Intellectual property and academic assistant system and implementation method and device thereof, electronic equipment and storage medium

Abstract

The invention provides an intellectual property and academic assistant system, an implementation method and device thereof, electronic equipment and a storage medium, and relates to the technical field of artificial intelligence. The method comprises the steps that question information is obtained, the question information is input into an intention classification model for intention classification, an intention classification result indicating question categories of the question information is obtained, and the question categories of the question information comprise an intellectual property question and answer category, a patent retrieval category and an academic question and answer category; if the question category of the question information is a patent retrieval category, inputting the question information into a patent retrieval model to generate a patent retrieval result; if the question category of the question information is an intellectual property question and answer category, inputting the question information into an intellectual property large model to generate an intellectual property answer result; and if the question category of the question information is an academic question and answer category, inputting the question information into an academic question and answer model to generate an academic question and answer result. The problem of poor efficiency of an intellectual property and academic assistant system in related technologies is solved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology. Specifically, the present application relates to an intellectual property and academic assistant system and its implementation method, device, electronic device and storage medium. Background Art

[0002] In the field of intellectual property and academic research, with the explosive growth of information, researchers and professionals are facing more and more information retrieval and management challenges. Traditional information retrieval and academic document management systems are often unable to cope with large-scale data and complex query requirements.

[0003] For example, in patent retrieval, researchers need to find relevant information from a large amount of patent documents. Traditional retrieval methods are usually based on keyword matching, which easily produces a large number of irrelevant results. In academic research, researchers need to find and organize a large amount of literature. Existing literature organization methods are too inefficient for organizing a large amount of patent documents.

[0004] From the above, we can see that the issue of how to improve the efficiency of the intellectual property and academic assistant systems remains to be resolved. Summary of the invention

[0005] The present application provides an intellectual property and academic assistant system and its implementation method, device, electronic device and storage medium, which can solve the problem of low efficiency of intellectual property and academic assistant systems in related technologies. The technical solution is as follows:

[0006] According to one aspect of the present application, an intellectual property and academic assistant system and an implementation method thereof are characterized by comprising:

[0007] Acquire question information, and input the question information into an intent classification model for intent classification, to obtain an intent classification result indicating a question category of the question information, wherein the question category of the question information includes an intellectual property question and answer category, a patent search category, and an academic question and answer category;

[0008] If the intent classification result indicates that the question category of the question information is a patent search category, inputting the question information into a patent search model to generate a patent search result;

[0009] If the intent classification result indicates that the question category of the question information is an intellectual property question and answer category, inputting the question information into the intellectual property big model to generate an intellectual property answer result;

[0010] If the intention classification result indicates that the question category of the question information is an academic question and answer category, the question information is input into the academic question and answer model to generate an academic question and answer result.

[0011] According to one aspect of the present application, an intellectual property and academic assistant system device is characterized by comprising:

[0012] An intent classification module is used to obtain question information, and input the question information into an intent classification model for intent classification, to obtain an intent classification result indicating a question category of the question information, wherein the question category of the question information includes an intellectual property question and answer category, a patent search category, and an academic question and answer category;

[0013] A patent search module, for inputting the question information into a patent search model to generate a patent search result if the intent classification result indicates that the question category of the question information is a patent search category;

[0014] an intellectual property question-answering module, configured to input the question information into an intellectual property big model to generate an intellectual property answer result if the intent classification result indicates that the question category of the question information is an intellectual property question-answering category;

[0015] The academic question-answering module is used to input the question information into the academic question-answering model to generate an academic question-answering result if the intention classification result indicates that the question category of the question information is an academic question-answering category.

[0016] In an exemplary embodiment, the patent search model includes a full-text search engine unit and a re-ranking unit;

[0017] The patent search module includes:

[0018] A keyword extraction unit, used to extract keywords from the question information and generate search term information;

[0019] A relevance search unit, used for inputting the search term information into a full-text search engine unit for relevance search, and obtaining a predetermined first number of search results;

[0020] A reordering unit is used to reorder the search results based on the reordering unit to obtain patent search results corresponding to the question information.

