Information interaction system and method and electronic equipment

Abstract

The invention discloses an information interaction system and method and electronic equipment, and relates to the field of large model technology and application development. The system comprises a client used for detecting a query request in an interactive interface, and the query request at least comprises query content to be replied; the text processing end is used for retrieving a text associated with the query content based on the query content in the query request; the guide information processing end is used for generating guide information based on the retrieved text and query content, and the guide information is used for generating reply information corresponding to the query content; and the information reply end is used for calling a machine learning model to analyze the guide information and outputting reply information. The technical problem of low accuracy of information interaction is solved.

Description

Technical Field

[0001] The present application relates to the field of large-scale model technology and application development, and more specifically, to an information interaction system, method and electronic device. Background Art

[0002] At present, it is usually possible to dispatch some foreign tools in an open source application development framework (for example, Lang Chain), and use computer programming languages ​​(Python or TypeScript) to give large language models (LLM) data perception and agent capabilities, and generate corresponding large language model applications to interact with clients and provide them with response information.

[0003] However, since the application development is limited to using foreign tools, there will be some limitations and obstacles when such applications are used domestically. In addition, when faced with query requirements from clients in different application scenarios, it is impossible to provide corresponding response information in a targeted manner. Therefore, there is still a technical problem of low accuracy of information interaction.

[0004] Currently, no effective solution has been proposed for the above-mentioned technical problems. Summary of the invention

[0005] The embodiments of the present application provide an information interaction system, method and electronic device to at least solve the technical problem of low accuracy of information interaction.

[0006] According to one aspect of an embodiment of the present application, an information interaction system is provided. The system may include: a client, used to detect a query request in an interactive interface, wherein the query request at least includes query content to be answered; a text processing end, used to retrieve text associated with the query content based on the query content in the query request; a guidance information processing end, used to generate guidance information based on the retrieved text and the query content, wherein the guidance information is used to generate reply information corresponding to the query content; an information reply end, used to call a machine learning model to analyze the guidance information and output reply information.

[0007] According to another aspect of the embodiment of the present application, an information interaction method is also provided. The method may include: detecting a query request in an interactive interface, wherein the query request at least includes query content to be answered; displaying text associated with the query content on the interactive interface in response to the query request, wherein the text is retrieved based on the query content; displaying guidance information on the interactive interface, wherein the guidance information is generated based on the retrieved text and the query content; and displaying reply information on the interactive interface, wherein the reply information is obtained by analyzing the guidance information based on a machine learning model.

[0008] According to another aspect of the embodiment of the present application, another information interaction method is also provided. The method may include:

[0009] Obtain a query request from an application platform, wherein the query request includes at least query content to be answered in an application scenario in which the application platform is located; in response to the query request, obtain text associated with the query content, wherein the text is retrieved based on the query content; generate guidance information based on the retrieved text and the query content, wherein the guidance information is used to generate reply information corresponding to the query content; call a machine learning model to analyze the guidance information to obtain reply information in the application scenario, wherein the machine learning model is trained based on application data samples in the application scenario; and output the reply information to the application platform.

[0010] According to another aspect of the embodiment of the present application, a method for generating information is also provided. The method may include: detecting a query request, wherein the query request includes at least a query content to be answered; based on the query content in the query request, retrieving text associated with the query content; based on the retrieved text and the query content, generating guidance information, wherein the guidance information is used to enable a machine learning model to analyze the guidance information to obtain answer information.

[0011] According to another aspect of an embodiment of the present application, an electronic device is also provided, which may include a memory and a processor: the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions. When the above-mentioned computer-executable instructions are executed by the processor, any one of the above-mentioned information interaction methods is implemented.

[0012] According to another aspect of an embodiment of the present application, a processor is further provided, and the processor is used to run a program, wherein any one of the above-mentioned information interaction methods is executed when the program is running.

[0013] According to another aspect of an embodiment of the present application, a computer-readable storage medium is further provided, the computer-readable storage medium including a stored program, wherein when the program is running, the device where the storage medium is located is controlled to execute any one of the above-mentioned information interaction methods.

[0014] In an embodiment of the present application, the query content in the query request in the user input interactive interface can be detected by the client, and can be sent to the text processing end to retrieve the query content, determine the text associated with the search content, and send the query content and the associated text to the guidance information processing end for processing to generate corresponding guidance information, and the guidance information can be sent to the information reply end, and the guidance information is analyzed by the machine learning model in the information reply end to determine the reply information corresponding to the query content, and can be fed back to the interactive interface of the corresponding client for display to complete the information interaction process. Considering that there are certain differences in the needs of different clients for reply information in different application scenarios, the corresponding query request can be used in a targeted manner to develop an information interaction system for outputting reply information that meets the corresponding application scenario, thereby achieving the purpose of significantly improving the analysis and processing capabilities of the query request, and then achieving the technical effect of improving the accuracy of information interaction, and solving the technical problem of low accuracy of information interaction.

[0015] It is easy to notice that the above general description and the following detailed description are only for the purpose of exemplifying and explaining the present application, and do not constitute a limitation of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0017] Figure 1 is a schematic diagram of an information interaction system according to an embodiment of the present application;

[0018] Figure 2 is a schematic diagram of an application scenario of an information interaction method according to an embodiment of the present application;

[0019] Figure 3 is a flow chart of an information interaction method according to an embodiment of the present application;

[0020] Figure 4 is a flow chart of another information interaction method according to an embodiment of the present application;

[0021] Figure 5 is a flow chart of an information generation method according to an embodiment of the present application;

[0022] Figure 6 It is a flow chart of an enterprise-specific big model application method based on a big model as the core according to an embodiment of the present application;

[0023] FIG. 7( a ) is a schematic diagram of a title chain of a group block according to an embodiment of the present application;

[0024] FIG7( b ) is a schematic diagram of a linear block obtained by processing a document according to an embodiment of the present application;

[0025] FIG7( c ) is a schematic diagram of a structural block obtained by performing structural processing on scene data of an offline data transmission type according to an embodiment of the present application;

[0026] Figure 8 It is a schematic diagram of a format for generating query instructions and document fragments according to rules according to an embodiment of the present application;

[0027] Fig. 9 is a schematic diagram of an information interaction device according to an embodiment of the present application;

[0028] Fig.10 is a schematic diagram of another information interaction device according to an embodiment of the present application;

[0029] Fig.11 is a schematic diagram of an information generating device according to an embodiment of the present application;

[0030] Fig.12 is a structural block diagram of a computer terminal according to an embodiment of the present application;

[0031] Fig.13 It is a block diagram of an electronic device according to an information interaction method of an embodiment of the present application. DETAILED DESCRIPTION

[0032] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0033] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0034] The technical solution provided in this application is mainly implemented by large-scale model technology. The large model here refers to a deep learning model with large-scale model parameters, which can usually contain hundreds of millions, tens of billions, hundreds of billions, trillions or even more than 10 trillion model parameters. The large model can also be called a cornerstone model / foundation model (Foundation Model). The large model is pre-trained through large-scale unlabeled corpus to produce a pre-trained model with more than 100 million parameters. This model can adapt to a wide range of downstream tasks, and the model has good generalization ability, such as a large-scale language model (LLM), a multi-modal pre-training model, etc.

[0035] It should be noted that when the large model is actually applied, the pre-trained model can be fine-tuned through a small number of samples so that the large model can be applied to different tasks. For example, the large model can be widely used in natural language processing (NLP), computer vision, speech processing and other fields. Specifically, it can be applied to computer vision tasks such as visual question answering (VQA), image caption (IC), image generation, etc. It can also be widely used in natural language processing tasks such as text-based sentiment classification, text summary generation, machine translation, etc. Therefore, the main application scenarios of the large model include but are not limited to digital assistants, intelligent robots, search, online education, office software, e-commerce, intelligent design, etc. In the embodiment of the present application, the application generation of the target language processing model obtained by training the generative large model in a dialogue scenario is taken as an example to explain,

[0036] First, some nouns or terms that appear in the description of the embodiments of the present application are subject to the following explanations:

[0037] Large models can be deep learning models that contain a large number of parameters and learning data. Such models require tens of billions of parameters for training and optimization, a large amount of computing resources and data sets to achieve the purpose of improving prediction accuracy, and the ability to follow instructions to output results;

[0038] Generative models (generative text models) refer to a type of natural language processing model based on deep learning, which can generate text similar to human natural language. The model is usually trained using machine learning algorithms to learn the syntax and semantic expressions of the language, and use this knowledge to execute the model when generating text. When generating text, this model receives a piece of text as input, then predicts the probability distribution of the next single word or character, generates the next word or character based on the probability distribution, and continuously iterates to generate the next part of the text.

[0039] Example 1

[0040] Figure 1 is a schematic diagram of an information interaction system according to the present application. Figure 1 As shown, the information interaction system 100 may include:

[0041] The client 102 is used to detect a query request in the interactive interface, wherein the query request at least includes query content to be answered.

[0042] In the technical solution provided by the client 102 of the present application, the query request on the interactive interface can be detected, wherein the client 102 can also be referred to as the user end. The query request (e.g., the query requested by the user) can at least include the query content to be answered, which can be the query content input by the user, for example, the parking location of a motor vehicle, the parking location of a non-motor vehicle, or the route to the corresponding location.

[0043] It should be noted that the query content in the above query request is only for illustration and is not specifically limited here, and may change according to the application scenario of the information interaction. The query content may be in the form of text, image, or audio. The application level corresponding to the interactive interface that processes the query request and generates reply information may be an enterprise-specific application using a large model, that is, a large model application, such as a large language model application. Application scenarios can be used to characterize the different usage areas and functional requirements of different enterprises for applications, such as navigation map scenarios, translation scenarios, writing assistant scenarios, and emotional information scenarios.

