Man-machine conversation method and device, storage medium and program product

Abstract

The invention provides a man-machine conversation method and device, a storage medium and a program product, and the method comprises the steps: determining a knowledge point index corresponding to each target professional knowledge point under the condition that an input problem is detected to be matched with at least one target professional knowledge point; constructing a large model cue word according to each knowledge point index and the input question; the large model cue words are input into the large language model to determine corresponding large model output content, and the large model output content comprises all knowledge point indexes; and calling corresponding target professional knowledge points based on the knowledge point indexes, and constructing reply answers according to the target professional knowledge points and the model output answers. Therefore, the powerful natural language understanding ability of the large language model is utilized, and more accurate and professional answers can be provided for users by quoting matched professional knowledge points which are related to the questions and are in the field of professional knowledge in the reply answers.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a human-computer dialogue method, device, storage medium and program product. Background Art

[0002] With the rapid development of artificial intelligence technology, intelligent dialogue systems have been widely used in many fields, such as customer service, intelligent assistants, education and training, etc. At present, large language models have made remarkable achievements in the field of natural language processing. Some human-computer interaction products have greatly improved the intelligence level of human-computer interaction products by integrating large language models.

[0003] Although the dialogue system based on the big model can effectively support a variety of user questions, the big model may not be accurate enough in applying professional knowledge in specific fields, and there are also limitations in knowledge customization, making it difficult to meet the needs of professional human-computer interaction in human-computer dialogues such as operating instructions based on home appliances or car computers, or other specific scenarios.

[0004] The industry has not yet proposed a better solution to the above problems. Summary of the invention

[0005] The present application provides a human-computer dialogue method, device, storage medium and program product, which are used to at least solve the problem that the response content of the existing large model-based dialogue system cannot meet the needs of professional question-and-answer guidance.

[0006] In a first aspect, an embodiment of the present application provides a human-computer dialogue method, comprising: when it is detected that an input question matches at least one target professional knowledge point, determining the knowledge point indexes corresponding to each of the target professional knowledge points; constructing a large model prompt word according to each of the knowledge point indexes and the input question; inputting the large model prompt word into a large language model to determine the corresponding large model output content, wherein the large model output content includes each of the knowledge point indexes; calling the corresponding target professional knowledge point based on each of the knowledge point indexes, and constructing a reply answer according to each of the target professional knowledge points and the model output answer.

[0007] In a second aspect, an embodiment of the present application provides an electronic device, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can perform the steps of the human-computer dialogue method of any embodiment of the present application.

[0008] In a third aspect, an embodiment of the present application provides a storage medium on which a computer program is stored, characterized in that when the program is executed by a processor, the steps of the human-computer dialogue method of any embodiment of the present application are implemented.

[0009] In a fourth aspect, an embodiment of the present application provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of the human-computer dialogue method of any embodiment of the present application.

[0010] The beneficial effects of the embodiments of the present application are: By detecting whether the input question involves at least one target professional knowledge point, it can intelligently identify whether the user's question belongs to the field of professional knowledge. If it involves professional knowledge points, the relevant knowledge point index is designed into the large model prompt words, so that the reply answer can not only include the large model's reading comprehension answer to the user's question, but also reference the target professional knowledge point based on the knowledge point index. In this way, it not only utilizes the powerful natural language understanding ability of the large language model, but also can provide users with more accurate and professional answers by citing matching professional knowledge points in the field of professional knowledge related to the question in the reply answer. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0012] Figure 1 An operation flow chart of an example of a human-computer dialogue method according to an embodiment of the present application is shown; Figure 2 A schematic diagram of an operation flow of an example of realizing multi-modal human-computer dialogue through a plug-in in the current related technology is shown; Figure 3 An operation flow chart of an example of constructing a reply answer based on target professional knowledge points and model output answers according to an embodiment of the present application is shown; Figure 4 An operation flow chart of an example of a human-computer dialogue method according to an embodiment of the present application is shown; Figure 5 It is a schematic structural diagram of an embodiment of an electronic device of the present application. DETAILED DESCRIPTION

[0013] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, 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 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 are within the scope of protection of this application.