[0021] In an exemplary embodiment, the academic question answering model includes a knowledge graph unit and a large language model unit;

[0022] The academic question-and-answer module includes:

[0023] A knowledge graph retrieval unit, used to input the question information into the knowledge graph unit for knowledge graph retrieval, and determine background retrieval information corresponding to the question information;

[0024] The academic question and answer result generating unit is used to input the background search information and the question information into a preset large language model unit to obtain the academic question and answer result corresponding to the question information.

[0025] In an exemplary embodiment, the apparatus further comprises:

[0026] Data acquisition module, used to obtain patent academic data;

[0027] An entity management processing module, used to perform entity association processing on the patent academic data to obtain key information entities and entity relationships between the key information entities, wherein the entity association processing includes word segmentation, named entity recognition and relationship extraction;

[0028] A storage module is used to store the key information entities and the corresponding entity relationships in a graph database to obtain a knowledge graph unit.

[0029] In an exemplary embodiment, the academic question-answering result generating unit includes:

[0030] A multi-dimensional analysis subunit, used to perform multi-dimensional analysis on key information entities in the background search information based on the large language model unit, and generate user portrait information corresponding to the key information entities;

[0031] The academic question and answer result generation subunit is used to generate academic question and answer results based on the user portrait information corresponding to each key information entity.

[0032] In an exemplary embodiment, the apparatus further comprises:

[0033] A first training module is used to obtain a first training set, and train a preset classification pre-training model based on the first training set to obtain a classification prediction result generated by the classification pre-training model, wherein the first training set includes intellectual property question and answer data corresponding to the intellectual property question and answer category, patent search data corresponding to the patent search category, and patent academic data corresponding to the academic question and answer category;

[0034] A performance evaluation module, configured to perform a model performance evaluation on the classification prediction result based on a loss function, and obtain an evaluation result indicating the prediction efficiency of the classification pre-training model for the data types of the first training set;

[0035] The model acquisition module is used to adjust the parameters of the pre-trained model based on the evaluation results and repeat the training until the prediction accuracy of the classification pre-trained model reaches a preset threshold, thereby obtaining an intent classification model.

[0036] In an exemplary embodiment, the apparatus further comprises:

[0037] A second training module, used to obtain a second training set, wherein the second training set includes intellectual property question and answer data and general question and answer data;

[0038] A parameter fine-tuning module is used to fine-tune the model parameters of the preset question-and-answer pre-training model based on the second training set to obtain a large intellectual property question-and-answer model.

[0039] According to one aspect of the present application, an electronic device includes at least one processor and at least one memory, wherein the memory stores computer-readable instructions; the computer-readable instructions are executed by one or more of the processors, so that the electronic device implements the intellectual property and academic assistant system and its implementation method as described above.

[0040] According to one aspect of the present application, a storage medium stores computer-readable instructions thereon, and the computer-readable instructions are executed by one or more processors to implement the intellectual property and academic assistant system and its implementation method as described above.

[0041] According to one aspect of the present application, a computer program product includes computer-readable instructions, which are stored in a storage medium. One or more processors of an electronic device read the computer-readable instructions from the storage medium, load and execute the computer-readable instructions, so that the electronic device implements the intellectual property and academic assistant system and its implementation method as described above.

[0042] The beneficial effects of the technical solution provided by this application are:

[0043] In the above technical solution, after intent classification is performed through the intent classification model, the question information is input into the corresponding model for processing. After the problem category of the question information is determined, the most appropriate large model is called for service retrieval according to the different fields to which the question information belongs, and finally the results that best meet user needs are obtained, thereby improving the pertinence of the intellectual property and academic assistant system and the efficiency of each model in processing question information, thereby effectively solving the problem of low efficiency of the intellectual property and academic assistant system existing in related technologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are only some embodiments of the present application, and those skilled in the art can obtain other drawings based on these drawings without creative work.