[0044] It should be further noted that the above application scenarios are only examples and are not specifically limited here. The interactive interface can be a generative dialogue interface, for example, it can be a graphical user interface of a dialogue generation system (Chat Generative Pre-trained Transformer, referred to as Chat GPT).

[0045] Optionally, if the user has some questions or needs to answer questions, he can perform relevant operations on the interactive interface of the dialogue generation system on the terminal device in the client 102, for example, he can enter the query content corresponding to the question to be queried, generate a corresponding query request, and upload the query request to the server for processing.

[0046] The text processing terminal 104 is used to retrieve text associated with the query content based on the query content in the query request.

[0047] In the technical solution provided by the text processing terminal 104 of the present application, after the client 102 detects the query request in the interactive interface, the query request can be sent to the text processing terminal 104 in the server by the client 102 to process the query request. In the text processing terminal 104, the text associated with the query content in the query request can be retrieved from the pre-stored documents, wherein the text processing terminal 104 can be called a document terminal, which can be a document search engine. The text associated with the query content can be a corresponding document or a document fragment.

[0048] Considering that if only a single large model application is used to uniformly answer and provide reply information to users, there will be problems of poor pertinence and low accuracy. However, in an embodiment of the present application, before information interaction, the unified large model application can be pre-processed in a targeted manner according to the different application scenarios of the large model application, so that the developed large model application has better adaptability. The large model in the large model application can be trained according to the corresponding scenario data of different application scenarios, and trained into a large model that better meets the application requirements of the corresponding application scenario, and an exclusive large model application after training is obtained. This embodiment can provide exclusive large model applications to users for use. When users have relevant needs, they can obtain corresponding exclusive large model applications to interact with information, thereby achieving the purpose of making corresponding exclusive large model applications for different application scenarios, and then achieving the technical effect of improving the accuracy of reply information provided to users.

[0049] For example, according to the needs of different companies for their own exclusive big model applications, the scenario data corresponding to the company's application scenarios can be input on the corresponding interactive interface to generate corresponding application generation instructions. Based on the application generation instructions, the scenario data in the application generation instructions can be called to train the unified big model to obtain the trained company's exclusive big model. An exclusive big model application is built based on the exclusive big model, where the unified big model can be a unified language big model, which can also be called a language processing model, for example, Tongyi Qianwen. Tongyi Qianwen is an ultra-large-scale language model whose functions can include multi-round dialogue, copywriting, logical reasoning, multimodal understanding, and multi-language support.

[0050] Optionally, after the interactive interface uploads the acquired query request to the server, the text processing terminal 104 in the server can retrieve documents or document fragments associated with the query content in the query request from the documents that have been established in the exclusive large model application.

[0051] The guidance information processing terminal 106 is used to generate guidance information based on the retrieved text and the query content, wherein the guidance information is used to generate reply information corresponding to the query content.

[0052] In the technical solution provided by the above-mentioned guidance information processing terminal 106 of the present application, after the text associated with the query content is retrieved by the text processing terminal 104, the text and the query content can be sent to the guidance information processing terminal 106 in the calling server for processing. The guidance information processing terminal 106 can generate guidance information (prompt) based on the received text and query content, wherein the guidance information processing terminal 106 can be a prompt processing terminal, for example, it can be a prompt processor module. The guidance information can be used to generate reply information corresponding to the query content, and the guidance information can be a high-quality prompt.

[0053] Optionally, the prompt processor module can perform prompt splicing processing on the retrieved text and the query content to generate spliced ​​guidance information, and can send the guidance information to the large model for processing according to certain rules.

[0054] The information reply terminal 108 is used to call the machine learning model to analyze the guidance information and output the reply information. In the technical solution provided by the above-mentioned information reply terminal 108 of the present application, after the text and the query content are processed by the guidance information processing terminal 106 to obtain the guidance information, the guidance information can be transmitted to the information reply terminal 108 through the guidance information processing terminal 106. The information reply terminal 108 can call the existing machine learning model Ali to analyze the guidance information, determine the corresponding reply information, and transmit the reply information to the interactive interface in the client 102 for display, wherein the machine learning model can be a trained exclusive large model that can meet different application scenarios, for example, it can be a trained language processing model. The application scenario can be the needs of different enterprises for different use fields of large models. The reply information can be the information to be answered by the query request, and the expression of the reply information can be consistent with the expression that can match the query request, or it can be the expression required by the user, for example, it can be a natural language expression, or it can be a computer language expression. It should be noted that the expression of the above-mentioned reply information is only for example, and no specific limitation is made here. The reply information may also be called a response, and its form may be text, image, audio, etc. It should be noted that the form of the above reply information is only for illustration and is not specifically limited here.

[0055] Optionally, after the guidance information is input into the machine learning model in the information reply terminal 108, the guidance information can be understood and processed by the machine learning model, thereby continuously generating and obtaining the final reply information, and the reply information can be sent to the interactive interface in the corresponding client 102 for display.

[0056] If only by dispatching foreign tools to develop applications in Lang chain, and generate corresponding information interaction applications, however, the developed applications will have obstacles when they are put into domestic use, and they cannot meet the development requirements of different application scenarios, so that the accuracy of the reply information provided to users through the above-mentioned applications cannot be guaranteed. Therefore, there is still a technical problem of low accuracy of reply information provided to the client. However, in the embodiment of the present application, considering the need to train the initial unified machine learning model, such as a large model or a language processing model, for different domestic development needs, to achieve targeted application development, so as to build an exclusive large model application that can meet different application scenarios, the application may include an interactive interface for collecting and uploading query requests, and also include a text processing end for receiving and processing the query content in the query request, a guiding information processing end, and an information reply end. The exclusive large model application processes the user's query request and feeds back the associated reply information to the user, thereby achieving the purpose of being able to design an exclusive large model application for different application scenarios, and then achieving the technical effect of improving the accuracy of the reply information provided to the client, wherein the exclusive large model application can be trained by using a unified language model as a base and using the scene data of the application scenario.

[0057] It should be noted that the various parts included in the above-mentioned large model application are only examples and are not specifically limited here. As long as the process and method of understanding and analyzing the scene data through the generative model to design an application that conforms to the scene data are within the protection scope of the embodiments of this application.

[0058] Through the above-mentioned information interaction system of the present application, the query content in the query request in the user input interactive interface can be detected by the client, and can be sent to the text processing end to retrieve the query content, determine the text associated with the search content, and send the query content and the associated text to the guidance information processing end for processing to generate corresponding guidance information, and the guidance information can be sent to the information reply end, and the guidance information can be analyzed by the machine learning model in the information reply end to determine the reply information corresponding to the query content, and can be fed back to the interactive interface of the corresponding client for display to complete the information interaction process. Considering that there are certain differences in the needs of different clients for reply information in different application scenarios, the corresponding query requests can be used in a targeted manner to develop an information interaction system for outputting reply information that meets the corresponding application scenarios, thereby achieving the purpose of significantly improving the analysis and processing capabilities of the query requests, and then achieving the technical effect of improving the accuracy of information interaction, and solving the technical problem of low accuracy of information interaction.

[0059] The above system of this embodiment is further introduced below.

[0060] As an optional implementation, the client is used to rewrite the query request in response to the rewrite request in the interactive interface, wherein the amount of information in the query content in the rewritten query request is greater than the amount of information in the query content in the query request before the rewrite; the guidance information processing end is used to generate guidance information based on the retrieved text and the query content in the rewritten query request.

[0061] In this embodiment, after the query request is detected, a corresponding rewriting request may be triggered, and the query request may be rewritten by the rewriting module to obtain a rewritten query request. The guidance information processing terminal may generate corresponding guidance information based on the retrieved text associated with the query content in the query request and the rewritten query request, wherein the amount of information of the query content in the rewritten query request may be greater than the amount of information of the query content in the query request before the rewriting.

[0062] Optionally, a rewrite module (Rewrite) can be deployed on the interactive interface. After detecting that a query request is written on the interactive interface, the query content in the query request (query) can be completed by the rewrite module. For example, keywords in the query content can be added to complete the query content. By completing the query content, the amount of information in the query content can be increased to obtain a completed query request (query').

[0063] Optionally, after obtaining the completed query request, the document or plug-in description can be searched (retrieved) through a document search engine to obtain the corresponding text. The text and the rewritten query request are sent to the promptprocessor module, which can perform prompt splicing on the retrieved text and the query content in the rewritten query request to generate corresponding guidance information.

[0064] As an optional implementation, the text processing end is used to structure the retrieved text, wherein the structured text is used to represent the hierarchical structure of the text; the guidance information processing end is used to generate guidance information based on the structured text and query content.

[0065] In this embodiment, the retrieved text can be structured by the text processing end, and the structured text can be sent to the guidance information processing end. The guidance information processing end can generate corresponding guidance information based on the structured text and query content, wherein the structured text can be used to represent the hierarchical structure of the text.

[0066] Optionally, the retrieved text can be subjected to structural processing by the text processing end to obtain the text after structural processing, wherein structural processing can be used to indicate that the text is subjected to structural chunk processing (chunk). The text after structural processing can be used to represent the hierarchical structure of the text, that is, structural processing can make the structure of the text more hierarchical. The hierarchical structure can be a hierarchical document representation, which can also be called document hierarchical structure information, and can be represented by a binary tree or the like. This is only for illustrative purposes, and no specific restrictions are placed on the form of expression of the hierarchical structure. Different hierarchies contain different document hierarchical structure information, that is, corresponding features of different hierarchies. For example, the root node of the binary tree contains the global features of the text, and the child nodes contain the internal features of the text.

