Knowledge-based question answering methods, devices, media, equipment, and products based on large models

Abstract

A knowledge question answering method, apparatus, medium, device, and product based on a large model, relating to the field of computer technology, is disclosed. This method determines the query category to which the query intent corresponding to the second content belongs through a first large model. When the query category is the first, the first and second content are input into a second large model to obtain the first query result corresponding to the query request output by the second large model. When the query category is the second, the first and second content are input into the first large model to obtain the third content output by the first large model. Then, based on the third content, the second query result corresponding to the query request is obtained. This method not only supports users to perform knowledge question answering using multimodal query requests but also eliminates the need for users to configure specific question answering strategies. For complex questions, it effectively converts the first content into searchable text questions, greatly improving the efficiency of users asking questions and increasing the accuracy of answers.

Description

Technical Field

[0001] This disclosure relates to the field of computer technology, and more specifically, to a knowledge question-answering method, apparatus, medium, device, and product based on a large model. Background Technology

[0002] In knowledge-based question-and-answer scenarios, plain text input is generally the primary method. If users need to combine image information with their questions, they typically need to manually describe the image content, which can lead to inaccurate descriptions and cumbersome operations. Although technological advancements have enabled question-and-answer systems to support complex queries containing multimodal data, the lack of dynamic adjustment strategies makes it difficult to guarantee the relevance and accuracy of the answers. Summary of the Invention

[0003] This summary section is provided to briefly introduce the concepts, which will be described in detail in the detailed description section below. This summary section is not intended to identify key or essential features of the claimed technical solution, nor is it intended to limit the scope of the claimed technical solution.

[0004] Firstly, this disclosure provides a knowledge question answering method based on a large model, including:

[0005] Obtain a query request, the query request including first content and second content, the first content and the second content belong to different content types, and the second content is used to describe the query intent;

[0006] The first major model is used to determine the query category to which the query intent corresponding to the second content belongs; wherein, the query category includes a first query category and a second query category, the first query category represents the generation of query results corresponding to the second content through the first content, and the second query category represents the generation of query results corresponding to the second content through knowledge retrieval;

[0007] In response to the query category corresponding to the second content being the first query category, the first content and the second content are input into the second large model to obtain the first query result corresponding to the query request output by the second large model;

[0008] In response to the query category corresponding to the second content being the second query category, the first content and the second content are input into the first large model to obtain the third content output by the first large model; wherein, the first large model is used to adjust the second content according to the information identified from the first content to obtain the third content, and the content type of the third content is consistent with the content type of the second content;

[0009] Based on the third content, the second query result corresponding to the query request is obtained.

[0010] Secondly, this disclosure provides a knowledge question answering device based on a large model, including:

[0011] The acquisition module is used to acquire a query request, which includes first content and second content. The first content and the second content belong to different content types, and the second content is used to describe the query intent.

[0012] The first determining module is used to determine the query category to which the query intent corresponding to the second content belongs through the first large model; wherein, the query category includes a first query category and a second query category, the first query category represents the query result corresponding to the second content generated through the first content, and the second query category represents the query result corresponding to the second content generated through knowledge retrieval;

[0013] The first response module is used to respond to the query category corresponding to the second content being the first query category, input the first content and the second content into the second large model, and obtain the first query result corresponding to the query request output by the second large model;

[0014] The second response module is used to respond to the query category corresponding to the second content being the second query category, inputting the first content and the second content into the first large model to obtain the third content output by the first large model; wherein, the first large model is used to adjust the second content according to the information identified from the first content to obtain the third content, and the content type of the third content is consistent with the content type of the second content;

[0015] The second determining module is used to obtain the second query result corresponding to the query request based on the third content.

[0016] Thirdly, this disclosure provides a computer-readable medium having a computer program stored thereon, which, when executed by a processing device, implements the steps of the method described in the first aspect.

[0017] Fourthly, this disclosure provides an electronic device, comprising:

[0018] A storage device on which computer programs are stored;

[0019] A processing device for executing the computer program in the storage device to implement the steps of the method described in the first aspect.

[0020] Fifthly, this disclosure provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the method described in the first aspect.

[0021] Based on the above technical solution, by acquiring a query request including first and second content, and using a first large model to determine the query category to which the query intent corresponding to the second content belongs, when the query category corresponding to the second content is the first query category, the first and second content are input into the second large model to obtain the first query result corresponding to the query request output by the second large model. When the query category corresponding to the second content is the second query category, the first and second content are input into the first large model to obtain the third content output by the first large model. Then, based on the third content, the second query result corresponding to the query request is obtained. This not only supports users to use multimodal query requests for knowledge-based question answering, making the knowledge-based question answering method more in line with users' daily usage habits, but also utilizes the capabilities of the large model in the front-end chain of the question-answering system to identify the question-answering strategy corresponding to the query request. Users do not need to configure specific question-answering strategies; instead, the large model automatically judges them, and users only need to perceive the query results at the end-to-end level. This simplifies user operations while enhancing the accuracy of knowledge-based question answering. Secondly, for complex questions, it can also effectively convert the first content into searchable second content, greatly improving the efficiency of users asking questions and increasing the accuracy of answers.

[0022] Other features and advantages of this disclosure will be described in detail in the following detailed description section. Attached Figure Description

[0023] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and the originals and elements are not necessarily drawn to scale. In the drawings:

[0024] Figure 1 This is a flowchart illustrating a knowledge question-answering method based on a large model, according to some embodiments.

[0025] Figure 2 This is a flowchart illustrating a knowledge question-answering method based on a large model, according to some other embodiments.

[0026] Figure 3 This is a schematic diagram of the module connections of a knowledge question-answering device based on a large model, provided according to an exemplary embodiment.