[0014] Figure 1 An operation flow chart of an example of a human-computer dialogue method according to an embodiment of the present application is shown.

[0015] like Figure 1 As shown, in step S110, when it is detected that the input question matches at least one target professional knowledge point, the knowledge point indexes corresponding to the respective target professional knowledge points are determined.

[0016] In some implementations, NLP (Natural Language Processing) technology is used to perform a preliminary analysis of the user input question, such as word segmentation, part-of-speech tagging, entity recognition, etc., to identify key information and core elements in the question. For example, in a conversation in the field of home appliances, if the user mentions "the washing machine does not work", the system will identify "washing machine" and "does not work" as key entities and status information.

[0017] Then, by using the pre-established professional knowledge base, the target professional knowledge points related to the input question are gradually retrieved and matched. The knowledge base contains professional terms, solutions, operating procedures, common problems, etc. in various fields, which are constructed into a series of knowledge point entries. It should be understood that these knowledge points in the professional knowledge base can be constructed through manual annotation or automatic learning (such as unsupervised learning based on domain text), which is not limited here.

[0018] Here, a variety of semantic matching algorithms (such as TF-IDF, vector distance matching, etc.) can be used to calculate the vector matching degree between each professional knowledge point and the input question to screen the target professional knowledge points with higher corresponding vector matching degrees. For example, in order for the large model to obtain more concise background knowledge, the complete background knowledge document can be segmented and vectorized in advance and stored in the vector database. When the user asks a question, the user's question is searched in the vector database, and the top-ranked paragraph fragments are selected as the target professional knowledge points. Then, the knowledge point index of each professional knowledge point is retrieved, such as the ID automatically generated by the system or the knowledge point in the library, which can uniquely identify and quickly locate the specific content of the knowledge point.

[0019] In step S120, a large model prompt word is constructed according to each knowledge point index and the input question.

[0020] In an example of an embodiment of the present application, each knowledge point index and input question are filled into the corresponding semantic slot of the large model prompt word template to construct a model prompt word (Prompt). Specifically, the large model prompt word template predefines that the large model output content includes a large model understanding answer part and a professional knowledge point reference part, and the professional knowledge point reference part is used to fill each knowledge point index.

[0021] Here, the format of the model's output content is constrained through the large model prompt template, so that on the one hand, it can give full play to the powerful language understanding ability of the large language model and give answers after understanding the semantics of the input question. It can also provide a channel for users to obtain professional guidance by calling relevant knowledge point indexes.

[0022] In another example of the embodiment of the present application, in addition to the reserved semantic slots for knowledge point index and input questions, the large model prompt word template also reserves semantic slots for various professional knowledge points, so that the large language model can combine various professional knowledge points and input questions to determine the output content of the large model. In this way, when the large language model performs reading comprehension for input questions, it can integrate and consider the influence of various professional knowledge points, thereby improving the accuracy and professionalism of the large model's understanding of the answer part.

[0023] In step S130, the large model prompt word is input into the large language model to determine the corresponding large model output content, wherein the large model output content includes various knowledge point indexes.

[0024] It should be understood that the types of large models can be diverse, such as Qwen or GPT series, etc., which are not limited here. Here, the output of the large language model is not only the natural language answer to the user's question, but also includes the part related to each professional knowledge point index. As described above, the output content of the large model is constrained by the prompt template, so that the output content can contain relevant professional knowledge items, which can improve the auxiliary effect of the output content in the professional field.

[0025] In step S140, the corresponding target professional knowledge points are called based on the indexes of each knowledge point, and a reply answer is constructed according to each target professional knowledge point and the model output answer.