[0067] In the embodiments of the present application, the provided exclusive large model application may include not only an online end but also an offline end. The online segment can interact with the client in real time when the client needs to answer questions. The offline end can be used for identification and indexing. It should be noted that no specific restrictions are made here on the roles and functions contained in the offline end and the online segment. The processing flow of the offline process can be divided into the offline end for processing, and the processing flow of the online process can be divided into the online end for processing. As long as the exclusive large model application is used to interact with the user and provide the user with functions and systems for answering information, they are all within the protection scope of the embodiments of the present invention.

[0068] In the embodiment of the present application, since different data have different transmission modes, in order to ensure the efficiency of data transmission, the data can be divided into data of offline transmission mode and data of online transmission mode according to the different data transmission modes. For the data of offline transmission mode, corresponding processing can be performed in the offline terminal. For the data of online transmission mode, corresponding processing can be performed in the online terminal. The data of offline transmission mode that needs to be processed offline can be processed in the offline terminal through offline processing processes such as document recognition, application programming interface (Application Programming Interface, referred to as API) registration and data vectorization. The data obtained by online transmission mode that needs to be processed online can be processed in the online segment through online processing processes such as question rewriting and related document recall, so that data with different needs can be accurately processed in a targeted manner, achieving the technical effect of improving the efficiency of data processing.

[0069] For example, the document can be represented by a document graph on the offline side to make full use of the structured information of the text.

[0070] Optionally, the following offline process is performed in the offline end: the offline end can perform document recognition on the received scene data of the offline data transmission type, and can use Doc Graph to perform document representation on the scene data after document recognition to obtain text data, and can identify global features (Global Features) and internal features (Inner Features) in the text data. The text data can be structured, that is, the text data can be divided into hierarchical structure chunks.

[0071] Optionally, in the process of determining the enterprise's exclusive big model based on the initial big model, the index information of the structured text data can be established through the offline end, and then the structured text and index information can be used to train the model, wherein the index information can be used to query the structured text data, and can be an identifier corresponding to each structured text, for example, the identifier can be represented as the chunk_id corresponding to each chunk. The structured document data may include paragraph information (paragraph) of the document data and information such as the title chain (title_chain). The title chain may include the chunk identifier chunk_id, the node relationship of the chunk, and the chunk.

[0072] Optionally, after representing the documents of different formats through DocGraph, the index of each structured chunk can be established, that is, each structured chunk can be labeled, for example, each structured chunk can be assigned its own chunk_id, and each chunk_id can be annotated with the paragraph information and title information of the corresponding document data. Thus, the index information of the structured text (chunks / docs) can be obtained, for example, it can be "chunk_id, title_chain, paragraph".

[0073] Optionally, the index information of the tools to be scheduled can also be established, that is, each scheduled tool can be numbered, for example, each tool can be assigned its own tool_id, and each tool_id can be marked with the corresponding manifest and specification. Thus, the index information of the tool (Tool Descs) can be obtained, for example, it can be "tool_id, Manifest, Spec".

[0074] Optionally, after establishing the index information of the document data and the index information of the tool, the index information may be embedded (Indexing) to obtain the final exclusive large model output text (file / Docs).

[0075] For example, in the offline end of the information interaction system, offline processes such as data upload, API registration, and data vectorization can be performed on the first target input text data.

[0076] Optionally, the text can be represented in the form of a binary tree, and divided layer by layer from the root node downwards, so that a structural chunk can be obtained. In the embodiment of the present application, considering that the global information in the document will be lost through linear chunks, it is proposed that the text data can be represented by structural chunks, thereby achieving the technical effect of being able to reuse the document hierarchical structure information and improving the accuracy of document analysis.

[0077] For example, the data (documents) that can be processed in the offline end may include plain text file format (PlainText File, referred to as txt), portable document format (Portable Document Format, referred to as pdf), document format (Document, referred to as doc), photocopy document, Uniform Resource Locator (Uniform Resource Locator, referred to as URL) web page / website and other formats. In the offline end, for the scene data in txt format, the document can be directly represented by Doc Graph. For the scene data in pdf format, doc format and photocopy format, the document recognition of the scene data in the above multiple formats can be performed by the artificial intelligence cloud service (Document Artificial Intelligence, referred to as DocAI) for document automation processing, and then the document can be represented by Doc Graph. For the scene data of url web page / website, the document recognition can be performed by a parser first, and then the document can be represented by DocGraph. The above scene data can be represented by Doc Graph as a document icon structure (Doc Graph Structure) such as text, table or image. It should be noted that this is only an example and does not impose any specific restrictions on the format of the offline processed data or the form of document representation.

[0078] As an optional implementation method, the text processing end is used to segment the retrieved text in response to the text length of the retrieved text being greater than the text length threshold; the guidance information processing end is used to generate guidance information based on the segmented text and query content.

[0079] In this embodiment, the text length of the retrieved text can be detected by the text processing end. If the text length exceeds the text length threshold, the corresponding text can be segmented and the segmented text can be sent to the guidance information processing end. The guidance information processing end can generate corresponding guidance information based on the segmented text and the query content, wherein the text length threshold can be a preset value or a value set according to actual conditions. It should be noted that the above-mentioned setting method of the text length threshold is only for example and is not specifically limited here. Text whose length exceeds the text length threshold can be called long text.

[0080] If a long text is cut into several smaller linear blocks, however, based solely on the above-mentioned cutting method, there will be a situation where the semantic information of the text is lost in the text, so there is still a technical problem of low accuracy of document analysis. However, in the embodiment of the present application, in order to avoid the situation where the semantic information of the text is lost directly in the text, the text can be structured so that it can be cut into hierarchical structural blocks for document representation, so that the hierarchical text can achieve the purpose of improving the understanding ability of document question and answer under the large model architecture, thereby achieving the technical effect of improving the accuracy of document analysis.

[0081] As an optional implementation method, the plug-in end is used to register the description data of the plug-in, wherein the description data is used to enable the information reply end to recognize the plug-in and execute the function corresponding to the plug-in.

[0082] In this embodiment, the description data of the plug-in related to the query request can be registered through the plug-in end in the document search engine, so that when the information reply end recognizes the plug-in, the corresponding function of the plug-in can be executed based on the description data of the plug-in. The description data can be used to enable the information reply end to recognize different plug-ins and execute the respective functions of the corresponding plug-ins. The description data can also be called a plug-in description.

[0083] As an optional implementation method, the text processing end is used to retrieve descriptive data associated with the query content based on the query content in the query request; the guidance information processing end is used to generate guidance information based on the retrieved text, the retrieved descriptive data and the query content.

[0084] In this embodiment, the text processing end can retrieve the query content in the received query request, retrieve the description data of the plug-in associated with the query content, and can send the retrieved text, description data, query content and other information to the guidance information processing end, and generate corresponding guidance information based on the above information.

[0085] Optionally, after obtaining the completed query request, the document corresponding to the query content in the completed query request can be searched (retrieved) through a document search engine to search for the description data of the plug-in to be used by the document, and the corresponding text can also be searched. The text, the completed query request and the description data are uniformly sent to the prompt processor module for processing to generate corresponding guidance information.

[0086] As an optional implementation manner, the guidance information processing end is used to splice the retrieved text and the retrieved description data to obtain a first splicing result, and generate guidance information based on the first splicing result and the query content.

[0087] In this embodiment, after the guidance information processing end receives the retrieved text, description data and completed query request, the retrieved text and description data can be spliced ​​to obtain a first splicing result, and corresponding guidance information can be generated based on the first splicing result and the query content, wherein the first splicing result can be used to indicate that both the text and the plug-in description data are spliced ​​together.

[0088] Optionally, if the guidance information processing end receives not only the retrieved text but also the description data of the retrieved plug-in, the prompt processor module can perform prompt splicing on the retrieved text, the description data and the query content in the completed query request to generate guidance information.

[0089] Optionally, the interactive interface can detect whether the user inputs a query request. When a query request is detected, the query request can be uploaded to the text processing end in the server to query the keywords in the description data of the plug-in. For example, the query request can be processed according to the context of the query content in the query request. Then, the description data of the required plug-in can be indexed from the pre-established plug-in end based on the processed query request. That is, the description data of the plug-in (for example, document fragments or plug-in descriptions, etc.) can be recalled from the document to Doc Cands and transmitted to the prompt processor module through Rank / MRC if too long. The document or plug-in description can be searched through the document search engine to obtain the search results, and the search results can be sorted and then sent to the prompt processor in the online end for prompt splicing.

[0090] As an optional embodiment, the guidance information processing end is used to at least splice the multiple retrieved texts to obtain a second splicing result, and generate guidance information based on the second splicing result and the query content.

[0091] In this embodiment, the guidance information processing end can at least splice the retrieved multiple texts to obtain a second splicing result, and can generate corresponding guidance information based on the second splicing result and the query content, wherein the second splicing result can be used to indicate that only the text is spliced.

[0092] Optionally, if the guidance information processing end only receives the retrieved text, the promptprocessor module can perform prompt concatenation on the retrieved text to obtain a second concatenation result, and can generate corresponding guidance information based on the second concatenation result and the query content in the completed query request.

[0093] As an optional embodiment, the text processing end is used to send the retrieved text to the client; the client is used to modify the text in response to the first modification operation on the interactive interface; the guidance information processing end is used to generate guidance information based on the modified text and query content.

[0094] In this embodiment, the text processing end can send the retrieved text to the client, and can display it on the interactive interface of the client. It can determine whether the text displayed on the interactive interface needs to be modified based on the user's needs. If the text needs to be modified, a first modification operation can be performed on the interactive interface to generate corresponding modification instructions to modify the text, and guidance information can be generated based on the modified text and query content through the guidance information processing end, wherein the first modification operation can be used to adjust the retrieved text.