[0027] Figure 4 This is a schematic diagram of module connections of an electronic device according to an exemplary embodiment. Detailed Implementation

[0028] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0029] It should be understood that the steps described in the method embodiments of this disclosure may be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of this disclosure is not limited in this respect.

[0030] The term "comprising" and its variations as used herein are open-ended inclusions, meaning "including but not limited to". The term "based on" means "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". Definitions of other terms will be given in the description below.

[0031] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are used only to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.

[0032] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0033] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0034] It is understood that before using the technical solutions disclosed in the various embodiments of this disclosure, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in this disclosure in an appropriate manner in accordance with relevant laws and regulations, and user authorization should be obtained.

[0035] For example, upon receiving a user's active request, a prompt message is sent to the user to explicitly inform them that the requested operation will require the acquisition and use of the user's personal information. This allows the user to independently choose whether to provide personal information to the software or hardware, such as the electronic device, application, server, or storage medium performing the operations of this disclosed technical solution, based on the prompt message.

[0036] As an optional but non-limiting implementation, in response to a user's active request, sending a prompt message to the user can be done via a pop-up window, where the prompt message can be presented in text format. Furthermore, the pop-up window can also include a selection control allowing the user to choose "agree" or "disagree" to provide personal information to the electronic device.

[0037] It is understood that the above notification and user authorization process are merely illustrative and do not constitute a limitation on the implementation of this disclosure. Other methods that comply with relevant laws and regulations may also be applied to the implementation of this disclosure.

[0038] Meanwhile, it is understood that the data involved in this technical solution (including but not limited to the data itself, the acquisition or use of the data) shall comply with the requirements of relevant laws, regulations and related provisions.

[0039] Figure 1 This is a flowchart illustrating a knowledge question-answering method based on a large model, according to some embodiments. For example... Figure 1 As shown, this disclosure provides a knowledge question answering method based on a large model, specifically executed by a knowledge question answering device based on a large model, which can be implemented in software and / or hardware. Figure 1 As shown, the method may include the following steps.

[0040] In step 110, a query request is obtained. The query request includes first content and second content. The first content and second content belong to different content types, and the second content is used to describe the query intent.

[0041] Here, a query request can refer to input content submitted by a user to a question-and-answer system for obtaining information or triggering an operation. In this embodiment of the disclosure, the query request includes first content and second content, and the first content and second content belong to different content types; that is, the first content and second content actually belong to different modalities. In the query request, the second content is used to describe the user's query intent, while the first content can be understood as reference information used to assist in describing the query intent.

[0042] It should be understood that a query request can be interpreted as an information retrieval request initiated by a user that involves two or more data modalities. Compared to single-text-modal query requests, multimodal query requests require the question-answering system to be able to process and understand multiple modalities of data simultaneously in order to more comprehensively capture user intent and provide accurate answers.

[0043] For example, in this embodiment of the disclosure, the first content may be at least one media content selected from images, documents, videos, and audio, and the second content may be a text question. The first content and the second content are data belonging to different modalities. A text question can refer to a natural language question expressed in natural language for querying knowledge. It should be understood that a text question can be a user request input by the user through natural language description. Of course, a text question can also be a user request expressed in natural language description obtained from the user's speech.

[0044] For example, a user can upload an image containing a chart and an accompanying text description to inquire about a scientific concept from the question-and-answer system. Accordingly, the uploaded image is the first content, and the text description is the second content; thus, the uploaded image and text description constitute the query request. It should be understood that, in this embodiment of the disclosure, there is no limit to the number of images a user can upload.

[0045] It should be noted that, in this embodiment of the disclosure, the user's input query request can be obtained through a displayed graphical user interface. For example, the graphical user interface can provide controls for uploading images and / or documents, and input boxes for entering text questions. Users can upload images and / or documents (first content) through the controls, and input second content through the input boxes. The images entered by the user can be displayed sequentially above the input boxes, and users can preview the images and delete slots, reducing the operational threshold for users. Through multimodal query requests, users can directly express their needs using images and other first content.

[0046] In step 120, the query category to which the query intent corresponding to the second content belongs is determined through the first big model; wherein, the query category includes a first query category and a second query category, the first query category represents the query result corresponding to the second content generated through the first content, and the second query category represents the query result corresponding to the second content generated through knowledge retrieval.

[0047] Here, the second piece of content can be input into the first main model, which outputs the query category to which the query intent corresponding to the second piece of content belongs. It should be understood that the first main model can perform semantic understanding on the second piece of content, thereby identifying the query intent described by the second piece of content, and then determining the query category to which the query intent belongs.

[0048] In this disclosure, "large model" can refer to a multimodal large model, which is a large-scale artificial intelligence model capable of simultaneously understanding, processing, and generating multiple types of information (i.e., "modalities"). For example, a multimodal large model can be a vision-language model (VLM). A vision-language large model is a multimodal generative artificial intelligence system that combines a large language model with a visual encoder to achieve comprehensive processing of images, text, and video.

[0049] In this embodiment of the disclosure, the query category to which the query intent corresponding to the second content belongs includes a first query category and a second query category. The first query category represents the generation of query results corresponding to the second content through the first content, while the second query category represents the generation of query results corresponding to the second content through knowledge retrieval.

[0050] It's important to clarify that the first query category essentially means that the first content contains sufficient information to generate the query results corresponding to the second content. In other words, for the first query category, the query results are directly derived from the first content provided by the user; the first content is the direct source of the query results corresponding to the query intent of the first query category. For example, if the first content is an image, and the query intent corresponding to the second content is asking how many apples are in the image, then the query intent corresponding to the second content can be answered directly through the image itself. Another example is if a user uploads a product image (equivalent to the first content), and the corresponding second content is "What is the brand of this product?", then the query results for this second content will be found in the product image. Yet another example is if a user uploads a screenshot of a table (equivalent to the first content), and the corresponding second content is "What was the sales volume in 2023?", then the query results will be found within the image content of the table screenshot.