[0026] In some embodiments, the knowledge point index is in a specific data format. The system detects the knowledge point index in the output content of the large language model through this data format, and retrieves the corresponding target professional knowledge point from the professional knowledge base based on the knowledge point index. Then, each target professional knowledge point is integrated with the model output answer to obtain a reply answer to provide feedback to the user. Accordingly, in the reply answer, the user can not only obtain the answer content output by the large language model through reading and understanding the input question, but also obtain relevant professional knowledge points in the professional knowledge field related to the question.

[0027] In another example of an embodiment of the present application, when it is detected that there is no matching target professional knowledge point for the input question, a large model prompt word is constructed according to the input question.

[0028] Specifically, if the input question does not involve any specific professional knowledge (such as "What's the weather like today?"), the large model prompt words are directly constructed based on the input question. Through the natural language understanding ability of the large language model, the dialogue system can still handle a variety of question types, such as common sense questions, small talk, weather queries, etc., without worrying about whether the user's question belongs to a specific professional field, and has strong flexibility.

[0029] Regarding the details of the knowledge point matching operation in step S110, in some embodiments, text matching is performed based on the input question and a preset background knowledge document to determine whether there is at least one matching target background knowledge, and the background knowledge document stores at least one background knowledge. Further, when it is determined that there is at least one matching target background knowledge, each target background knowledge is determined as a corresponding target professional knowledge point.

[0030] It should be understood that background knowledge documents are core resources for storing target professional knowledge points, and the business scenarios supported by the documents can be diverse, such as home appliance maintenance instructions, vehicle computer instructions, etc., which should not be limited here. In addition, each background knowledge entry can be identified by a unique ID and can also be equipped with corresponding keyword tags, which is conducive to supporting more efficient and accurate knowledge matching operations.

[0031] Specifically, when the input question has a high degree of match with certain knowledge points in the background knowledge document (for example, exceeding a preset threshold), these matching background knowledge entries are confirmed and marked as target background knowledge points (or target professional knowledge points), providing a basis for the subsequent generation of response answers with professional information.

[0032] In some examples of the embodiments of the present application, the background knowledge document may be maintained or updated by the user to meet the user's personalized customization needs for specialized human-computer interaction scenarios. Specifically, the document uploaded by the user is obtained, and the background knowledge document is constructed or updated based on the document uploaded by the user.

[0033] In some embodiments, by providing an upload data interface, users are supported to upload documents in various formats (such as PDF, Word, Excel, etc.), and NLP technology is used to automatically extract potential target background knowledge points from the uploaded documents and store them in a structured manner in the background knowledge documents. As a result, users can upload, maintain and update background knowledge documents according to their needs, ensuring that the background knowledge base of the dialogue system can keep up with changes in business needs. In some industry-specific or very detailed areas, such as product instructions, internal corporate system inquiries, etc., the customized documents uploaded by users enable the dialogue system to provide highly relevant professional knowledge and answers, thereby improving the professionalism and personalization of dialogue services.

[0034] It should be noted that the large language model has a strong natural language understanding ability, but its ability to understand the content of media materials is weak, especially in some professional interaction scenarios. The generated pictures or videos of the large language model lack constraints and cannot meet the interactive goal of providing professional guidance to users using specific media materials. For example, in some home appliance and car manual question and answer scenarios, customers expect to be able to achieve multi-modal responses of text, pictures, and videos.

[0035] At present, large language models can better output text answers when faced with text questions, which is determined by the specifications of the large language models themselves. Since the technical framework and data used in the pre-training of large language models are all text-related, multi-modal picture and text replies cannot be achieved in similar technologies, which is a long-standing problem of large language models.

[0036] In order to achieve the above goals, some experts and scholars in the industry have proposed to generate images and videos by using multimodal large models or plug-ins. For example, by adopting multimodal large models or using plug-ins, the demand can be met in theory. However, the technical threshold and cost of multimodal large models are very high, and their output content lacks constraints and may be difficult to meet the requirements. For most practitioners, it is impossible to achieve this, so using plug-ins to achieve multimodality has become a common solution for practitioners.