[0095] Optionally, after analyzing and understanding the query content in the query request at the text processing end and retrieving the text associated with the query content, the retrieved text can be sent to the corresponding user's interactive interface in the client for display, so that the user can confirm whether there is inaccurate or even wrongly retrieved text. The user can confirm and filter the text according to their own needs, and can remove or modify some unnecessary or problematic text. That is, corresponding modification or deletion operations can be performed on the interactive interface to generate corresponding operation instructions to modify or delete the corresponding text, thereby ensuring the accuracy of the text.

[0096] In an embodiment of the present application, since the retrieved text can be fed back to the user in the client, the user is provided with the execution authority of operations such as self-confirmation and modification of the text, that is, the user can adjust the retrieved text by operating on the interactive interface according to his own needs, etc., so that the reply information to be fed back to the user later is more accurate and more in line with the user's needs, thereby achieving the purpose of improving the interactivity between the exclusive language model application and the user, thereby realizing the technical effect of improving the accuracy of the reply information provided to the client.

[0097] As an optional implementation manner, the guidance information processing end is used to send guidance information to the client; the client is used to modify the guidance information in response to the second modification operation acting on the interactive interface.

[0098] In this embodiment, the guidance information processing end can send the generated guidance information to the client, and can display it on the interactive interface of the client. It can determine whether the guidance information displayed on the interactive interface needs to be modified according to the user's needs. If it needs to be modified, a second modification operation can be performed on the interactive interface to generate corresponding modification instructions to modify the guidance information, and the final reply information can be generated by the modified guidance information, wherein the second modification operation can be used to adjust the generated guidance information.

[0099] In an embodiment of the present application, after the guidance information processing end generates corresponding guidance information for the received modified text and other information, it can be sent to the corresponding user's interactive interface in the client for display, so that the user can confirm whether there is inaccurate or erroneous guidance information. The user can filter the guidance information according to his or her own needs, and can modify the guidance information by eliminating or modifying problematic or unnecessary guidance information. Considering that the guidance information can be modified and adjusted through interaction with the user, the accuracy of the generated reply information is ensured.

[0100] As an optional implementation method, the information reply end is used to display the modified guidance information, and call the machine learning model to analyze the modified guidance information and output reply information.

[0101] In this embodiment, after the client modifies the original guidance information received, the modified guidance information can be sent to the information reply end, and the machine learning model can be called to analyze the modified guidance information to output the final reply information.

[0102] In the embodiment of the present application, the text and guidance information are modified through the above-mentioned two interactions with the user, which greatly ensures the accuracy of the text and guidance information, further ensures the accuracy of the reply information provided to the user, and the reply information better meets the user's needs, thereby solving the technical problem of low accuracy of reply information used by the client.

[0103] Optionally, in the process of training the machine learning model to obtain an exclusive large model, after the query, document fragments, etc. enter the prompt-processor, the query, document fragments, etc. can be organized into a certain format and sent to the machine learning model according to certain rules.

[0104] For example, you can combine queries, document fragments, etc. into the following format:

[0105] Humans: Please keep in mind the following information, as it may be helpful to answer certain questions. If you use a particular instruction, please include the number that preceded the instruction in your response:

[0106] [1] “

Title

[0107] [Main text] Dear, the non-motorized parking lot is located opposite the Tangqiao Ferry (near the security booth), and there is currently no charging equipment available.

[0108] Warm reminder: The underground parking lot of the park is a motor vehicle parking lot, and motorcycles / battery vehicles / bicycles are not allowed to park there. ”

[0109] Virtual Assistant: OK, I've got that information. Is there anything I can help you with?

[0110] Human: Can bicycles be parked in office areas?

[0111] Virtual Assistant:.

[0112] It should be noted that humans can also be represented by "Human" and virtual assistants can also be represented by "Assistant". The above formats and expressions fed into the big model are only examples and are not specifically limited here.

[0113] For another example, based on the above method, the virtual assistant can determine the corresponding information to be remembered for learning the application requirements of different application scenarios, so that an exclusive large model can be trained, so that when the user sends relevant questions to the virtual assistant, the virtual assistant can answer the user's relevant questions based on the pre-learned documents.

[0114] In the embodiment of the present application, considering the need to develop targeted applications according to the development needs of different domestic enterprises, the initial unified big model can be called and used as a base for training. After training, an exclusive big model that meets the needs of the corresponding application scenario can be obtained to develop applications, thereby achieving the purpose of being able to develop corresponding applications for different application scenarios, and further realizing the technical effect of improving the applicability of application development.

[0115] As an optional implementation method, the information reply end is used to send the retrieved text and reply information to the client; the client is used to display the text on the interactive interface in response to the query operation of the reply information.

[0116] In this embodiment, when the information reply end processes the modified guidance information using the machine learning model and generates the reply information, the text retrieved from the query content of the query request and the generated reply information can be sent to the client, and the reply information can be displayed on the interactive interface of the corresponding user in the client. If the user needs to query the retrieved text corresponding to the reply information, the query operation can be performed on the interactive interface to generate the corresponding query instruction so that the text is displayed in the interactive interface.

[0117] Optionally, after the information reply end generates the reply information corresponding to the query request, the generated reply information can be fed back to the client and displayed on the interactive interface. If the user needs to view the text used to generate the reply information, the corresponding query operation can be performed on the interactive interface, such as clicking the query control or voice expressing the query request to generate the corresponding query instruction to query the text corresponding to the reply information. It should be noted that the above-mentioned operation method of querying the text is only for illustration and is not specifically limited here.

[0118] In an embodiment of the present application, the query content in the query request in the user input interactive interface can be detected by the client, and can be sent to the text processing end to retrieve the query content, determine the text associated with the search content, and send the query content and the associated text to the guidance information processing end for processing to generate corresponding guidance information, and the guidance information can be sent to the information reply end, and the guidance information is analyzed by the machine learning model in the information reply end to determine the reply information corresponding to the query content, and can be fed back to the interactive interface of the corresponding client for display to complete the information interaction process. Considering that there are certain differences in the needs of different clients for reply information in different application scenarios, the corresponding query request can be used in a targeted manner to develop an information interaction system for outputting reply information that meets the corresponding application scenario, thereby achieving the purpose of significantly improving the analysis and processing capabilities of the query request, and then achieving the technical effect of improving the accuracy of information interaction, and solving the technical problem of low accuracy of information interaction.

[0119] Example 2

[0120] According to an embodiment of the present application, an information interaction method is provided, and the information interaction method is used to be executed by the information interaction system in Example 1. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that here.

[0121] Considering the huge amount of model parameters of the large model and the limited computing resources of the mobile terminal, the above information interaction method provided in the embodiment of the present application can be applied to Figure 2 The application scenarios shown are not limited to these. Figure 2 In the application scenario shown, the large model is deployed in the server 10, and the server 10 can be connected to one or more client devices 20 through a local area network connection, a wide area network connection, an Internet connection, or other types of data networks. The client device 20 here may include but is not limited to: a smart phone, a tablet computer, a laptop computer, a PDA, a personal computer, a smart home device, a vehicle-mounted device, etc. An application platform for detecting and obtaining query requests can be deployed on the graphical user interface of the client device. There is an interactive interface on the application platform for information interaction, and the interactive interface can be a generative dialogue interface. The client device 20 can interact with the user through the graphical user interface to realize the call of the large model, thereby realizing the information interaction method provided in the embodiment of the present application.

[0122] In an embodiment of the present application, the system composed of a client device and a server can perform the following steps: if there is a user in the client who wants to obtain reply information for the corresponding content from the application platform, the query request corresponding to the corresponding content (query content) to be replied can be entered on the interactive interface on the client device. The client device can obtain the query request and send the query request to the server through the network. After receiving the query request, the server can execute: step S202, obtain the query request from the application platform; step S204, in response to the query request, obtain the text associated with the query content; step S206, generate guidance information based on the retrieved text and query content; step S208, call the machine learning model to analyze the guidance information and obtain the reply information in the application scenario; step S210, output the reply information to the application platform. The server can send the reply information generated by calling the machine learning model to the corresponding client device through the network. It should be noted that when the operating resources of the client device can meet the deployment and operation conditions of the large model, the embodiment of the present application can be performed in the client device.

[0123] Under the above operating environment, Example 2 of the present application provides the following Figure 3 The model-based application generation method shown.

[0124] The embodiment of the present application provides an information interaction method from the client side, wherein the client side may be a user end. It should be noted that the steps shown in the flowchart of the accompanying drawings may be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that here.

[0125] Figure 3 is a flow chart of an information interaction method according to an embodiment of the present application. Figure 3 As shown, the method may include the following steps:

[0126] Step S302: detecting a query request in the interactive interface, wherein the query request at least includes query content to be answered.

[0127] In the technical solution provided in the above step S302 of the present application, it is possible to detect whether there is a query request on the interactive interface, wherein the query request may at least include query content that needs to be answered by the large model application, and its expression form includes but is not limited to text, image or audio. It should be noted that the expression form of the above query request is only for example and is not specifically limited here.

[0128] Optionally, when the user needs to provide reply information through the corresponding large model application, the user can click the corresponding application on the client device to display the interactive interface. The user can enter the query request including the query content to be answered on the interactive interface. During this process, the client can detect whether there is a query request on the interactive interface. When a query request is detected, it can be sent to the server for processing.

[0129] Step S304, in response to the query request, displaying text associated with the query content on the interactive interface, wherein the text is retrieved based on the query content.

[0130] In the technical solution provided in the above step S304 of the present application, text associated with the query content in the query request can be displayed on the interactive interface, wherein the text can be retrieved based on the query content.