[0051] The second question type indicates that the information in the first content is insufficient to generate query results corresponding to the query intent of the second content. In other words, for the second content of the second query category, the corresponding query results cannot be obtained solely from the first content provided by the user; knowledge retrieval is required. For example, if the first content is a design image, and the second content is "search for other design drafts similar to the design image," other similar design drafts cannot be directly obtained from the design image; therefore, the query category for this second content belongs to the second query category. Another example is a user uploading a picture of medicine packaging; the corresponding second content is "What are the side effects of this medicine?" The packaging image may only show the drug name; the side effects need to be retrieved from a knowledge base.

[0052] In step 130, in response to the query category corresponding to the second content being the first query category, the first content and the second content are input into the second large model to obtain the first query result corresponding to the query request output by the second large model.

[0053] Here, if the query category corresponding to the second content is the first query category, then the first content and the second content can be input into the second large model to obtain the first query result corresponding to the query request.

[0054] It should be understood that when the query category corresponding to the second content is the first query category, the query result representing the query intent described by the second content is actually within the first content provided by the user. Therefore, by performing multimodal understanding of the first and second content through the second major model, the first query result corresponding to the second content of the first query category can be generated.

[0055] Taking the first content as including images and documents as an example, the document content can be extracted from the document, and then the images, document content and the second content can be input into the second large model. The second large model outputs the first query result corresponding to the query intent of the second content based on the images, document content and the second content, thereby directly generating the first query result corresponding to the second content of the first query category by utilizing the model capabilities of the second large model.

[0056] It is worth noting that for the second content in the first query category, since the query results corresponding to the second content are already in the first content provided by the user, there is no need to obtain the query results corresponding to the second content through a question-and-answer system, thereby improving the accuracy and efficiency of generating query results.

[0057] In some embodiments, in response to a selection operation for a target large model among multiple candidate large models, the selected target large model can be determined as the second large model.

[0058] This system allows users to pre-configure multiple large models for selection. These models can have different performance characteristics or be responsible for different types of tasks. For example, it can provide large model A, large model B, and large model C for users to choose from. Large model A excels at text-based question answering. Large model B has high accuracy but high computational cost. Large model C is lightweight but has low accuracy.

[0059] Once a user selects a target large model from multiple available large models through a selection operation, the selected target large model will be used as the second large model. Users can select the target large model through methods such as clicking, specifying via a drop-down menu, or passing in parameters.

[0060] It should be understood that by selecting the target large model as the second large model, users can choose different large models as the second large model according to different needs, thereby adapting to different task requirements of users.

[0061] In step 140, in response to the query category corresponding to the second content being the second query category, the first content and the second content are input into the first large model to obtain the third content output by the first large model.

[0062] Here, if the query category corresponding to the second content is the second query category, the first content and the second content can be input into the first large model to obtain the third content output by the first large model. The first large model is used to adjust the second content based on the information identified from the first content to obtain the third content, and the content type of the third content is consistent with the content type of the second content. For example, the third content can be a text question.

[0063] It should be understood that when the query category corresponding to the second content is the second query category, it actually means that the query result corresponding to the query intent described by the second content cannot be obtained through the first content. Instead, the query result corresponding to the query intent described by the second content needs to be obtained through knowledge retrieval. In this case, the first model can be used to rewrite the second content based on the first and second content to obtain the third content. The third content includes the information identified by the first model from the first content.

[0064] Rewriting the second content using the first model can be achieved by combining the information identified by the first model from the first content with the second content, thereby rewriting the second content to obtain a more complete and specific third content. This rewritten third content more clearly describes the user's query intent. The information identified by the first model from the first content can be textual information suitable for knowledge retrieval.

[0065] For example, combining the information identified by the first model from the first content with the second content can be used to reconstruct the second content using this information, resulting in the third content. Alternatively, combining the information identified by the first model from the first content with the second content can also involve adding the identified information as supplementary information, scenarios, conditions, or constraints to the second content, thus obtaining the third content.

[0066] Taking the first content as an image as an example, suppose a user uploads a picture of Lina Belle, and the corresponding text question is "Please introduce the cartoon character in the picture." The first model recognizes the following information from the picture: "There is a pink fox in the picture, with a round face, big sapphire blue eyes and long eyelashes, big fluffy ears on its head, a light purple orchid on its ear, and a fluffy tail trailing behind it." The rewritten third content can then be: "Please introduce the cartoon character in the picture. There is a pink fox in the picture, with a round face, big sapphire blue eyes and long eyelashes, big fluffy ears on its head, a light purple orchid on its ear, and a fluffy tail trailing behind it."

[0067] It is worth noting that in this embodiment of the disclosure, the first content and the second content are input into the first large model. The first large model determines the query category corresponding to the second content. If the query category corresponding to the second content is the second query category, the first large model rewrites the second content based on the first content and the second content to obtain the third content.

[0068] In other words, the first model can be used to identify the query category corresponding to the second content, and then select different question-answering strategies to obtain the query results corresponding to the query intent described by the second content based on the different identified query categories. It can also be used to rewrite the second content to obtain the third content.

[0069] It's important to note that by rewriting the second content based on the first content, the second content, and the first main model to obtain the third content, the first content can be converted into textual information capable of knowledge retrieval. This improves the efficiency of users asking questions and enhances the accuracy of query results generated by the question-and-answer system. Furthermore, users are no longer required to manually describe the first content using natural language, reducing the difficulty of asking questions and improving the accuracy of the answers. For example, if the first content includes an image, the image can be converted into textual information capable of knowledge retrieval. Rewriting the second content using the first content allows users to express their needs using images and other elements of the first content, combining them with the second content to generate the third content, thus making the questioning method more aligned with everyday communication habits.