[0037] Figure 2 The present invention shows an operation flow diagram of an example of realizing multi-modal human-computer dialogue through plug-ins in the current related technology.

[0038] like Figure 2As shown in the figure, by calling the plug-in to support multimodal responses in the interaction process of the large language model, however, because calling the plug-in is a preset step, each round of question and answer will generate pictures and videos, which is costly and unnecessary; and each round adds a process of calling the plug-in, which greatly increases the time consumption. In addition, the pictures and videos generated by the text-generated picture and text-generated video models cannot effectively meet the rigorous effect of text-image consistency due to the uncertainty of model attributes and in some professional interaction scenarios (for example, home appliances and car machine manuals).

[0039] It should be noted that the purpose of the above description of the current related technologies is only to enable the public to better understand the inventive spirit of the present application, and is not intended to limit the scope of the present invention, and these current related technologies may be undisclosed technologies, such as technologies in the experimental stage, etc.

[0040] In view of this, Figure 3 An operational flowchart of an example of constructing a reply answer based on target professional knowledge points and model output answers according to an embodiment of the present application is shown.

[0041] like Figure 3 As shown, in step S310, it is detected whether there is a matching associated media material for each target professional knowledge point.

[0042] In some implementations, the professional knowledge points in the professional knowledge base are pre-matched with corresponding media materials, such as pictures and videos associated with the knowledge points. It should be understood that the pre-matching relationship between the media materials and the knowledge points can be defined in a variety of ways, such as by specifying through interactive operations or automatically matching through semantic analysis, which is not limited here.

[0043] In terms of business scenarios, in the maintenance guidance of home appliances, specific troubleshooting steps may be associated with related images, charts, or operation videos. Or, in the vehicle manual scenario, specific operation procedures may correspond to schematic diagrams or video tutorials. Therefore, if a certain professional knowledge point is associated with related images, videos, and other materials, the related call information of the associated materials can be recorded.

[0044] In step S320, when it is determined that there are matching associated media materials, a reply answer is constructed according to each target professional knowledge point, each associated media material and the model output answer.

[0045] On the other hand, when it is determined that there is no matching associated media material, a reply answer may be constructed based on various target professional knowledge points and the model output answer.

[0046] Once it is determined that there are matching associated media materials for the target professional knowledge points of the input question, the associated media materials can be integrated with the text answers generated by the large language model and the text content of the professional knowledge points to form a multimodal response answer.

[0047] For example, when answering questions about home appliance repair, the system can generate a text answer and attach a diagram of the operating steps, or if it is a complex operating process, a video link can also be attached.

[0048] It should be understood that the multimodal output format can be diverse. For example, according to different types of materials (images, videos, text, etc.), the system will use different formats for output, which is not limited here. For example, images can be directly embedded in the reply content, videos can provide a play link or an embedded play box, and the text part is returned in a conventional text format.

[0049] Through the embodiments of the present application, by combining text and related multimodal materials such as images and videos, users can be provided with more comprehensive and easy-to-understand answers. In particular, in some professional scenarios that require operation steps or visual assistance (such as home appliance repair, car machine operation, etc.), this multimodal approach greatly improves the user experience.

[0050] It should be noted that each associated media material is matched based on professional knowledge points, which can effectively avoid the situation where the multimodal large model lacks constraints and the output media material is irrelevant to the question, making the reply have more rigorous media materials, and also avoiding the external plug-in generating materials for each round of question and answer through plug-in calls, which can effectively improve the response efficiency of the system. As a result, it not only optimizes the generation process of multimodal replies, reduces unnecessary resource waste and processing time, but also greatly improves the accuracy, professionalism and user experience of the answers.

[0051] In some examples of the embodiments of the present application, the user can independently choose to match the corresponding media material for each professional knowledge point. Specifically, when a material matching instruction for a first professional knowledge point is detected, the first associated media material corresponding to the material matching instruction is obtained, and a matching relationship is established between the first associated media material and the first professional knowledge point.