[0131] Optionally, after detecting that there is a query request on the interactive interface, the query request may be sent to a text processing terminal in the server to process and retrieve the query request to obtain text associated with the query content in the query request.

[0132] Optionally, a rewriting module may be deployed on the interactive interface. After detecting that a query request is written on the interactive interface, the rewriting module may be used to complete the query content in the query request (query) to obtain a completed query request (query').

[0133] Optionally, after obtaining the completed query request, the document or plug-in description may be retrieved through a document search engine to obtain the corresponding text.

[0134] In the embodiment of the present application, in order to ensure the accuracy of the retrieved text, the retrieved text can be sent to the corresponding user's interactive interface in the client through the text search engine for display, so that the user can confirm whether there is inaccurate or even wrongly retrieved text. The user can confirm and filter the text according to his own needs, and can eliminate or modify some unnecessary or problematic text. That is, the corresponding modification or deletion operation can be performed on the interactive interface to generate the corresponding operation instruction to modify or delete the corresponding text, thereby ensuring the accuracy of the text.

[0135] Step S306: displaying guidance information on the interactive interface, wherein the guidance information is generated based on the retrieved text and the query content.

[0136] In the technical solution provided in the above step S306 of the present application, after the retrieved text and query content are processed to obtain guidance information, the guidance information can be fed back to the corresponding interactive interface in the client for display, wherein the guidance information can be generated based on the retrieved text and query content.

[0137] Optionally, the retrieved text and the completed query request or the description information of the plug-in may be uniformly sent to the guidance information processing end for processing. For example, prompt splicing may be performed to generate corresponding guidance information.

[0138] In an embodiment of the present application, in order to ensure the accuracy of the generated guidance information, it can be sent to the interactive interface of the corresponding user in the client for display, so that the user can confirm whether there is inaccurate or erroneous guidance information. The user can filter the guidance information according to his or her own needs, and can eliminate or modify problematic or unnecessary guidance information to modify the guidance information. Considering that the guidance information can be modified and adjusted through interaction with the user, the accuracy of the generated reply information is ensured.

[0139] Step S308, displaying reply information on the interactive interface, wherein the reply information is obtained by analyzing the guidance information based on the machine learning model.

[0140] In the technical solution provided in the above step S308 of the present application, after the guidance information is analyzed based on the machine learning model to obtain the reply information, the reply information can be sent to the interactive interface in the client for display, wherein the reply information can be obtained by analyzing the guidance information based on the machine learning model.

[0141] Optionally, after the guidance information is input into the machine learning model in the information reply end, the guidance information can be understood and processed by the machine learning model, thereby continuously generating and obtaining the final reply information, and the reply information can be sent to the interactive interface in the corresponding client for display.

[0142] Through the above steps S302 to S308 of the present application, a query request in the interactive interface is detected, wherein the query request at least includes query content to be answered; in response to the query request, text associated with the query content is displayed on the interactive interface, wherein the text is retrieved based on the query content; guidance information is displayed on the interactive interface, wherein the guidance information is generated based on the retrieved text and the query content; reply information is displayed on the interactive interface, wherein the reply information is obtained by analyzing the guidance information based on a machine learning model, thereby achieving a technical effect of improving the accuracy of information interaction and solving the technical problem of low accuracy of information interaction.

[0143] The embodiment of the present application also provides an information interaction method from the server side, wherein the server can be an enterprise platform. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0144] Figure 4 FIG. 1 is a flow chart of another information interaction method according to an embodiment of the present application. Figure 4 As shown, the method may include the following steps:

[0145] Step S402: obtaining a query request from the application platform, wherein the query request at least includes query content to be answered in the application scenario where the application platform is located.

[0146] In the technical solution provided in step S402 of the present application, a query request from an application platform may be received, wherein the query request may at least include query content to be answered in the application scenario in which the application platform is located. The application platform may be a platform for an application program established through a dedicated large model. The application platform may include an interactive interface and a server corresponding to the application.

[0147] Optionally, the interactive interface of the application platform on the user's client device may be detected in real time to obtain a query request from the interactive interface, and the query request may be sent to a server in the application platform for processing.

[0148] Step S404, in response to the query request, obtaining text associated with the query content, wherein the text is retrieved based on the query content.

[0149] In the technical solution provided in the above step S404 of the present application, after the server of the application platform detects the query request, the query content in the query request can be retrieved to retrieve the text associated therewith, wherein the text can be retrieved based on the query content.

[0150] Optionally, a rewriting module can be deployed on the interactive interface of the application platform. After detecting that a query request is written on the interactive interface, the rewriting module can be used to complete the query content in the query. For example, keywords in the query content can be added to complete the query content. By completing the query content, the amount of information in the query content can be increased to obtain a completed query'. The completed query' can be sent to the text processing end in the application platform. The text processing end can be used to retrieve it and obtain the corresponding text.

[0151] Step S406, generating guidance information based on the retrieved text and the query content, wherein the guidance information is used to generate reply information corresponding to the query content.

[0152] In the technical solution provided in the above step S406 of the present application, the retrieved text and query content can be processed to generate corresponding guidance information, wherein the guidance information can be used to generate response information required for the query content through processing by a machine learning model.

[0153] Optionally, the retrieved text and the completed query request or the description information of the plug-in may be uniformly sent to the guidance information processing terminal in the server of the application platform for processing. For example, prompt splicing may be performed to generate corresponding guidance information.

[0154] Step S408, calling a machine learning model to analyze the guidance information to obtain response information in the application scenario, wherein the machine learning model is trained based on application data samples in the application scenario.

[0155] In the technical solution provided in the above step S408 of the present application, after the guidance information is generated by the retrieved text and query content, the machine learning model can be called to analyze the guidance information to obtain the response information in the application scenario, wherein the machine learning model can be trained based on the application data samples in the application scenario. The application scenario can be used to represent the different uses and application requirements of the application platform. The application data samples can be the corresponding data samples deployed for the application scenario that are exclusive to the enterprise.

[0156] Optionally, for the initial unified application platform, due to the different application scenarios and application requirements of different enterprises, the big model used in the unified application platform can be pre-trained according to the application data samples corresponding to the application scenarios of each enterprise, so as to obtain the big model exclusive to each enterprise, thereby building an application platform exclusive to each enterprise.

[0157] Optionally, after building an enterprise-specific application platform, the exclusive large model in the application platform can be called to analyze the guidance information to obtain response information that meets the query request in accordance with the corresponding application scenario.

[0158] Step S410: output reply information to the application platform.

[0159] In the technical solution provided in the above step S410 of the present application, after the guidance information is processed by the machine learning model and the reply information is obtained, it can be sent to and displayed on the interactive interface in the application platform used to interact with the user.

[0160] Through the above steps S402 to S410 of the present application, a query request from the application platform is obtained, wherein the query request at least includes the query content to be answered in the application scenario where the application platform is located; in response to the query request, text associated with the query content is obtained, wherein the text is retrieved based on the query content; guidance information is generated based on the retrieved text and the query content, wherein the guidance information is used to generate reply information corresponding to the query content; a machine learning model is called to analyze the guidance information to obtain reply information in the application scenario, wherein the machine learning model is trained based on application data samples in the application scenario; and the reply information is output to the application platform, thereby achieving the technical effect of improving the accuracy of information interaction and solving the technical problem of low accuracy of information interaction.

[0161] The embodiment of the present application also provides a method for generating information. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that here.

[0162] Figure 5 is a flow chart of an information generation method according to an embodiment of the present application. Figure 5 As shown, the method may include the following steps:

[0163] Step S502: detecting a query request, wherein the query request at least includes query content to be answered.

[0164] In the technical solution provided in the above step S502 of the present application, the query request may be detected, wherein the query request may at least include the query content to be answered.

[0165] Optionally, an interactive interface deployed in the client in the exclusive large model application can be used to detect whether the user has entered a query request. If a query request is detected, it can be input into the server in the application for processing.

[0166] Optionally, a rewriting module may be deployed on the interactive interface. After detecting that a query request is written on the interactive interface, the rewriting module may be used to complete the query content in the query to obtain a completed query request (query').

[0167] Step S504: based on the query content in the query request, retrieve text associated with the query content.

[0168] In the technical solution provided in the above step S504 of the present application, in the server in the application, a search can be performed according to the query content in the query request to retrieve text associated with the query content.

[0169] Optionally, after the completed query request is obtained, it can be sent to a document search engine in the server, and the document or plug-in description can be retrieved through the document search engine to obtain the corresponding text.

[0170] Step S506, generating guidance information based on the retrieved text and query content, wherein the guidance information is used to enable the machine learning model to analyze the guidance information to obtain response information.

[0171] In the technical solution provided in the above step S506 of the present application, after retrieving the text associated with the query content, corresponding guidance information can be generated based on the retrieved text and the query content, wherein the guidance information can be used to analyze the guidance information using a machine learning model to obtain reply information.

[0172] Optionally, the retrieved text and the completed query request or the description information of the plug-in may be uniformly sent to a guidance information processing terminal in the server for processing. For example, prompt splicing may be performed to generate corresponding guidance information.

[0173] Optionally, after the guidance information is input into the machine learning model in the information reply end, the guidance information can be understood and processed by the machine learning model, thereby continuously generating and obtaining the final reply information, and the reply information can be sent to the interactive interface in the corresponding client for display.

[0174] Through the above steps S502 to S506 of the present application, a query request is detected, wherein the query request at least includes query content to be answered; based on the query content in the query request, text associated with the query content is retrieved; based on the retrieved text and the query content, guidance information is generated, wherein the guidance information is used to enable a machine learning model to analyze the guidance information and obtain reply information, thereby achieving the technical effect of improving the accuracy of information interaction and solving the technical problem of low accuracy of information interaction.