[0070] In step 150, the second query result corresponding to the query request is obtained based on the third content.

[0071] Here, for example, a third type of content can be input into the question-answering system to obtain a second query result corresponding to the query request output by the question-answering system. The question-answering system is used to generate query results through knowledge retrieval. Specifically, the question-answering system obtains query results through a knowledge-based question-answering chain, which refers to the complete process of a series of automated processing modules or technical steps that the question-answering system undergoes from receiving a text question to generating the final query result. The text question is input into the question-answering system, which then sequentially performs a series of operations such as intent recognition, knowledge retrieval, reasoning, and answer generation to obtain the query result output by the question-answering system.

[0072] For example, a question-answering system may include an understanding module, a context enhancement module, a knowledge retrieval module, a knowledge fusion module, an answer generation module, and a post-processing module. The understanding module performs operations such as word segmentation, entity recognition, intent classification, and query category determination on the text question. The knowledge retrieval module retrieves documents or facts related to the user's intent corresponding to the text question from a knowledge base. The knowledge fusion module scores the retrieval results and selects the most relevant segments. The answer generation module generates natural language answers based on the retrieval results. The post-processing module performs operations such as format validation, sensitivity filtering, confidence assessment, and citation tracing on the answer content.

[0073] It should be understood that the configuration functions of different types of question-answering systems may vary. However, regardless of the specific principles behind a question-answering system, it is used to obtain query results corresponding to the input text question. In other words, different types of question-answering systems are at least used to: retrieve knowledge fragments related to the text question from the knowledge base, input the knowledge fragments and the text question into a large model, and obtain query results corresponding to the text question.

[0074] It is worth noting that, in this embodiment of the disclosure, before the question-answering system, the query category to which the query intent corresponding to the second content belongs is determined in advance through a first large model. Then, different question-answering strategies are selected according to different query categories to obtain the query results corresponding to the query request. When the query category to which the query intent corresponding to the second content belongs is a first query category, the corresponding question-answering strategy is to input the first content and the second content into the second large model to obtain the first query result corresponding to the query request output by the second large model. When the query category to which the query intent corresponding to the second content belongs is a second query category, the corresponding question-answering strategy is to input the first content and the second content into the first large model to obtain the third content output by the first large model, and obtain the second query result corresponding to the query request based on the third content.

[0075] In other words, in the front-end of a question-and-answer system, a large model can be used to identify user-initiated query requests and automatically plan answer strategies. Users do not need to configure specific question-and-answer strategies; instead, the large model makes automatic judgments, and users can perceive the query results in the end-to-end effect.

[0076] Therefore, by acquiring a query request including first and second content, and using a first large model to determine the query category to which the query intent corresponding to the second content belongs, when the query category corresponding to the second content is the first query category, the first and second content are input into the second large model to obtain the first query result corresponding to the query request output by the second large model. When the query category corresponding to the second content is the second query category, the first and second content are input into the first large model to obtain the third content output by the first large model. Then, based on the third content, the second query result corresponding to the query request is obtained. This not only supports users to use multimodal query requests for knowledge-based question answering, making the knowledge-based question answering method more in line with users' daily usage habits, but also utilizes the capabilities of the large model in the front-end chain of the question-answering system to identify the question-answering strategy corresponding to the query request. Users do not need to configure specific question-answering strategies; instead, the large model automatically judges them, and users only need to perceive the query results at the end-to-end level. This simplifies user operations while enhancing the accuracy of knowledge-based question answering. Secondly, for complex questions, it can also effectively convert the first content into searchable second content, greatly improving the efficiency of users asking questions and increasing the accuracy of answers.

[0077] In some feasible implementations, in step 120, the query intent corresponding to the second content can be determined by the first large model, and then, in response to the query intent being a search type, the query category to which the second content belongs is determined as the second query category, and in response to the query intent being a non-search type, the query category to which the second content belongs is determined as the first query category.

[0078] Here, the first and second content can be input into the first big model. The first big model can perform semantic parsing on the second content input by the user to identify its corresponding query intent.

[0079] In this embodiment of the disclosure, the query intent can include search-type intent and non-search-type intent. Search-type intent refers to a query intent implied by the second content initiated by the user that requires knowledge retrieval to achieve, while non-search-type intent refers to a query intent implied by the second content initiated by the user that can be achieved without knowledge retrieval. For example, if the second content initiated by the user is "Help me search for the latest industry news," its query intent is a search-type intent. If the second content initiated by the user is "How many apples are in the picture?", its query intent is a non-search-type intent.

[0080] When the query intent corresponding to the second piece of content initiated by the user is determined to be a search-type intent using the first major model, the query category to which the second piece of content belongs is determined to be the second query category. When the query intent corresponding to the second piece of content initiated by the user is determined to be a non-search-type intent using the first major model, the query category to which the second piece of content belongs is determined to be the first query category.

[0081] It should be understood that the first model can obtain the query intent corresponding to the second content through semantic recognition, or it can combine the second content and the first content to obtain the query intent corresponding to the second content. Moreover, after determining the query category to which the second content belongs, if the query category to which the second content belongs is the first query category, the first model sends the second content and the first content to the second model, so that the second model can obtain the first query result corresponding to the query request based on the first content and the second content. If the query category to which the second content belongs is the second query category, the first content and the second content can be input into the first model to obtain the third content output by the first model, and the third content is sent to the question answering system to obtain the second query result corresponding to the user's query request.

