervals to further optimize the preset intent recognition model. Among them, the collected user's actual query questions can be classified into one type according to the degree of similarity, and multiple questions with high similarity can be classified into one type.

[0084] S103, if the user interaction intention is a business-related intention, a small business model is used to generate corresponding target information based on the target problem. The small business model is generated based on the large model and training samples. The training samples are related to the preset business data.

[0085] In this embodiment, the small business model is part of the large model, and the small business model is generated after the large model is trained with training samples related to business data. When the user's interaction intention is business-related, the small business model can be used in combination with the business processing logic pre-set by the developer to analyze the target problem and determine the target information corresponding to the target problem, where the business processing logic is a way of using the small model pre-configured by the developer. The target information can be the business information, parameter information, or device information that the user needs to query, etc.

[0086] Optionally, the small business model can also perform corresponding processing based on the current round of historical session information and the target question, thereby improving the adaptability of the subsequently generated first processing result to the actual needs of the user.

[0087] S104, executing a processing flow related to the target information according to the target information, generating a first processing result, and feeding back the first processing result to the user.

[0088] In this embodiment, the processing flow related to the target information is related to the target information. If the target information is the device information to be operated, the processing flow is the process related to how to operate the corresponding device. If the target information is the business information or parameter information to be queried, the processing flow is the process related to obtaining the corresponding business information or parameter information.

[0089] The first processing result may be data that ultimately needs to be fed back to the user, processing process data, a total processing report, and the like.

[0090] The embodiment of the present application provides a human-computer interaction method based on a large model, the method comprising: obtaining a target question input by a user. Inputting the target question into a preset intention recognition model to generate a corresponding user interaction intention. If the user interaction intention is a business-related intention, a small business model is used to generate corresponding target information based on the target question. The small business model is generated based on the large model and training samples. The training samples are related to preset business data. A processing flow related to the target information is executed according to the target information to generate a first processing result, and the first processing result is fed back to the user.

[0091] The human-computer interaction method based on the big model of the present application identifies the target question input by the user through a preset intention recognition model, and generates the corresponding user interaction intention. If the user interaction intention is a business-related intention, the small business model is used to generate the corresponding target information based on the target question. At the same time, the processing flow related to the target information is executed according to the target information, a first processing result is generated, and the first processing result is fed back to the user. Thereby realizing the function of the big model to parse business-related problems, so that users can get answers to business-related questions through the big model, and improve the user's interactive experience.

[0092] Figure 3 Schematic diagram of the process of the human-computer interaction method based on the large model provided in this application Figure 2 ,like Figure 3 As shown, the human-computer interaction method based on a large model provided in this embodiment is further refined on the basis of the human-computer interaction method based on a large model provided in the previous embodiment of this application, and the solution content of professional knowledge and general questions and answers is added. The human-computer interaction method based on a large model provided in this embodiment includes the following steps.

[0093] S201, obtaining a target question input by a user.

[0094] The implementation of S201 in this embodiment is similar to the implementation of S101 in the previous embodiment, and will not be described in detail here.

[0095] S202, input the target question into a preset intention recognition model to generate a corresponding user interaction intention.

[0096] The implementation of S202 in this embodiment is similar to the implementation of S102 in the previous embodiment, and will not be repeated here.

[0097] S203, if the user interaction intention is a business-related intention and the target problem is an operating system-related problem, the small business model is used to identify the operating system-related problem, generate corresponding device information, and use the device information as the target information.

[0098] In this embodiment, the user's business-related intentions can be divided into two aspects: operating system and query data. The operating system involves operating business-related equipment, such as cameras, computers and other equipment in the coal industry. Query data involves business-related information, parameter information, etc., such as coal production and energy consumption.

[0099] If the target problem is an operating system-related problem, the operating system-related problem can be identified through the small business model to determine the device information corresponding to the device that needs to be operated. For example, if the camera numbered 10 needs to be operated, the relevant device information of camera numbered 10 can be determined.

[0100] S204, if the user interaction intention is a business-related intention and the target question is a business data-related question, a small business model is used to identify the business data-related question, corresponding business information and corresponding parameter information are generated, and the business information and parameter information are used as target information.