[0175] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0176] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the described order of actions, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present application.

[0177] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, disk, CD), including several instructions for enabling a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0178] Example 3

[0179] At present, since the advent of Chat GPT, major companies have launched their own large models, but there is no marked solution on how to use large models to solve actual problems of enterprises. In related technologies, it is usually possible to dispatch some foreign tools in Lang Chain, and use Python or TypeScript to give LLM) data perception and agent capabilities to develop corresponding applications, and interact with users according to the corresponding applications to provide users with the required response information. However, due to the limitation of dispatching foreign tools for application development, there will be some limitations and obstacles when such applications are used domestically, and when faced with different development needs of different companies, it is impossible to develop the required applications in a targeted manner, which makes the applicability of application development low. Therefore, there is still a technical problem of low accuracy of information interaction.

[0180] Optionally, the present application provides an enterprise-exclusive big model application method based on a big model as the core, which solves the technical problem of low accuracy of reply information provided to the client. It is different from traditional solutions that do not target the application development needs of different enterprises. It avoids the limitations and obstacles of using foreign tools for application development, which causes the reply information provided to users to produce errors or inaccuracies, and the technical problem of low accuracy of information interaction, thereby solving the technical problem of low accuracy of information interaction.

[0181] In an embodiment of the present application, the query content in the query request in the user input interactive interface can be detected by the client, and can be sent to the text processing end to retrieve the query content, determine the text associated with the search content, and send the query content and the associated text to the guidance information processing end for processing to generate corresponding guidance information, and the guidance information can be sent to the information reply end, and the guidance information is analyzed by the machine learning model in the information reply end to determine the reply information corresponding to the query content, and can be fed back to the interactive interface of the corresponding client for display to complete the information interaction process. Considering that there are certain differences in the needs of different clients for reply information in different application scenarios, the corresponding query request can be used in a targeted manner to develop an information interaction system for outputting reply information that meets the corresponding application scenario, thereby achieving the purpose of significantly improving the analysis and processing capabilities of the query request, and then achieving the technical effect of improving the accuracy of information interaction, and solving the technical problem of low accuracy of information interaction.

[0182] The above method of this embodiment is further introduced below.

[0183] In this embodiment, Figure 6 is a flow chart of an enterprise-specific big model application method based on a big model as the core according to an embodiment of the present application, such as Figure 6 As shown, the method may include the following steps:

[0184] Step S601: document recognition scene data to obtain text data.

[0185] In the technical solution provided in the above step S601 of the present application, if a unified large model application needs to be trained, it is necessary to obtain the scene data required for the corresponding application scenario, and document recognition can be performed on the scene data to obtain text data, wherein the scene data can be the data required for the application scenarios of different enterprises.

[0186] Optionally, if an enterprise needs to develop an enterprise-specific large-scale model application that meets its own application scenarios and other requirements, relevant operations can be performed on the interactive interface of the dialogue generation system in the enterprise's client device. For example, the application scenario and scenario data to be used by the application to be generated can be input on the interactive interface. When the development requirements of the application such as scenario data are received on the interactive interface, the corresponding application generation instruction can be generated. The application generation instruction can also be uploaded to the server for processing to generate an application that meets the application scenario of the enterprise.

[0187] Optionally, the received data can be subjected to document recognition in the offline terminal, and the document recognized data can be represented by Doc Graph to obtain text data, and global features and internal features in the text data can be identified. The text data can be structured, that is, the text data can be divided into hierarchical structured chunks.

[0188] Optionally, the scene data for offline processing may include formats such as plain text document format, pdf format, doc format, photocopy document, url web page / website, etc. In the offline end, for scene data in plain text document format, the document can be directly represented by a document icon (Doc Graph). For scene data in pdf format, doc format, and photocopy format, Doc AI can be used to perform document recognition on the scene data in the above multiple formats, and then the document can be represented by a document icon. For scene data in url web page / website, document recognition can be performed by a parser first, and then the document can be represented by a document icon. The above scene data can be represented by a document icon structure such as text form, table form, or image form through a document icon, and the global features and internal features in the scene data can be distinguished.

[0189] Step S602: Create index information of structured text data.

[0190] In the technical solution provided in the above step S602 of the present application, index information of structured text data can be established.

[0191] Optionally, after document representation of scene data in different formats through document icons, an index of each structured chunk after structuring can be established, that is, each structure chunk can be numbered, for example, each structure chunk can be assigned its own chunk_id, and each chunk_id can be annotated with the paragraph information and title information of the corresponding document data. Thus, the index information (chunks / docs) of the structured text data is obtained, for example, it can be "chunk_id, title_chain, paragraph". The index information of the tool to be scheduled can also be established, that is, each scheduled tool (tool) can be numbered, for example, each tool can be assigned its own tool_id, and each tool_id can be annotated with the corresponding manifest (Manifest) and specification (Spec). Thus, the index information of the tool (Tool Descs) is obtained, for example, it can be "tool_id, Manifest, Spec".

[0192] Figure 7(a) is a schematic diagram of a title chain of a chunk according to an embodiment of the present application. As shown in Figure 7(a), the title chain (title_chain) of the chunk chunk may include the chunk identifier (chunk_id) and the connection relationship between each node in the binary tree of the structural chunk, as well as the chunk.

[0193] FIG7(b) is a schematic diagram of a linear block obtained by processing a document according to an embodiment of the present application. As shown in FIG7(b), a linear block is a structured method in the prior art. This method can cut a long text into several smaller linear blocks. However, based on the above-mentioned cutting method, the text's subsequent text will lose semantic information, that is, the linear block will lose document level information. Therefore, there is still a technical problem of low accuracy of document analysis. However, in an embodiment of the present application, in order to avoid the situation where the semantic information of the subsequent text of the long text is lost directly, the scene data can be structured to obtain a structural block as shown in FIG7(c), so that the scene data can be cut into hierarchical structural blocks to represent the document, so that the hierarchical scene data can reuse the document level through the structural blocks under the large model architecture, so as to achieve the purpose of improving the understanding ability of document question answering, thereby achieving the technical effect of improving the accuracy of document analysis.

[0194] FIG7(c) is a schematic diagram of a structural block obtained by structurally processing the scene data of an offline data transmission type according to an embodiment of the present application. As shown in FIG7(c), in the embodiment of the present application, the text data can be represented in the form of a binary tree, and divided layer by layer from the root node downward, so that the structural block can be obtained. In the present application, considering that the global information in the document will be lost through linear blocks, it is proposed that the text data can be represented by structural blocks, so as to achieve the technical effect of being able to reuse the document hierarchical structure information and improving the accuracy of document analysis.

[0195] Optionally, after the index information of the document data and the index information of the tool are established, the Indexing (Moss Embedding) index information may be embedded into the offline language processing model to obtain the file Docs.

[0196] For example, in the offline end, offline processes such as data upload, API registration, and data vectorization can also be performed on the first target input text data.

[0197] Step S603: Complete the query request.

[0198] In the technical solution provided in the above step S603 of the present application, the query request may be completed to obtain a completed query request.

[0199] Optionally, the context (query context) of the query instruction can be queried through the rewriting module (Rewrite) in the online end, and the information therein can be completed. For example, keywords can be added to the query content in the query request to obtain the completed query request (query').

[0200] Step S604: recall the document and obtain the plug-in data.

[0201] In the technical solution provided in the above step S604 of the present application, the plug-in data can be obtained through document recall.

[0202] Optionally, the plug-in data can be recalled to Doc Cands and transmitted to the prompt processor module through Rank / MRC if too long. The document or plug-in description can be searched through the document search engine in the online end to obtain the search results, which can be sorted and then sent to the prompt processor in the online end for prompt splicing.

[0203] Step S605: splicing the plug-in data or documents.

[0204] In the technical solution provided in the above step S605 of the present application, prompt splicing can be performed on the plug-in data or document.

[0205] Optionally, the document or plug-in description can be searched through the document search engine in the online end to obtain search results, and the search results can be sorted and then sent to the prompt processor in the online end for prompt splicing.

[0206] Optionally, if the guidance information processing end receives not only the retrieved text but also the description data of the retrieved plug-in, the prompt processor module can perform prompt splicing on the retrieved text, the description data and the query content in the completed query request to generate guidance information.

[0207] Optionally, if the guidance information processing end only receives the retrieved text, the promptprocessor module can perform prompt concatenation on the retrieved text to obtain a second concatenation result, and can generate corresponding guidance information based on the second concatenation result and the query content in the completed query request.

[0208] Optionally, after entering the prompt-processor, the query, document fragments, etc. can be formatted into a certain format according to certain rules and sent to the large model for training.

[0209] Optionally, Figure 8 is a schematic diagram of a format for generating query instructions and document fragments according to rules according to an embodiment of the present application, such as Figure 8 As shown, the query, document fragments, etc. can be composed into the following format:

[0210] Humans: Please keep in mind the following information, as it may be helpful to answer certain questions. If you use a particular instruction, please include the number that preceded the instruction in your response:

[0211] [1] “

Title

[0212] [Main text] Dear, the non-motorized parking lot is located opposite the Tangqiao Ferry (near the security booth), and there is currently no charging equipment available.

[0213] Warm reminder: The underground parking lot of the park is a motor vehicle parking lot, and motorcycles / battery vehicles / bicycles are not allowed to park there. ”

[0214] Virtual Assistant: OK, I've got that information. Is there anything I can help you with?

[0215] Human: Can bicycles be parked in office areas?

[0216] Virtual Assistant:.

[0217] Step S606, generating a reply box.