[0082] Therefore, the first major model can accurately identify the query intent corresponding to the user's second piece of content, thereby accurately classifying the second piece of content and accurately planning the question-and-answer strategy for the user's query request. Users do not need to configure specific question-and-answer strategies; instead, the first major model automatically makes the judgment, and users only need to perceive the query results in the end-to-end effect. This simplifies user operations while enhancing the accuracy of knowledge-based question answering. Secondly, for complex questions, it can also effectively convert the first piece of content into searchable second piece of content, greatly improving the efficiency of users asking questions and increasing the accuracy of answers.

[0083] In some feasible implementations, in response to a query intent of the type of search, the query object corresponding to the query intent of the second content can be determined in the first content based on the second content, and in response to the query object containing non-textual information that cannot be converted into text, the query category to which the second content belongs can be determined as the first query category.

[0084] Here, when determining that the query intent corresponding to the second content is a search-type intent, the query object targeted by the query intent corresponding to the second content can be determined from the first content based on the second content. The query object refers to the specific target that the query intent corresponding to the second content points to in the first content. For example, the query object could be a certain area in an image, a certain action in a video, a certain sound in an audio file, etc.

[0085] For example, if the first content is an image, and a user uploads a design draft image and asks, "Is there a latest design for this draft?", then the query intent for the second content is targeting the image itself, not the specific content within the image. As another example, if the first content is an image, and a user uploads a picture of Rena Belle, and the corresponding second content is "Please describe the cartoon character in the picture," then the query intent for the second content is targeting the content within the image.

[0086] After identifying the query object, it is determined whether the query object contains non-textual information that cannot be rewritten as text. If it does contain non-textual information that cannot be rewritten as text, the query category to which the query intent corresponding to the second content belongs is determined to be the first query category. If it does not contain non-textual information that cannot be rewritten as text, the query category to which the query intent corresponding to the second content belongs is determined to be the second query category.

[0087] It should be understood that non-textual information in a query object that cannot be rewritten into text refers to information that cannot be completely, accurately, and losslessly converted into a natural language text description. For example, taking an image as the first content, if a user uploads an image containing a math problem, which includes a textual stem and geometric figures, the geometric figures are non-textual information that cannot be rewritten into text. In other words, during the adjustment of the second content, the geometric figures cannot be accurately rewritten into the textual information in the third content. Given this inaccuracy, if a question-and-answer strategy of "inputting the first and second content into a first-level model, obtaining the third content output by the first-level model, and then obtaining the second query result corresponding to the query request based on the third content" is used to obtain the second query result, the final result may also be inaccurate. Therefore, when a query object contains non-textual information that cannot be rewritten into text, the query category to which the query intent corresponding to the second content belongs can be determined as the first query category. Thus, a question-and-answer strategy of "inputting the first and second content into a second-level model, obtaining the first query result corresponding to the query request output by the second-level model" can be used to obtain the first query result corresponding to the query request.

[0088] It's important to note that if the query object contains non-textual information that cannot be rewritten into text, it means the corresponding query object cannot be accurately described using natural language. Accordingly, it can be understood that the query object cannot be rewritten into a third type of content. Therefore, a question-and-answer strategy of "inputting the first and second types of content into the second main model to obtain the first query result corresponding to the query request output by the second main model" is needed to obtain the query result corresponding to the query request. When the query object does not contain non-textual information that cannot be rewritten into text, then the query object can be rewritten into a third type of content. Accordingly, a question-and-answer strategy of "inputting the first and second types of content into the first main model to obtain the third type of content output by the first main model; and obtaining the second query result corresponding to the query request based on the third type of content" can be used to obtain the query result corresponding to the query request.

[0089] Therefore, through the above implementation method, the query category to which the query intent corresponding to the second content belongs can be accurately determined, thereby accurately determining the question-and-answer strategy corresponding to the query request, so that the answer can be more accurate.

[0090] In some feasible implementations, step 150 may further verify whether the third content conforms to a preset rule; in response to the third content conforming to the preset rule, a second query result corresponding to the query request is obtained based on the third content; in response to the third content not conforming to the preset rule, the first content and the second content are input into the second large model to obtain the third query result corresponding to the query request output by the second large model.

[0091] Here, the third content is generated by rewriting the second and first content in combination with the first major model. It may have semantic errors, unclear references, and other problems. By verifying whether the third content conforms to the preset rules, we can determine whether the quality of the third content obtained by rewriting the second content meets the standards.

[0092] For example, preset rules may include semantic integrity rules, referential validity rules, etc. Semantic integrity rules are used to restrict the rewritten third content from being a plain text description of the first content. For instance, suppose a user uploads a picture of Rena Belle, and the corresponding second content is "Please describe the cartoon character in the picture." The third content rewritten by the first model is "Please describe the cartoon character in the picture. The picture shows a pink fox with a round face, large sapphire blue eyes and long eyelashes, large fluffy ears on its head, a pale purple orchid on its ear, and a fluffy tail trailing behind it." The third content, "The picture shows a pink fox with a round face, large sapphire blue eyes and long eyelashes, large fluffy ears on its head, a pale purple orchid on its ear, and a fluffy tail trailing behind it," is a plain text description of the picture content. Clearly, since the first model does not recognize the cartoon character in the picture as Rena Belle, the query results generated by the question-answering system using this rewritten third content are highly likely to be inaccurate. Therefore, it can be determined that the third content does not conform to the preset rules. Accordingly, the third query result corresponding to the query request can be obtained through a question-and-answer strategy of "inputting the first and second content into the second large model to obtain the third query result corresponding to the query request output by the second large model". It should be understood that the principle of obtaining the third query result is the same as that of obtaining the first query result, and will not be elaborated here.

[0093] The referential validity rule restricts the rewritten third content from referencing the first content. For example, the third content cannot contain descriptions such as "this picture" or "that picture." However, if the rewritten third content from the first model is "Please introduce Rena Belle" or "Please introduce the cartoon character Rena Belle," then the rewritten third content conforms to the preset rule. Accordingly, the second query result corresponding to the query request can be obtained through a question-and-answer strategy of "obtaining the second query result corresponding to the query request based on the third content."