[0101] In this embodiment, if the target question is a business data-related question, the business data-related question can be identified through the business model, such as determining the specific business information and corresponding parameter information that need to be queried. For example, if the user needs to query the energy consumption data of the second production line from 9 am to 11 am, the energy consumption information of the second production line from 9 am to 11 am can be determined as the target information, where the parameters are time, energy consumption, and production line identification, and the business information is the information of the production line.

[0102] Optionally, in this embodiment, if the user interaction intention is a professional knowledge query intention, this embodiment further includes the following process:

[0103] The knowledge model is used to search for matching knowledge information from the preset professional knowledge database based on the target problem. The knowledge model is generated based on the large model and training samples related to the preset professional knowledge. The preset professional knowledge database stores relevant professional knowledge of the coal industry.

[0104] A large model is used to perform data analysis and data organization on the target problem and knowledge information to generate the second processing result.

[0105] Determine whether it is necessary to call data from the knowledge graph database based on the second processing result.

[0106] If it is determined that there is no need to call the data in the knowledge graph database, the second processing result is fed back to the user.

[0107] If it is determined that data in the knowledge graph database needs to be called, entity extraction processing is performed on the target question to generate entity data to be queried.

[0108] Obtain the corresponding knowledge graph data from the knowledge graph database based on the entity data.

[0109] The knowledge graph data and the target question are input into the big model, the interaction results corresponding to the target question are generated, and the interaction results are fed back to the user.

[0110] In this embodiment, under normal circumstances, knowledge sources can be divided into knowledge graphs, knowledge bases, and general documents according to different knowledge storage media. Among them, knowledge base is a broad concept, including any form of knowledge storage, while knowledge graph is a specific form of knowledge base that emphasizes the semantic relationship between entities and is easy to query and reason. Different vertical fields have independent knowledge graphs and document libraries.

[0111] The knowledge graph database and the professional knowledge base support each other. When searching the professional knowledge base, the knowledge with the highest relevance or the top three relevance can be obtained through the small knowledge model according to the user's target question. At the same time, the knowledge and target question are handed over to the large model for reasoning, analysis and sorting to generate processing results. If further reasoning is required, the knowledge graph data can be retrieved and the corresponding interactive results can be generated based on the knowledge graph data.

[0112] For example, if the user needs to query the meaning of a word, the interaction with the user can be completed only through the knowledge of the professional knowledge base. If the user needs to know the semantic relationship between each word, it may be necessary to further generate corresponding interaction results based on the knowledge graph data.

[0113] The knowledge graph database contains entities and semantic relationships between entities. When searching the knowledge graph database, the question can be subjected to entity extraction, wherein entity extraction can use professional knowledge extraction algorithm tools. After extracting the entity, the knowledge graph data is obtained from the knowledge graph database. At the same time, the target question and the knowledge graph data are handed over to the big model for analysis and reasoning to improve the answer effect. The big model can also generate summary report data for the analysis process, analysis results, etc., and use the summary report data as the processing result of interaction with the user, thereby improving the detail of the answer question data and improving the question and answer effect.

[0114] In addition, when the user's question involves knowledge graph content and business data content, the knowledge graph data can be used to call business-related data such as production system data to obtain the content the user needs.

[0115] Optionally, in this embodiment, if the user interaction intention is a general query intention, the solution content of this embodiment also includes:

[0116] Get the historical session messages of this round.

[0117] The target question and the historical conversation messages of this round are input into the big model to generate the corresponding query result data.

[0118] Call the preset data parsing tool to convert the data format of the query result data into the standard data format required for display.

[0119] The converted query result data is displayed in the preset display area.

[0120] In this embodiment, the historical conversation messages of this round refer to the historical interaction information of the users of this round. Generally, the conversation messages of a preset number of times, such as three times, can be selected as the historical conversation messages of this round.

[0121] General query intent refers to the general content query involved in the original language big model or the big model, such as daily conversations, non-professional knowledge content questions and answers, etc. When the user interaction intention is a general query intention, the target question and the historical conversation message of this round can be input into the big model to generate the corresponding query result data through the function of the big model itself. At the same time, since the data format of the data generated by the big model is generally different from the data format of the user interaction page, the preset data parsing tool can be called to convert the data format of the query result data into the standard data format required for display, and the converted query result data is displayed in the preset display area.

[0122] In addition, the data of business data, professional knowledge database, and knowledge graph database may also be in a different format from the data format of the user interaction page, and a preset data parsing tool may be used to convert the data format. The preset data parsing tool can be written in advance by the developer and is suitable for data format conversion.