[0218] In the technical solution provided in the above step S606 of the present application, a reply box can be generated.

[0219] Optionally, in the process of inputting the above format into the conversational language model (Moss) for training, if a reply box needs to be generated, the corresponding API can be called, that is, if call API, the corresponding reply box can be generated through the tool application programming interface (ToolAPI) through the application programming interface response (API Response).

[0220] Optionally, based on the above method, the virtual assistant can determine the corresponding information to be remembered for learning the application requirements of different application scenarios, so that an exclusive large model can be trained, so that when the user sends relevant questions to the virtual assistant, the virtual assistant can answer the user's relevant questions based on the pre-learned documents.

[0221] In the embodiment of the present application, considering the need to develop targeted applications according to the development needs of different domestic enterprises, the initial unified big model can be called and used as a base for training. After training, an exclusive big model that meets the needs of the corresponding application scenario can be obtained to develop applications, thereby achieving the purpose of being able to develop corresponding applications for different application scenarios, and further realizing the technical effect of improving the applicability of application development.

[0222] It should be noted that the preferred implementation scheme involved in the above embodiments of the present application is the same as the scheme provided in Example 1, as well as the application scenario and implementation process, but is not limited to the scheme provided in Example 1.

[0223] Example 4

[0224] According to an embodiment of the present application, a method for implementing the above Figure 3 An information interaction device of the information interaction method shown.

[0225] Fig. 9 is a schematic diagram of an information interaction device according to an embodiment of the present application. Fig. 9 As shown, the information interaction device 900 may include: a first detection unit 902 , a first display unit 904 , a second display unit 906 and a third display unit 908 .

[0226] The first detection unit 902 is used to detect a query request in the interactive interface, wherein the query request at least includes query content to be answered.

[0227] The first display unit 904 is used to display text associated with the query content on the interactive interface in response to the query request, wherein the text is retrieved based on the query content.

[0228] The second display unit 906 is used to display guidance information on the interactive interface, wherein the guidance information is generated based on the retrieved text and the query content.

[0229] The third display unit 908 is used to display the reply information on the interactive interface, wherein the reply information is obtained by analyzing the guidance information based on the machine learning model.

[0230] It should be noted that the first detection unit 902, the first display unit 904, the second display unit 906 and the third display unit 908 correspond to steps S302 to S308 in Example 2, and the four units and the corresponding steps implement the same examples and application scenarios, but are not limited to the contents disclosed in Example 2. It should be noted that the above units can be hardware components or software components stored in a memory and processed by one or more processors, and the above units can also be run in the computer terminal provided in Example 5 as part of the device.

[0231] According to an embodiment of the present application, a method for implementing the above Figure 4 An information interaction device of the information interaction method shown.

[0232] Fig.10 is a schematic diagram of another information interaction device according to an embodiment of the present application. Fig.10 As shown, the information interaction device 1000 may include: a first acquisition unit 1002 , a second acquisition unit 1004 , a first generation unit 1006 , a calling unit 1008 and an output unit 1010 .

[0233] The first acquisition unit 1002 is used to acquire a query request from the application platform, wherein the query request at least includes query content to be answered in the application scenario where the application platform is located.

[0234] The second acquisition unit 1004 is configured to acquire text associated with the query content in response to the query request, wherein the text is retrieved based on the query content.

[0235] The first generating unit 1006 is used to generate guidance information based on the retrieved text and the query content, wherein the guidance information is used to generate reply information corresponding to the query content.

[0236] The calling unit 1008 is used to call the machine learning model to analyze the guidance information to obtain the response information in the application scenario, wherein the machine learning model is trained based on the application data samples in the application scenario.

[0237] The output unit 1010 is used to output the reply information to the application platform.

[0238] It should be noted that the first acquisition unit 1002, the second acquisition unit 1004, the first generation unit 1006, the calling unit 1008 and the output unit 1010 correspond to steps S402 to S410 in Example 2, and the five units and the corresponding steps implement the same examples and application scenarios, but are not limited to the contents disclosed in the above-mentioned Example 2. It should be noted that the above-mentioned units can be hardware components or software components stored in a memory and processed by one or more processors, and the above-mentioned units can also be run in the computer terminal provided in Example 5 as part of the device.

[0239] According to an embodiment of the present application, a method for implementing the above Figure 5 The information generating device shown.

[0240] Fig.11 is a schematic diagram of an information generating device according to an embodiment of the present application. Fig.11 As shown, the fine generation device 1100 may include: a second detection unit 1102 , a retrieval unit 1104 and a second generation unit 1106 .

[0241] The second detection unit 1102 is configured to detect a query request, wherein the query request at least includes query content to be answered.

[0242] The retrieval unit 1104 is configured to retrieve text associated with the query content based on the query content in the query request.

[0243] The second generating unit 1106 is used to generate guidance information based on the retrieved text and query content, wherein the guidance information is used to enable the machine learning model to analyze the guidance information to obtain response information.

[0244] It should be noted that the second detection unit 1102, the retrieval unit 1104 and the second generation unit 1106 correspond to steps S502 to S506 in Example 2, and the three units and the corresponding steps implement the same examples and application scenarios, but are not limited to the contents disclosed in Example 2. It should be noted that the above units can be hardware components or software components stored in a memory and processed by one or more processors, and the above units can also be run in the computer terminal provided in Example 5 as part of the device.

[0245] In the above-mentioned device, the query content in the query request in the user input interactive interface can be detected by the client, and can be sent to the text processing end to retrieve the query content, determine the text associated with the search content, and send the query content and the associated text to the guidance information processing end for processing to generate corresponding guidance information, and the guidance information can be sent to the information reply end, and the guidance information can be analyzed by the machine learning model in the information reply end to determine the reply information corresponding to the query content, and can be fed back to the interactive interface of the corresponding client for display to complete the information interaction process. Considering that there are certain differences in the needs of different clients for reply information in different application scenarios, the corresponding query request can be used in a targeted manner to develop an information interaction system for outputting reply information that meets the corresponding application scenario, thereby achieving the purpose of significantly improving the analysis and processing capabilities of the query request, and then achieving the technical effect of improving the accuracy of information interaction, and solving the technical problem of low accuracy of information interaction.

[0246] Example 5

[0247] The embodiment of the present application may provide a computer terminal, which may be any computer terminal device in a computer terminal group. Optionally, in this embodiment, the computer terminal may also be replaced by a terminal device such as a mobile terminal.

[0248] Optionally, in this embodiment, the computer terminal may be located in at least one network device among a plurality of network devices of the computer network.

[0249] In this embodiment, the above-mentioned computer terminal can execute the program code of the following steps in the model-based application generation method: detecting an application generation instruction acting on an interactive interface, wherein the application generation instruction at least includes: scenario data required for the application to be generated in the corresponding application scenario; in response to the application generation instruction, calling an initial language processing model, wherein the initial language processing model is used to determine the semantics of initial input data and convert the initial input data into initial output data, and the initial output data is used to describe a natural language that matches the semantics of the initial input data; using the scenario data, training the initial language processing model into a target language processing model in the application scenario; generating an application based on the target language processing model, wherein the target input data of the application is converted via the target language processing model to obtain target output data, the target output data satisfies the attributes of the application scenario, and the target output data is used to describe a natural language that matches the semantics of the target input data.

[0250] Optionally, Fig.12 is a structural block diagram of a computer terminal according to an embodiment of the present application. Fig.12As shown, the computer terminal A may include: one or more (only one is shown in the figure) processors 1202, a memory 1204, a storage controller, and a peripheral interface, wherein the peripheral interface is connected to a radio frequency module, an audio module, and a display.

[0251] Among them, the memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the model-based application generation method and device in the embodiment of the present application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, the above-mentioned model-based application generation method is realized. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include a memory remotely arranged relative to the processor, and these remote memories can be connected to the computer terminal A via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0252] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: detect a query request in the interactive interface, wherein the query request includes at least the query content to be answered; display text associated with the query content on the interactive interface in response to the query request, wherein the text is retrieved based on the query content; display guidance information on the interactive interface, wherein the guidance information is generated based on the retrieved text and the query content; display reply information on the interactive interface, wherein the reply information is obtained by analyzing the guidance information based on the machine learning model.

[0253] As another optional example, the processor may call the information and application stored in the memory through the transmission device to perform the following steps: obtain a query request from the application platform, wherein the query request includes at least the query content to be answered in the application scenario in which the application platform is located; in response to the query request, obtain text associated with the query content, wherein the text is retrieved based on the query content; generate guidance information based on the retrieved text and the query content, wherein the guidance information is used to generate reply information corresponding to the query content; call a machine learning model to analyze the guidance information to obtain reply information in the application scenario, wherein the machine learning model is trained based on application data samples in the application scenario; and output the reply information to the application platform.

[0254] As another optional example, the processor may call the information and application stored in the memory through the transmission device to perform the following steps: detecting a query request, wherein the query request includes at least query content to be answered; retrieving text associated with the query content based on the query content in the query request; and generating guidance information based on the retrieved text and query content, wherein the guidance information is used to enable a machine learning model to analyze the guidance information to obtain reply information.

[0255] By adopting the embodiment of the present application, the query content in the query request in the user input interactive interface can be detected by the client, and can be sent to the text processing end to retrieve the query content, determine the text associated with the search content, and send the query content and the associated text to the guidance information processing end for processing to generate corresponding guidance information, and the guidance information can be sent to the information reply end, and the guidance information can be analyzed by the machine learning model in the information reply end to determine the reply information corresponding to the query content, and can be fed back to the interactive interface of the corresponding client for display to complete the information interaction process. Considering that there are certain differences in the needs of different clients for reply information in different application scenarios, the corresponding query request can be used in a targeted manner to develop an information interaction system for outputting reply information that meets the corresponding application scenario, thereby achieving the purpose of significantly improving the analysis and processing capabilities of the query request, and then achieving the technical effect of improving the accuracy of information interaction, and solving the technical problem of low accuracy of information interaction.