[0094] It should be understood that the first model can be used to verify whether the third content conforms to preset rules. When using the first model to verify the third content's conformity, since the first model is a black-box model, pre-configured prompts can be used to guide it to conform to the preset rules indicated by the prompts when rewriting the second content. If the third content obtained after multiple rewrites by the first model still does not conform to the preset rules, a question-and-answer strategy of "inputting the first and second content into the second model to obtain the third query result corresponding to the query request output by the second model" can be used to obtain the third query result corresponding to the query request. In other words, if the third content obtained after multiple rewrites by the first model does not conform to the preset rules, answering with the first content itself may yield better results. If the third content ultimately obtained by the first model conforms to the preset rules, a question-and-answer strategy of "obtaining the second query result corresponding to the query request based on the third content" can be used to obtain the second query result corresponding to the query request.

[0095] Therefore, through the above implementation method, by verifying whether the third content conforms to the preset rules, different question-and-answer strategies can be selected to obtain the query results corresponding to the query request, thereby ensuring that the query results corresponding to the query request can be more accurate.

[0096] In some feasible implementations, after obtaining the query request, the request length corresponding to the query request can also be determined; if the request length is less than or equal to a preset threshold, the query category to which the query intent corresponding to the second content belongs is determined through the first major model; if the request length is greater than the preset threshold, the first content and the second content are input into the third major model to obtain the fourth content output by the third major model; wherein, the third major model is used to adjust the second content according to the information identified from the first content to obtain the fourth content, and the content type of the fourth content is consistent with the content type of the second content; based on the fourth content, the fourth query result corresponding to the query request is obtained.

[0097] Here, after obtaining the query request, the request length corresponding to the obtained query request can be determined. The request length can refer to the length of the document corresponding to the user-initiated query request. For example, the request length corresponding to the query request can include a first number of documents included in the query request, a second number of passages included in the query request, and a third number of tokens used by the query request during model processing.

[0098] Different preset thresholds can be assigned to the first, second, and third quantities. For example, if the query request includes documents, the corresponding request length, including the first quantity of documents, has a preset threshold of 5 documents; that is, the first quantity of documents is less than or equal to 5. If the query request includes articles, the corresponding request length, including the second quantity of articles, has a preset threshold of 100 articles; that is, the second quantity of articles is less than or equal to 100. The number of articles in the query request does not exceed 100. If the query request length includes the third quantity of tokens, the corresponding preset threshold is 30kb; that is, the third quantity of tokens in the query request cannot exceed 30kb.

[0099] By determining whether the length of the query request is less than or equal to a preset threshold, it can be judged whether the query request is excessively long. If the request length is less than or equal to the preset threshold, the query request can be sent to the first main model to determine the query category to which the query intent corresponding to the second content belongs. If the request length exceeds the preset threshold, the first and second content can be input into the third main model to obtain the fourth content output by the third main model. The fourth content is then input into the question-answering system to obtain the fourth query result corresponding to the query request output by the question-answering system.

[0100] It should be understood that the concept of the fourth content is consistent with the concept of the third content, and the principle for obtaining the fourth content is consistent with the principle for obtaining the third content, which will not be elaborated here. Similarly, the principle for the fourth query result corresponding to the query request output by the question-and-answer system is consistent with the principle for the third query result corresponding to the query request output by the question-and-answer system, which will not be elaborated here.

[0101] For example, if the request length exceeds a preset threshold, the first and second content can be input into a third model. The third model then rewrites the second content to obtain the fourth content. Based on the fourth content and the question-and-answer system, the fourth query result corresponding to the query request is obtained. It should be understood that the specific logic for obtaining the fourth content is consistent with the logic for obtaining the third content, and will not be elaborated here.

[0102] By determining the length of the query request, it can be identified whether the query request is excessively long. If the query request is excessively long, the fourth content can be obtained directly through the third model and sent to the question-and-answer system to obtain the fourth query result corresponding to the user's query request.

[0103] It is worth noting that after obtaining the fourth content, it is also possible to verify whether the fourth content conforms to the preset rules. If the fourth content conforms to the preset rules, the fourth query result corresponding to the query request is obtained based on the fourth content and the question-and-answer system. If the fourth content does not conform to the preset rules, the query result corresponding to the query request is obtained based on the first content, the second content, and the second major model.

[0104] Of course, in other implementations, it can be determined in advance whether the query request contains an image. If the query request contains an image, the length of the corresponding query request can be further determined. If the query request contains an image, the query request can be directly sent to the question-and-answer system.

[0105] Therefore, through the above implementation methods, different strategies can be used to process query requests of different lengths, thereby enabling the question-and-answer system to automatically plan the question-and-answer strategy corresponding to the query request.

[0106] The following is in conjunction with the appendix Figure 2 The above-described embodiments will be described in detail.

[0107] Figure 2 This is a flowchart illustrating a knowledge question-answering method based on a large model, according to some other embodiments. For example... Figure 2 As shown, a query request is obtained, and then it is determined whether the request length is less than or equal to a preset threshold. If the request length is less than or equal to the preset threshold, and if the first content corresponding to the query request includes a document, the document content is obtained. Furthermore, the query category to which the query intent corresponding to the second content belongs is determined through a first main model. If it belongs to the first query category, the first and second content are input into a second main model to obtain the first query result corresponding to the query request output by the second main model. If it belongs to the second query category, the first and second content are input into the first main model to obtain the third content output by the first main model. The third content is then input into the question-answering system to obtain the second query result corresponding to the query request output by the question-answering system. If the request length is greater than the preset threshold, the first and second content can be input into a third main model to obtain the fourth content output by the third main model. The fourth content is then input into the question-answering system to obtain the fourth query result corresponding to the query request output by the question-answering system.