[0052] More specifically, users can match corresponding media materials for professional knowledge points according to business needs. For example, users can trigger material matching instructions through material-knowledge point pairing operations, and build matching relationships between various knowledge points in background knowledge documents or professional knowledge bases and their desired associated media materials. For example, users can match associated media materials for a certain home appliance maintenance professional knowledge point through operations, so that when designing the professional knowledge point through input questions asked by (the same or different) users, the corresponding associated media materials can be quickly called.

[0053] Through the embodiments of the present application, the dialogue system can respond to user needs and match professional knowledge points with related materials (such as pictures, videos, audio, etc.), so that in the process of human-computer dialogue, users can not only obtain text replies, but also have an in-depth understanding of professional knowledge through auxiliary materials such as images and videos. Especially in scenarios that require operational guidance, complex processes or troubleshooting, it can provide richer and more vivid answers in multimodal interactions, thereby improving the user immersion in system professional knowledge field questions and answers.

[0054] Figure 4 An operation flow chart of an example of a human-computer dialogue method according to an embodiment of the present application is shown.

[0055] like Figure 4 As shown in FIG. 1 , in the multimodal reply method for home appliance and vehicle manual question and answer based on a large language model, the following operations are specifically included: 1) Upload the instruction manual.

[0056] The document management system supports uploading and parsing documents in various formats such as pdf, docx, pptx, etc. for home appliances or car manuals.

[0057] 2) Segmentation and vectorization The document management system will automatically segment, number and vectorize the parsed documents and save them persistently.

[0058] 3) Upload relevant pictures or videos for knowledge point segments Users need to actually check the manual of home appliances or car computers, and store and save the pictures or video links that introduce the relevant functions in the manual, and then find the corresponding knowledge point segment that introduces the function in the document system, upload the pictures and videos, and associate the knowledge point segment. For example, in the manual of the car computer, extract a paragraph of instructions on how to open the hood to determine the corresponding knowledge point segment, and then the user marks the associated video or picture of opening the hood to the knowledge point segment.

[0059] The preparation of the document system is now complete. If there are any problems during the Q&A process, you can return to the document management system for further tuning. Therefore, document upload management is more flexible and convenient for associating vectorized text with pictures and videos.

[0060] 4) User questions and vector retrieval When a user asks a question, the user's question is searched in the document vector database, and several qualified background knowledge fragments are screened and recalled. Each background knowledge fragment has an independent number.

[0061] Specifically, the angle between the user question and the target vector in the vector database is calculated. The smaller the angle, the greater the correlation. The corresponding background knowledge is then referenced or distinguished by the background knowledge number.

[0062] 5) Large Model Reading Comprehension Several recalled background knowledge fragments are numbered and filled into the placeholder of the prompt word, and the prompt word requires the large model to include the cited background knowledge number when outputting the response.

[0063] The following is an example of a prompt word template: "If your answer comes from background knowledge, your answer should cite the context in the corresponding part of your answer in the format of reference number [data:x]. If a sentence comes from multiple contexts, please list all relevant reference numbers, such as [data:3][data:5].

[0064] Here is an example response: Background knowledge: [data:001] The rearview mirrors of a car are on the left and right sides of the front row of the car.

[0065] Question: Where are the rearview mirrors in a car? Answer: The rearview mirrors of a car are on the left and right sides of the front row of the car. [data:001].

[0066] Also, if no relevant content is found in the background knowledge, do not output the reference number. " In this way, by utilizing the large model reasoning generation capability, the referenced background knowledge fragment numbers are output in an agreed format to facilitate subsequent indexing.

[0067] 6) Determine whether the cited source is related to pictures or videos After the big model outputs the reply content, the background knowledge fragments in the reply are numbered and indexed in the document system. If pictures or videos are associated, they will be displayed in the reply results.

[0068] Therefore, pictures and videos will only be displayed if the user associates them with the knowledge point fragment. There is no need to display them in every round. The displayed content is completely defined by the user, which is more rigorous and reliable. There is no need to call additional plug-ins, and no extra time is added.