[0123] S205, executing a processing flow related to the target information according to the target information, generating a first processing result, and feeding back the first processing result to the user.

[0124] In this embodiment, if the target information is related to the operating system, various related functions can be called to implement specific operations. If the target information is related to business information, various related functions can be called to obtain data in the business interface to determine the required corresponding information.

[0125] Optionally, in this embodiment, S205 may be specifically as follows:

[0126] A preset function calling tool is used to call the corresponding function in the large model based on the target information to execute the processing flow corresponding to the target information.

[0127] In this embodiment, the preset function calling tool can be pre-written and developed by the developer to call the relevant functions in the large model. If the target information is related to the operating system, the relevant functions can be called layer by layer. For example, the query function is called to determine the path of the device to be operated, and the permission-related function is called to grant the permission to operate the device.

[0128] Optionally, in this embodiment, when a data or process failure occurs, a retry mechanism may be added, as follows:

[0129] When a data or process failure occurs, corresponding fault information is generated.

[0130] The fault information and recorded process execution information are sent to the big model to generate the corresponding retry related information. The retry related information is to re-execute the entire process or to re-execute the process from a certain process in the middle.

[0131] Execute the corresponding process according to the retry related information.

[0132] In this embodiment, when a multi-step task is executed through big model scheduling, the process execution information will be recorded, including the process stage being performed, the parameter data involved, etc. When a data or process fails, the big model can determine whether to re-execute the entire process or re-execute the process from a certain point in the middle based on the recorded process execution information and fault information.

[0133] This function of large models can be achieved by developers pre-configuring prompts (instructions). At the same time, the number of retries can also be manually configured to avoid wasting resources.

[0134] The human-computer interaction method based on the large model of this embodiment collects professional knowledge and trains it into the small knowledge model as fine-tuning corpus, and loads the corresponding model into the large model as needed. In addition, by combining knowledge graphs and business data, the professional knowledge capabilities of the large language model are enriched.

[0135] In order to further explain the human-computer interaction method based on the large model of the present application in detail, the following will be further explained in conjunction with the accompanying drawings. Figure 4 As shown, when the user is interacting with the conversation, he or she enters a corresponding question, which can be a question generated during production work. At this time, the intent recognition model determines the intent based on the question. If it is determined to be a general query intent, it can be processed based on the capabilities of the large model itself. Specifically:

[0136] Select the dialogue model function, calculate the token data (instruction card in programming) available in the dialogue model context, obtain the available historical conversation length, process the dialogue content based on the big model, and generate the big model processing results.

[0137] After generating the large model processing result, call the data parsing tool to convert the data format (specifically including determining the scenario of the current returned result and packaging the data into the scenario setting format), and use the converted data as the final processing result and return the result.

[0138] If it is determined to be a professional knowledge query intention, it can be processed according to the professional knowledge database and the knowledge graph database. Specifically:

[0139] Query the data in the professional knowledge database and determine whether the knowledge graph data is needed. If it is determined that the knowledge graph data is needed, extract the entity information of the question, and obtain the corresponding knowledge graph data based on the entity information, and hand over the question and knowledge to the knowledge model for processing to generate the knowledge graph processing result. At the same time, if the corresponding knowledge graph data is not obtained based on the entity information, return to the previous step and process the data based on the professional knowledge database.

[0140] If it is determined that the knowledge graph data is not needed, the data queried from the professional knowledge database and the user's questions are directly processed by the large model based on the small knowledge model to generate professional knowledge processing results.

[0141] In addition, if it is determined that business data auxiliary processing is required, corresponding processing can also be performed in combination with business data and / or business services. At the same time, the result generated by the final processing can call a data parsing tool to convert the data format and return the result.

[0142] If it is determined to be a business-related intention, it is determined that business-related data is required, such as if it needs to be implemented in combination with business data in the business system, and the corresponding processing results are generated.

[0143] The method of this embodiment integrates a large model of professional fields, which can answer professional questions in specific fields. At the same time, it solves the problem of large models not being able to process professional knowledge in a timely manner through external professional knowledge.

[0144] We developed an intent recognition model and packaged it into a tool to solve the problem of the system identifying the correct execution route based on natural language. Different recognition models can be called according to the corresponding steps to complete intent recognition, improve system flexibility, and greatly reduce the cost of business logic research and judgment.