[0108] Figure 3 This is a schematic diagram of the module connections of a knowledge question-answering device based on a large model, provided according to an exemplary embodiment. For example... Figure 3 As shown, this disclosure provides a knowledge question answering device 300 based on a large model, which may include:

[0109] The acquisition module 301 is used to acquire a query request, the query request including first content and second content, the first content and the second content belong to different content types, and the second content is used to describe the query intent;

[0110] The first determining module 302 is used to determine the query category to which the query intent corresponding to the second content belongs through the first large model; wherein, the query category includes a first query category and a second query category, the first query category represents the query result corresponding to the second content generated through the first content, and the second query category represents the query result corresponding to the second content generated through knowledge retrieval;

[0111] The first response module 303 is used to respond to the query category corresponding to the second content being the first query category, input the first content and the second content into the second large model, and obtain the first query result corresponding to the query request output by the second large model;

[0112] The second response module 304 is used to respond to the query category corresponding to the second content being the second query category, inputting the first content and the second content into the first large model to obtain the third content output by the first large model; wherein, the first large model is used to adjust the second content according to the information identified from the first content to obtain the third content, and the content type of the third content is consistent with the content type of the second content;

[0113] The second determining module 305 is used to obtain the second query result corresponding to the query request based on the third content.

[0114] Optionally, the first determining module 302 is specifically used for:

[0115] The first major model is used to determine the query intent corresponding to the second content;

[0116] In response to the query intent being a search type, the query category to which the second content belongs is determined to be the second query category;

[0117] In response to the query intent being a non-search type intent, the query category to which the second content belongs is determined to be the first query category.

[0118] Optionally, the first determining module 302 is specifically used for:

[0119] In response to the query intent being a search type, the query object targeted by the query intent corresponding to the second content is determined in the first content based on the second content;

[0120] In response to the query object containing non-textual information that cannot be converted into text, the query category to which the second content belongs is determined to be the first query category.

[0121] Optionally, the second determining module 305 is specifically used for:

[0122] Verify whether the third content conforms to preset rules;

[0123] In response to the third content conforming to preset rules, a second query result corresponding to the query request is obtained based on the third content;

[0124] In response to the fact that the third content does not conform to the preset rule, the first content and the second content are input into the second large model to obtain the third query result corresponding to the query request output by the second large model.

[0125] Optionally, the knowledge question answering device 300 based on a large model further includes:

[0126] The third determining module is used to determine the selected target large model as the second large model in response to the selection operation of the target large model among multiple candidate large models.

[0127] Optionally, the first content includes at least one of images, documents, videos, and audio.

[0128] Optionally, the first determining module 302 is specifically used for:

[0129] Determine the request length corresponding to the query request;

[0130] If the request length is less than or equal to a preset threshold, the query category to which the query intent corresponding to the second content belongs is determined through the first major model.

[0131] If the request length is greater than the preset threshold, the first content and the second content are input into the third model to obtain the fourth content output by the third model; wherein, the third model is used to adjust the second content according to the information identified from the first content to obtain the fourth content, and the content type of the fourth content is consistent with the content type of the second content;

[0132] Based on the fourth content, the fourth query result corresponding to the query request is obtained.

[0133] Regarding the knowledge question-answering device 300 based on a large model in the above embodiments, the method logic executed by each functional module has been described in detail in the section on methods, and will not be repeated here.

[0134] The following is for reference. Figure 4The diagram illustrates a structural schematic of an electronic device (e.g., a terminal device or a server) 400 suitable for implementing embodiments of the present disclosure. The terminal device in the embodiments of the present disclosure may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 4 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments disclosed herein.

[0135] like Figure 4 As shown, electronic device 400 may include a processing device (e.g., a central processing unit, a graphics processor, etc.) 401, which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 402 or a program loaded from storage device 408 into random access memory (RAM) 403. RAM 403 also stores various programs and data required for the operation of electronic device 400. Processing device 401, ROM 402, and RAM 403 are interconnected via bus 404. Input / output (I / O) interface 405 is also connected to bus 404.

[0136] Typically, the following devices can be connected to I / O interface 405: input devices 406 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 407 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 408 including, for example, magnetic tapes, hard disks, etc.; and communication devices 409. Communication device 409 allows electronic device 400 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 4 An electronic device 400 with various devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively.

[0137] In particular, according to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication device 409, or installed from storage device 408, or installed from ROM 402. When the computer program is executed by processing device 401, it performs the functions defined in the methods of embodiments of this disclosure.

[0138] It should be noted that the computer-readable medium described in this disclosure can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this disclosure, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this disclosure, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0139] In some implementations, clients and servers can communicate using any currently known or future-developed network protocol such as HTTP (Hypertext Transfer Protocol) and can interconnect with digital data communication (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (“LANs”), wide area networks (“WANs”), the Internet (e.g., the Internet of Things), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future-developed networks.

[0140] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device.

[0141] The aforementioned computer-readable medium carries one or more programs. When the electronic device executes the aforementioned one or more programs, the electronic device causes the following to occur: 1. Obtain a query request, the query request including first content and second content, the first content and the second content belonging to different content types, and the second content describing a query intent; 2. Determine the query category to which the query intent corresponding to the second content belongs through a first large model; wherein the query category includes a first query category and a second query category, the first query category representing the generation of a query result corresponding to the second content through the first content, and the second query category representing the generation of a query result corresponding to the second content through knowledge retrieval; 3. In response to the query category corresponding to the second content being the first query category, input the first content and the second content into a second large model to obtain a first query result corresponding to the query request output by the second large model; 4. In response to the query category corresponding to the second content being the second query category, input the first content and the second content into the first large model to obtain a third content output by the first large model; wherein the first large model is used to adjust the second content based on information identified from the first content to obtain the third content, and the content type of the third content is consistent with the content type of the second content; 5. Obtain a second query result corresponding to the query request based on the third content.