[0069] It should be noted that in the current related technologies, some large models can output the reference source while replying, but they are mainly attached search engine links and accompanying content. However, in the field of question and answer based on business instructions (for example, home appliances, car computers, etc.), the business requires relatively accurate output of relevant pictures, so searching for pictures through search engines or generating pictures through multimodal large models cannot achieve the expected effect.

[0070] Through the embodiment of the present application, first, the background knowledge document is uploaded to the platform for segment numbering and vectorization, and the segmented knowledge point fragments can be viewed on the platform. Then, by manually taking screenshots and annotating, the drawings and videos related to the knowledge point fragments in the manual are uploaded and associated with the corresponding fragments. When arranging the prompt words, the large model is required to output the knowledge point fragment number referenced by the answer question. Furthermore, during the human-computer dialogue, the pictures or videos associated with it are displayed according to the fragment number index output by the large model.

[0071] For example, in the dialogue scenario of business manuals, when users associate these manuals as background knowledge, they can refer to these background knowledge when responding. However, when the recalled background knowledge fragments are all below the set threshold (for example, vector matching will have a score, which represents the relevance to the user's question. If it is not relevant, it is often below the threshold), these background knowledge can be discarded, so that the dialogue system can also respond to the user's questions that are not related to background knowledge as an ordinary large model, and there is no need to provide pictures at this time.

[0072] As a result, the dialogue system can provide pictures and videos on demand, and the pictures and videos are manually annotated, so they are more accurate. If there is no associated video or picture for the knowledge segment corresponding to the user's question, the picture or video will not be called, and the reply will only be made when the picture or video is needed. In addition, the big model capability is used to output the background knowledge reference source and index the related pictures and videos. The pictures and videos replied must be relevant and rigorous to the content, and should not increase the link time unless necessary.

[0073] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity of description, they are all expressed as a series of actions combined, 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. In the above embodiments, 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.

[0074] In some embodiments, an embodiment of the present application provides a non-volatile computer-readable storage medium, in which one or more programs including execution instructions are stored. The execution instructions can be read and executed by an electronic device (including but not limited to a computer, a server, or a network device, etc.) to execute any of the above-mentioned human-computer interaction methods of the present application.

[0075] In some embodiments, the embodiments of the present application also provide a computer program product, which includes a computer program stored on a non-volatile computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer executes any one of the above-mentioned human-computer interaction methods.

[0076] In some embodiments, an embodiment of the present application also provides an electronic device, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute a human-computer dialogue method.

[0077] Figure 5 is a schematic diagram of the hardware structure of an electronic device for executing a human-computer dialogue method provided by another embodiment of the present application, such as Figure 5 As shown, the device includes: One or more processors 510 and memory 520, Figure 5 A processor 510 is taken as an example.

[0078] The device for executing the human-machine dialogue method may further include: an input device 530 and an output device 540 .

[0079] The processor 510, the memory 520, the input device 530 and the output device 540 may be connected via a bus or other means. Figure 5 The example of connecting through bus is taken in the following.

[0080] The memory 520 is a non-volatile computer-readable storage medium that can be used to store non-volatile software programs, non-volatile computer executable programs and modules, such as program instructions / modules corresponding to the human-computer dialogue method in the embodiment of the present application. The processor 510 executes various functional applications and data processing of the server by running the non-volatile software programs, instructions and modules stored in the memory 520, that is, realizing the human-computer dialogue method in the above method embodiment.

[0081] The memory 520 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the electronic device, etc. In addition, the memory 520 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory 520 may optionally include a memory remotely arranged relative to the processor 510, and these remote memories may be connected to the electronic device 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.

[0082] The input device 530 may receive input digital or character information and generate signals related to user settings and function control of the electronic device. The output device 540 may include a display device such as a display screen.