[0145] Diversified data analysis provides a unified data analysis interface. Operation and maintenance personnel can configure multiple data formats according to usage scenarios, enrich data display effects, improve user experience, and avoid only text-formatted data without further intuitive display effects.

[0146] Figure 5A schematic diagram of the structure of the human-computer interaction device based on the large model provided in this application, such as Figure 5 As shown, in this embodiment, the human-computer interaction device 300 based on the large model can be set in an electronic device, and the human-computer interaction device 300 based on the large model includes:

[0147] The acquisition module 301 is used to acquire the target question input by the user.

[0148] The first generating module 302 is used to input the target question into a preset intention recognition model to generate a corresponding user interaction intention.

[0149] The second generation module 303 is used to generate corresponding target information based on the target problem using a small business model if the user interaction intention is a business-related intention. The small business model is generated based on the large model and training samples. The training samples are related to the preset business data.

[0150] The feedback module 304 is used to execute a processing flow related to the target information according to the target information, generate a first processing result, and feed back the first processing result to the user.

[0151] The human-computer interaction device based on the large model provided in this embodiment can be executed Figure 2 The technical solution of the method embodiment shown in the figure has the same implementation principle and technical effect as Figure 2 The method embodiments shown are similar and will not be described in detail here.

[0152] The human-computer interaction device based on a large model provided in the present application is further refined on the basis of the human-computer interaction device based on a large model provided in the previous embodiment. The human-computer interaction device based on a large model 300 includes:

[0153] Optionally, in this embodiment, when the second generating module 303 uses the small business model to generate corresponding target information based on the target problem, it is specifically used to:

[0154] If the target problem is an operating system-related problem, the small business model is used to identify the operating system-related problem, generate the corresponding device information, and use the device information as the target information. If the target problem is a business data-related problem, the small business model is used to identify the business data-related problem, generate the corresponding business information and corresponding parameter information, and use the business information and parameter information as the target information.

[0155] Optionally, in this embodiment, when the feedback module 304 executes a processing flow related to the target information according to the target information to generate a first processing result, it is specifically used to:

[0156] A preset function calling tool is used to call the corresponding function in the large model based on the target information to execute the processing flow corresponding to the target information.

[0157] Optionally, in this embodiment, if the user interaction intention is a professional knowledge query intention, the human-computer interaction device 300 based on the big model further includes:

[0158] The professional knowledge processing module is used to use a small knowledge model to search for matching knowledge information from a preset professional knowledge database based on a target question. The small knowledge model is generated based on the large model and training samples related to the preset professional knowledge. The preset professional knowledge database stores relevant professional knowledge of the coal industry. The large model is used to perform data analysis and data sorting on the target question and knowledge information to generate a second processing result. According to the second processing result, it is determined whether it is necessary to call the data in the knowledge graph database. If it is determined that it is not necessary to call the data in the knowledge graph database, the second processing result is fed back to the user. If it is determined that the data in the knowledge graph database needs to be called, entity extraction processing is performed on the target question to generate entity data to be queried. Obtain the corresponding knowledge graph data from the knowledge graph database based on the entity data. The knowledge graph data and the target question are input into the large model to generate an interactive result corresponding to the target question, and the interactive result is fed back to the user.

[0159] Optionally, in this embodiment, if the user interaction intention is a general query intention, the human-computer interaction device 300 based on the large model further includes:

[0160] The general processing module is used to obtain the historical conversation messages of this round. The target question and the historical conversation messages of this round are input into the large model to generate the corresponding query result data. The preset data parsing tool is called to convert the data format of the query result data into the standard data format required for display. The converted query result data is displayed in the preset display area.

[0161] Optionally, in this embodiment, the human-computer interaction device 300 based on the large model further includes:

[0162] The retry module is used to generate corresponding fault information when a data or process fails. The fault information and the recorded process execution information are sent to the big model to generate corresponding retry related information. The retry related information is to re-execute the entire process or to re-execute the process from a certain process in the middle. The corresponding process is executed according to the retry related information.

[0163] The human-computer interaction device based on the large model provided in this embodiment can be executed Figure 2-Figure 4 The technical solution of the method embodiment shown in the figure has the same implementation principle and technical effect as Figure 2-Figure 4 The method embodiments shown are similar and will not be described in detail here.