[0142] Computer program code for performing the operations of this disclosure can be written in one or more programming languages ​​or a combination thereof, including but not limited to object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0143] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0144] The modules described in the embodiments of this disclosure can be implemented in software or hardware. The names of the modules are not, in some cases, intended to limit the functionality of the module itself.

[0145] The functions described above in this document can be performed at least in part by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoCs), complex programmable logic devices (CPLDs), and so on.

[0146] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0147] The above description is merely a preferred embodiment of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features disclosed in this disclosure that have similar functions.

[0148] Furthermore, while the operations are described in a specific order, this should not be construed as requiring these operations to be performed in the specific order shown or in a sequential order. In certain environments, multitasking and parallel processing may be advantageous. Similarly, while several specific implementation details are included in the above discussion, these should not be construed as limiting the scope of this disclosure. Certain features described in the context of individual embodiments may also be implemented in combination in a single embodiment. Conversely, various features described in the context of a single embodiment may also be implemented individually or in any suitable sub-combination in multiple embodiments.

[0149] Although the subject matter has been described using language specific to structural features and / or methodological logic, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. Rather, the specific features and actions described above are merely illustrative examples of implementing the claims. Regarding the apparatus in the above embodiments, the specific manner in which the various modules perform their operations has been described in detail in the embodiments relating to the method, and will not be elaborated upon here.

Claims

1. A knowledge question answering method based on a large model, characterized in that, include: Obtain a query request, the query request including first content and second content, the first content and the second content belong to different content types, and the second content is used to describe the query intent; The first major model is used to determine the query category to which the query intent corresponding to the second content belongs; wherein, the query category includes a first query category and a second query category, the first query category represents the generation of query results corresponding to the second content through the first content, and the second query category represents the generation of query results corresponding to the second content through knowledge retrieval; In response to the query category corresponding to the second content being the first query category, the first content and the second content are input into the second large model to obtain the first query result corresponding to the query request output by the second large model; In response to the query category corresponding to the second content being the second query category, the first content and the second content are input into the first large model to obtain the third content output by the first large model; wherein, the first large model is used to adjust the second content according to the information identified from the first content to obtain the third content, and the content type of the third content is consistent with the content type of the second content; Based on the third content, the second query result corresponding to the query request is obtained.

2. The method according to claim 1, characterized in that, The process of determining the query category to which the query intent corresponding to the second content belongs through the first major model includes: The first major model is used to determine the query intent corresponding to the second content; In response to the query intent being a search type, the query category to which the second content belongs is determined to be the second query category; In response to the query intent being a non-search type intent, the query category to which the second content belongs is determined to be the first query category.

3. The method according to claim 2, characterized in that, The step of determining the query category to which the second content belongs as the second query category in response to the query intent being a search type includes: In response to the query intent being a search type, the query object targeted by the query intent corresponding to the second content is determined in the first content based on the second content; In response to the query object containing non-textual information that cannot be converted into text, the query category to which the second content belongs is determined to be the first query category.

4. The method according to any one of claims 1-3, characterized in that, The step of obtaining the second query result corresponding to the query request based on the third content includes: Verify whether the third content conforms to preset rules; In response to the third content conforming to preset rules, a second query result corresponding to the query request is obtained based on the third content; In response to the fact that the third content does not conform to the preset rule, the first content and the second content are input into the second large model to obtain the third query result corresponding to the query request output by the second large model.

5. The method according to any one of claims 1-3, characterized in that, The second major model was determined through the following steps: In response to the selection operation of the target large model among multiple candidate large models, the selected target large model is determined as the second large model.

6. The method according to any one of claims 1-3, characterized in that, The process of determining the query category to which the query intent corresponding to the second content belongs through the first major model includes: Determine the request length corresponding to the query request; If the request length is less than or equal to a preset threshold, the query category to which the query intent corresponding to the second content belongs is determined through the first major model. If the request length is greater than the preset threshold, the first content and the second content are input into the third model to obtain the fourth content output by the third model; wherein, the third model is used to adjust the second content according to the information identified from the first content to obtain the fourth content, and the content type of the fourth content is consistent with the content type of the second content; Based on the fourth content, the fourth query result corresponding to the query request is obtained.

7. A knowledge question-answering device based on a large model, characterized in that, include: The acquisition module is used to acquire a query request, which includes first content and second content. The first content and the second content belong to different content types, and the second content is used to describe the query intent. The first determining module is used to determine the query category to which the query intent corresponding to the second content belongs through the first large model; wherein, the query category includes a first query category and a second query category, the first query category represents the query result corresponding to the second content generated through the first content, and the second query category represents the query result corresponding to the second content generated through knowledge retrieval; The first response module is used to respond to the query category corresponding to the second content being the first query category, input the first content and the second content into the second large model, and obtain the first query result corresponding to the query request output by the second large model; The second response module is used to respond to the query category corresponding to the second content being the second query category, inputting the first content and the second content into the first large model to obtain the third content output by the first large model; wherein, the first large model is used to adjust the second content according to the information identified from the first content to obtain the third content, and the content type of the third content is consistent with the content type of the second content; The second determining module is used to obtain the second query result corresponding to the query request based on the third content.

8. A computer-readable medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processing device, it implements the method of any one of claims 1-6.

9. An electronic device, characterized in that, include: A storage device on which computer programs are stored; A processing device for executing the computer program in the storage device to implement the method of any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1-6.

Images