[0256] It can be understood by those skilled in the art that Fig.12 The structure shown is for illustration only, and the computer terminal may also be a smart phone (such as an Android phone, an iOS phone, etc.), a tablet computer, a PDA, a mobile Internet device (Mobile Internet Devices, MID), a PAD, or other terminal devices. Fig.12 It does not limit the structure of the above electronic device. For example, the computer terminal A may also include Fig.12 More or fewer components (such as network interfaces, display devices, etc.) shown in, or having Fig.12 Different configurations shown.

[0257] A person of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the hardware related to the terminal device through a program, and the program can be stored in a computer-readable storage medium, and the storage medium may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0258] Example 6

[0259] The embodiment of the present application also provides a computer-readable storage medium. Optionally, in this embodiment, the computer-readable storage medium can be used to store the program code executed by the model-based application generation method provided in the first embodiment.

[0260] Optionally, in this embodiment, the above storage medium may be located in any computer terminal in a computer terminal group in a computer network, or in any mobile terminal in a mobile terminal group.

[0261] Optionally, in this embodiment, the storage medium is configured to store program code for performing the following steps: detecting a query request in an interactive interface, wherein the query request includes at least query content to be answered; displaying text associated with the query content on the interactive interface in response to the query request, wherein the text is retrieved based on the query content; displaying guidance information on the interactive interface, wherein the guidance information is generated based on the retrieved text and the query content; and displaying reply information on the interactive interface, wherein the reply information is obtained by analyzing the guidance information based on a machine learning model.

[0262] In an embodiment of the present application, the query content in the query request in the user input interactive interface can be detected by the client, and can be sent to the text processing end to retrieve the query content, determine the text associated with the search content, and send the query content and the associated text to the guidance information processing end for processing to generate corresponding guidance information, and the guidance information can be sent to the information reply end, and the guidance information is analyzed by the machine learning model in the information reply end to determine the reply information corresponding to the query content, and can be fed back to the interactive interface of the corresponding client for display to complete the information interaction process. Considering that there are certain differences in the needs of different clients for reply information in different application scenarios, the corresponding query request can be used in a targeted manner to develop an information interaction system for outputting reply information that meets the corresponding application scenario, thereby achieving the purpose of significantly improving the analysis and processing capabilities of the query request, and then achieving the technical effect of improving the accuracy of information interaction, and solving the technical problem of low accuracy of information interaction.

[0263] Example 7

[0264] An embodiment of the present application may provide an electronic device, which may include a memory and a processor. Fig.13It is a block diagram of an electronic device according to a model-based application generation method according to an embodiment of the present application. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. 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.

[0265] like Fig.13 As shown, the device 1300 includes a computing unit 1301, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 1302 or a computer program loaded from a storage unit 1308 to a random access memory (RAM) 1303. In RAM 1303, various programs and data required for the operation of the device 1300 can also be stored. The computing unit 1301, ROM 1302, and RAM 1303 are connected to each other via a bus 1304. An input / output (I / O) interface 1305 is also connected to the bus 1304.

[0266] A number of components in the device 1300 are connected to the I / O interface 1305, including: an input unit 1306, such as a keyboard, a mouse, etc.; an output unit 1304, such as various types of displays, speakers, etc.; a storage unit 1308, such as a disk, an optical disk, etc.; and a communication unit 13010, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 13010 allows the device 1300 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0267] The computing unit 1301 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 1301 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The computing unit 1301 performs the various methods and processes described above, such as a model-based application generation method. For example, in some embodiments, the model-based application generation method may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 1308. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 1300 via the ROM 1302 and / or the communication unit 13010. When the computer program is loaded into RAM 1303 and executed by computing unit 1301, one or more steps of the model-based application generation method described above may be performed. Alternatively, in other embodiments, computing unit 1301 may be configured to perform the model-based application generation method in any other appropriate manner (e.g., by means of firmware).

[0268] Various embodiments of the systems and techniques described above herein may be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard parts (ASSPs), system on chip systems (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0269] The program code for implementing the method of the present invention can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code can be executed entirely on the machine, partially on the machine, partially on the machine as a stand-alone software package and partially on a remote machine, or entirely on a remote machine or server.

[0270] In the context of the present application, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0271] To provide interaction with a user, the systems and techniques described herein may be implemented on a computer having: a display device (e.g., a cathode ray tube or liquid crystal display, monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or path ball) through which the user can provide input to the computer. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0272] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: local area networks, wide area networks, and the Internet.

[0273] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0274] It should be noted that the serial numbers of the above-mentioned embodiments of the present application are only for description and do not represent the advantages or disadvantages of the embodiments.

[0275] In the above embodiments of the present application, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0276] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic, for example, the division of units is only a logical function division, and there may be other division methods in actual implementation, for example, 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 through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0277] 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.

[0278] In addition, each functional unit in each embodiment of the present application 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. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0279] If the integrated unit 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 application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes 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 various embodiments of the present application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, read-only memory random access memories, mobile hard disks, magnetic disks or optical disks.

[0280] The above are only preferred implementations of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. An information interaction system, characterized in that: include: The client is used to detect a query request in the interactive interface, wherein the query request at least includes query content to be answered; A text processing terminal, used for retrieving text associated with the query content based on the query content in the query request; A guidance information processing terminal, used to generate guidance information based on the retrieved text and the query content, wherein the guidance information is used to generate reply information corresponding to the query content; The information reply end is used to call the machine learning model to analyze the guidance information and output the reply information.

2. The system according to claim 1, characterized in that The client is used to rewrite the query request in response to the rewrite request in the interactive interface, wherein the amount of information of the query content in the rewritten query request is greater than the amount of information of the query content in the query request before the rewrite; The guidance information processing end is used to generate the guidance information based on the retrieved text and the rewritten query content in the query request.

3. The system according to claim 1, characterized in that The text processing end is used to perform structural processing on the retrieved text, wherein the structured text is used to represent the hierarchical structure of the text; The guidance information processing end is used to generate the guidance information based on the structured text and the query content.

4. The system according to claim 1, characterized in that The text processing end is used for segmenting the retrieved text in response to the text length of the retrieved text being greater than a text length threshold; The guidance information processing end is used to generate the guidance information based on the segmented text and the query content.

5. The system according to claim 1, characterized in that The system further comprises: The plug-in end is used to register the description data of the plug-in, wherein the description data is used to enable the information reply end to identify the plug-in and execute the function corresponding to the plug-in.

6. The system according to claim 5, characterized in that The text processing end is used to retrieve the description data associated with the query content based on the query content in the query request; The guidance information processing end is used to generate the guidance information based on the retrieved text, the retrieved description data and the query content.

7. The system according to claim 6, characterized in that The guide information processing end is used to splice the retrieved text and the retrieved description data to obtain a first splicing result, and generate the guide information based on the first splicing result and the query content.

8. The system according to claim 1, characterized in that The guidance information processing end is used to at least splice the multiple retrieved texts to obtain a second splicing result, and generate the guidance information based on the second splicing result and the query content.

9. The system according to claim 1, characterized in that The text processing end is used to send the retrieved text to the client; The client is used to modify the text in response to a first modification operation performed on the interactive interface; The guidance information processing end is used to generate the guidance information based on the modified text and the query content.

10. The system according to claim 1, characterized in that The guidance information processing end is used to send the guidance information to the client; The client is used to modify the guidance information in response to a second modification operation performed on the interactive interface.

11. The system according to claim 10, characterized in that The information reply end is used to display the modified guidance information, call the machine learning model to analyze the modified guidance information, and output the reply information.

12. The system according to any one of claims 1 to 11, characterized in that: The information reply terminal is used to send the retrieved text and the reply information to the client; The client is used to respond to the query operation of the reply information and display the text on the interactive interface.

13. An information interaction method, characterized in that: include: Detecting a query request in the interactive interface, wherein the query request at least includes query content to be answered; In response to the query request, displaying text associated with the query content on the interactive interface, wherein the text is retrieved based on the query content; Displaying guidance information on the interactive interface, wherein the guidance information is generated based on the retrieved text and the query content; The reply information is displayed on the interactive interface, wherein the reply information is obtained by analyzing the guidance information based on a machine learning model.

14. An information interaction method, characterized in that: include: Obtaining a query request from an application platform, wherein the query request at least includes query content to be answered in an application scenario where the application platform is located; In response to the query request, obtaining text associated with the query content, wherein the text is retrieved based on the query content; Generate guidance information based on the retrieved text and the query content, wherein the guidance information is used to generate reply information corresponding to the query content; Calling a machine learning model to analyze the guidance information to obtain the response information in the application scenario, wherein the machine learning model is trained based on application data samples in the application scenario; The reply information is output to the application platform.

15. The method according to claim 14, characterized in that The reply information satisfies the attributes of the application scenario and is used to describe multimodal information that matches the semantics of the query content, and the multimodal information includes at least one of the following: text information containing character information, video frame information containing frame image information, and audio information.

16. A method for generating information, characterized in that: include: Detecting a query request, wherein the query request at least includes query content to be answered; Based on the query content in the query request, retrieving text associated with the query content; Based on the retrieved text and the query content, guidance information is generated, wherein the guidance information is used to enable a machine learning model to analyze the guidance information to obtain response information.

17. An electronic device, characterized in that: include: A memory storing an executable program; A processor, configured to run the program, wherein the program executes the method according to any one of claims 13 to 16 when running.

Images