[0164] According to an embodiment of the present application, the present application also provides an electronic device, a computer-readable storage medium and a computer program product.

[0165] like Figure 6 As shown, Figure 6 is a schematic diagram of the structure of the electronic device provided by the present application. The electronic device is intended to be various forms of electronic devices, such as a tablet computer, a computer, etc. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present application described and / or required herein.

[0166] like Figure 6 As shown, the electronic device includes: a processor 401 and a memory 402. The various components are connected to each other using different buses and can be installed on a common mainboard or in other ways as required. The processor can process instructions executed in the electronic device.

[0167] The memory 402 is a non-transitory computer-readable storage medium provided in the present application. The memory stores instructions executable by at least one processor to enable at least one processor to execute the human-computer interaction method based on a large model provided in the present application. The non-transitory computer-readable storage medium of the present application stores computer instructions, which are used to enable a computer to execute the human-computer interaction method based on a large model provided in the present application.

[0168] The memory 402 is a non-transient computer-readable storage medium that can be used to store non-transient software programs, non-transient computer executable programs and modules, such as the program instructions / modules corresponding to the large model-based human-computer interaction method in the embodiment of the present application (for example, the attached Figure 5 The processor 401 executes various functional applications and data processing of the electronic device by running the non-transient software programs, instructions and modules stored in the memory 402, that is, the human-computer interaction method based on the large model in the above method embodiment is realized.

[0169] In the above embodiments, it should be understood that the processor can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), etc. A general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in the invention can be directly implemented as a hardware processor, or can be implemented by a combination of hardware and software modules in the processor.

[0170] The memory may include a high-speed memory (Random Access Memory, RAM), and may also include a non-volatile memory (NVM), such as at least one disk storage.

[0171] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, the bus in the drawings of this application is not limited to only one bus or one type of bus.

[0172] At the same time, this embodiment also provides a computer product. When the instructions in the computer product are executed by the processor of the electronic device, the electronic device can execute the human-computer interaction method based on the large model of the above embodiment.

[0173] The above-mentioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. The readable storage medium can be any available medium that can be accessed by a general or special-purpose computer.

[0174] An exemplary readable storage medium is coupled to a processor so that the processor can read information from the readable storage medium and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can be located in an application specific integrated circuit (Application Specific Integrated Circuits, referred to as: ASIC). Of course, the processor and the readable storage medium can also exist in the device as discrete components.

[0175] The division of units is only a logical function division, and there may be other divisions in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interface, device or unit, which can be electrical, mechanical or other forms.

[0176] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0177] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0178] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0179] Those skilled in the art will appreciate that all or part of the steps of implementing the above-mentioned method embodiments can be completed by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, the steps of the above-mentioned method embodiments are executed. The aforementioned storage medium includes: ROM, RAM, magnetic disk or optical disk, etc., various media that can store program codes.

[0180] Those skilled in the art will readily come up with other implementations of the embodiments of the present application after considering the specification and practicing the invention disclosed herein. The present application is intended to cover any modifications, uses or adaptations of the embodiments of the present application, which follow the general principles of the embodiments of the present application and include common knowledge or customary technical means in the art that are not disclosed in the embodiments of the present application.

[0181] It should be understood that the embodiments of the present application are not limited to the precise structures described above and shown in the drawings, and various modifications and changes can be made without departing from the scope thereof. The scope of the embodiments of the present application is limited only by the appended claims.