[0045] Figure 1 It is a schematic diagram of the implementation environment involved in this application;

[0046] Figure 2 is a flow chart of an intellectual property and academic assistant system and an implementation method thereof according to an exemplary embodiment;

[0047] Figure 3 yes Figure 2 A flowchart of step 230 in one embodiment corresponding to the embodiment;

[0048] Figure 4 yes Figure 2 A flowchart of an embodiment corresponding to step 270 in an embodiment;

[0049] Figure 5 is a flowchart of a process of constructing a knowledge graph according to an exemplary embodiment;

[0050] Figure 6 yes Figure 4 A flowchart of an embodiment corresponding to step 273 in an embodiment;

[0051] Figure 7 is a flowchart of a process of constructing an intent classification model according to an exemplary embodiment;

[0052] Figure 8 is a flow chart showing a process of constructing a large model of intellectual property question and answer according to an exemplary embodiment;

[0053] Fig. 9 It is a specific implementation diagram of an intellectual property and academic assistant system and its implementation method in an application scenario;

[0054] Fig.10 is a structural block diagram of an intellectual property and academic assistant system device according to an exemplary embodiment;

[0055] Fig.11 is a hardware structure diagram of a server according to an exemplary embodiment;

[0056] Fig.12 The present invention is a structural block diagram of an electronic device according to an exemplary embodiment. DETAILED DESCRIPTION

[0057] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, and cannot be interpreted as limiting the present application.

[0058] It will be understood by those skilled in the art that, unless expressly stated, the singular forms "one", "said", and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present disclosure refers to the presence of the features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. It should be understood that when we refer to an element as being "connected" or "coupled" to another element, it may be directly connected or coupled to the other element, or there may be intermediate elements. In addition, the "connection" or "coupling" used herein may include wireless connection or wireless coupling. The term "and / or" used herein includes all or any unit and all combinations of one or more associated listed items.

[0059] As mentioned above, existing intellectual property and academic assistant systems are usually based on keyword matching, which easily produces a large number of irrelevant results. In academic research, researchers need to find and organize a large amount of literature. The existing literature organization methods are too inefficient for organizing a large number of patent documents. As can be seen from the above, the related art still has the defect of low efficiency of intellectual property and academic assistant systems.

[0060] To this end, the intellectual property and academic assistant system and its implementation method provided in the present application can effectively improve the efficiency of the intellectual property and academic assistant system. Accordingly, the intellectual property and academic assistant system and its implementation method are suitable for the intellectual property and academic assistant system device, and the intellectual property and academic assistant system device can be deployed in an electronic device. The electronic device can be a computer device configured with a von Neumann architecture, for example, the computer device includes a desktop computer, a laptop computer, a server, etc.; the electronic device can also be an electronic device with a central control function, for example, the electronic device includes a gateway, etc.; the electronic device can also refer to a portable and mobile electronic device, for example, the electronic device includes a smart phone, a tablet computer, etc.

[0061] In order to make the objectives, technical solutions and advantages of the present application clearer, the implementation methods of the present application will be further described in detail below with reference to the accompanying drawings.

[0062] like Figure 1 As shown, Figure 1It is a system structure diagram of an embodiment of the intellectual property and academic assistant system of the present invention; the present invention provides an embodiment of the intellectual property and academic assistant system based on a large language model; the intellectual property and academic assistant system 100 includes an acquisition module 101, an intent classification model 102, a database 103 and a generative large language model 104; wherein, the acquisition module 101 is used to acquire question information; the intent classifier 102 is used to identify and classify the question information; the database 103 is used to store real intellectual property and academic assistant data, and according to the question information of the intellectual property and academic assistant related categories, generate patents corresponding to the question information; the generative large language model 104 is used to receive the classified question information and the legal provisions, and output the final answer information.

[0063] In this embodiment, the user of the acquisition module 101 can input question information through text or voice, and the acquired question information will be further identified and classified by the intent classifier 102. Specifically, during the identification and classification process, if the question information includes content related to patent search, the question information is classified into the patent search category; if the question information includes content related to intellectual property questions and answers, the question information is classified into the intellectual property questions and answers category; if the question information includes content related to academic questions and answers, the question information is classified into the academic questions and answers category.