[0083] The one or more modules are stored in the memory 520, and when executed by the one or more processors 510, the human-computer dialogue method in any of the above method embodiments is executed.

[0084] The above-mentioned product can execute the method provided in the embodiment of the present application, and has the functional modules and beneficial effects corresponding to the execution method. For technical details not fully described in this embodiment, please refer to the method provided in the embodiment of the present application.

[0085] The electronic devices of the embodiments of the present application exist in various forms, including but not limited to: (1) Mobile communication equipment: This type of equipment is characterized by having mobile communication functions and its main purpose is to provide voice and data communications. This type of terminal includes: smart phones, multimedia phones, functional phones, and low-end phones.

[0086] (2) Ultra-mobile personal computer devices: These devices belong to the category of personal computers, have computing and processing functions, and generally also have mobile Internet access features. These terminals include: PDA, MID and UMPC devices, etc.

[0087] (3) Portable entertainment devices: These devices can display and play multimedia content. They include audio and video players, handheld game consoles, e-books, smart toys, and portable car navigation devices.

[0088] (4) Other onboard electronic devices with data interaction functions, such as on-board devices installed in vehicles.

[0089] The device embodiments described above are merely illustrative, wherein 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 modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0090] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a general hardware platform, and of course, by hardware. Based on this understanding, the above technical solution, in essence, or the part that contributes to the relevant technology, can be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiment.

[0091] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A human-computer dialogue method, comprising: In the case where it is detected that the input question matches at least one target professional knowledge point, determining the knowledge point indexes corresponding to each of the target professional knowledge points; Constructing a large model prompt word according to each of the knowledge point indexes and the input question; Inputting the large model prompt words into a large language model to determine corresponding large model output content, wherein the large model output content includes each of the knowledge point indexes; Based on each of the knowledge point indexes, the corresponding target professional knowledge points are called, and a reply answer is constructed according to each of the target professional knowledge points and the model output answer.

2. The method according to claim 1, wherein: The step of constructing a reply answer according to each of the target professional knowledge points and the model output answer includes: Detecting whether there is any matching associated media material for each of the target professional knowledge points; In the case where it is determined that there are matching associated media materials, a reply answer is constructed according to each of the target professional knowledge points, each of the associated media materials and the model output answer.

3. The method according to claim 2, further comprising: When a material matching instruction for a first professional knowledge point is detected, a first associated media material corresponding to the material matching instruction is acquired, and a matching relationship is established between the first associated media material and the first professional knowledge point.

4. The method according to claim 1, further comprising: Performing text matching based on the input question and the preset background knowledge document to determine whether there is at least one matching target background knowledge; The background knowledge document stores at least one background knowledge; When it is determined that there is at least one matching target background knowledge, each of the target background knowledge is determined as a corresponding target professional knowledge point.

5. The method according to claim 4, further comprising: The document uploaded by the user is obtained, and the background knowledge document is constructed or updated based on the document uploaded by the user.

6. The method according to claim 1, wherein: The step of constructing a large model prompt word according to each of the knowledge point indexes and the input question includes: Fill each of the knowledge point indexes and the input question into the corresponding semantic slot of the large model prompt word template to construct a model prompt word; wherein the large model prompt word template predefines the large model output content to include a large model understanding answer part and a professional knowledge point reference part, and the professional knowledge point reference part is used to fill each of the knowledge point indexes.

7. The method according to claim 6, wherein: The large model prompt word template also reserves semantic slots for each of the professional knowledge points, so that the large language model can comprehensively consider each of the professional knowledge points and the input questions to determine the output content of the large model.

8. The method according to any one of claims 1 to 7, further comprising: When it is detected that there is no matching target professional knowledge point for the input question, a large model prompt word is constructed according to the input question.

9. A storage medium having a computer program stored thereon, wherein: When the program is executed by a processor, the steps of the method described in any one of claims 1 to 8 are implemented.

10. An electronic device, comprising: At least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the steps of the method described in any one of claims 1 to 8.

Images