Claims

1. A human-computer interaction method based on a large model, characterized in that: include: Get the target question input by the user; Input the target question into a preset intention recognition model to generate a corresponding user interaction intention; If the user's interaction intention is business-related, a small business model is used to generate corresponding target information based on the target question; The small business model is generated based on the large model and training samples; The training samples are related to preset business data; Execute a processing flow related to the target information according to the target information, generate a first processing result, and feed back the first processing result to a user; If the target information is information about the device to be operated, then the processing flow is a flow related to how to operate the corresponding device; If the target information is the business information or parameter information to be queried, the processing flow is a related flow for obtaining the corresponding business information or parameter information; The adopting of the small business model to generate corresponding target information based on the target problem includes: If the target problem is an operating system-related problem, a small business model is used to identify the operating system-related problem, generate corresponding device information, and use the device information as the target information; If the target problem is a problem related to business data, a small business model is used to identify the problem related to the business data, corresponding business information and corresponding parameter information are generated, and the business information and the parameter information are used as the target information; If the user interaction intention is a professional knowledge query intention, the method further includes: A small knowledge model is used to search for matching knowledge information from a preset professional knowledge database based on the target problem; the small knowledge model is generated based on the large model and training samples related to the preset professional knowledge; the preset professional knowledge database stores relevant professional knowledge of the coal industry; Using a large model to perform data analysis and data sorting on the target problem and the knowledge information to generate a second processing result; Determine whether it is necessary to call data in the knowledge graph database according to the second processing result; If it is determined that there is no need to call the data in the knowledge graph database, the second processing result is fed back to the user; If it is determined that data in the knowledge graph database needs to be called, entity extraction processing is performed on the target question to generate entity data to be queried; Acquire corresponding knowledge graph data from the knowledge graph database according to the entity data; Input the knowledge graph data and the target question into a large model, generate an interaction result corresponding to the target question, and feed back the interaction result to the user; When the user's question involves knowledge graph content and business data content, the knowledge graph data is used to call business-related data to obtain the content the user needs.

2. The method according to claim 1, characterized in that The step of executing a processing flow related to the target information according to the target information to generate a first processing result includes: A preset function calling tool is used to call a corresponding function in the large model based on the target information to execute a processing flow corresponding to the target information.

3. The method according to claim 1, characterized in that: If the user interaction intention is a general query intention, the method further includes: Get the historical conversation messages of this round; Input the target question and the historical conversation messages of this round into the big model to generate corresponding query result data; Calling a preset data parsing tool to convert the data format of the query result data into a standard data format required for display; The converted query result data is displayed in the preset display area.

4. The method according to any one of claims 1 to 3, characterized in that: The method further comprises: When data or process fails, corresponding fault information is generated; Send the fault information and the recorded process execution information to the big model to generate corresponding retry related information; the retry related information is to re-execute the entire process or to re-execute the process from a certain process in the middle; Execute the corresponding process according to the retry related information.

5. A human-computer interaction device based on a large model, characterized in that: include: An acquisition module is used to obtain the target question input by the user; A first generating module, used to input the target question into a preset intention recognition model to generate a corresponding user interaction intention; A second generating module is used to generate corresponding target information based on the target problem using a small business model if the user interaction intention is a business-related intention; The small business model is generated based on the large model and training samples; The training samples are related to preset business data; A feedback module, configured to execute a processing flow related to the target information according to the target information, generate a first processing result, and feed back the first processing result to a user; If the target information is information about the device to be operated, then the processing flow is a flow related to how to operate the corresponding device; If the target information is the business information or parameter information to be queried, the processing flow is a related flow for obtaining the corresponding business information or parameter information; The second generating module is specifically configured to, if the target problem is an operating system-related problem, use a small business model to identify the operating system-related problem, generate corresponding device information, and use the device information as the target information; If the target problem is a problem related to business data, a small business model is used to identify the problem related to the business data, corresponding business information and corresponding parameter information are generated, and the business information and the parameter information are used as the target information; A professional knowledge processing module, used for searching matching knowledge information from a preset professional knowledge database based on the target problem using a small knowledge model; The small knowledge model is generated based on the large model and training samples related to preset professional knowledge; the preset professional knowledge database stores relevant professional knowledge of the coal industry; Using a large model to perform data analysis and data collation on the target problem and the knowledge information to generate a second processing result; judging whether it is necessary to call data in the knowledge graph database according to the second processing result; if it is determined that it is not necessary to call data in the knowledge graph database, feeding back the second processing result to the user; If it is determined that data in the knowledge graph database needs to be called, entity extraction processing is performed on the target question to generate entity data to be queried; corresponding knowledge graph data is obtained from the knowledge graph database according to the entity data; the knowledge graph data and the target question are input into the big model to generate an interaction result corresponding to the target question, and the interaction result is fed back to the user; When the user's question involves knowledge graph content and business data content, the knowledge graph data is used to call business-related data to obtain the content the user needs.

6. An electronic device, characterized in that: include: Memory and processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the large model-based human-computer interaction method as described in any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the large model-based human-computer interaction method as described in any one of claims 1 to 4.

8. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the human-computer interaction method based on a large model as described in any one of claims 1 to 4 is implemented.

Images