[0064] Then, the question information of the patent search type is input into the generative large language model 104, and the patent related to the question information is searched and generated in the database 103. The question information of the general intellectual property question and answer type is directly input into the generative large language model 104, and then the generative large language model 104 will directly output the final answer information based on the question information or based on the question information and the generated patent information. The question information of the academic question and answer type is input into the generative large language model 104. The generative large language model 104 performs semantic understanding and information retrieval based on the question information, queries and associates relevant information in the database 103, and finally generates analysis results related to the question information. The problem of poor efficiency of intellectual property and academic assistant systems existing in the relevant technology is solved.

[0065] See also Figure 2 The embodiment of the present application provides an intellectual property and academic assistant system and an implementation method thereof, the method is applicable to an electronic device, the electronic device may be Figure 1 The intellectual property and academic assistant system 100 in the illustrated implementation environment may also be a desktop computer, a laptop computer, a server, and the like.

[0066] In the following method embodiments, for ease of description, the execution subject of each step of the method is taken as an electronic device as an example for illustration, but this does not constitute a specific limitation.

[0067] like Figure 2 As shown, the method may include the following steps:

[0068] Step 210, obtain question information, and input the question information into an intent classification model for intent classification, to obtain an intent classification result indicating the question category of the question information.

[0069] Among them, the question categories of question information include intellectual property questions and answers, patent search, and academic questions and answers.

[0070] Specifically, the question category of the question information is the result of the determination by the intellectual property and academic assistant system according to the intention of the question information input by the user. By classifying the acquired question information, the processing flow of the question information is determined.

[0071] In one possible implementation, the intent classification model is a three-classification model based on BERT.

[0072] Step 230: If the intent classification result indicates that the question category of the question information is a patent search category, the question information is input into a patent search model to generate a patent search result.

[0073] Among them, the patent search model is a model obtained by targeted training for patent search. The patent search model includes a full-text search engine unit and a re-ranking unit.

[0074] For example, in response to the question information entered by the user, the patent search model quickly searches the patent documents stored in the database through metadata such as the patent title, abstract, applicant, inventor, application date, patent classification number, etc. in the question information to generate corresponding patent search results.

[0075] Step 250: If the intent classification result indicates that the question category of the question information is an intellectual property question and answer category, the question information is input into the intellectual property big model to generate an intellectual property answer result.

[0076] Among them, the intellectual property big model is a model that generates targeted answers to questions related to intellectual property in question information.

[0077] For example, when the user inputs a question to find the specific content of a certain article in the Patent Law, the intellectual property big model can generate an intellectual property answer result corresponding to the specific content of the article.

[0078] Step 270: If the intent classification result indicates that the question category of the question information is an academic question and answer category, the question information is input into the academic question and answer model to generate an academic question and answer result.

[0079] Among them, the academic question answering model is a model that generates corresponding literature reviews based on the question information input by the user. The academic question answering model includes a knowledge graph unit and a large language model unit.

[0080] For example, when users input various academic questions, such as "What are the latest research progress in a certain field?" or "What are the best solutions to a certain research problem?", the intellectual property and academic assistant system generates an academic review report in the field based on the academic question-answering model, and provides it to users through visualization tools, such as generating a research field heat map, cooperation network map, and results distribution map for the field.

[0081] Through the above process, after the intent classification is performed through the intent classification model, the question information is input into the corresponding model for processing, which improves the pertinence of the intellectual property and academic assistant system and the efficiency of each model in processing question information, implements natural language processing of users, can understand the natural language questions of users, and provide accurate and useful answers based on large language models and knowledge graphs. It greatly improves the efficiency and accuracy of users in the knowledge retrieval and decision-making process, and can automatically generate comprehensive relevant results according to user needs, improve the efficiency of users in obtaining information based on the results, and can effectively solve the problem of low efficiency of intellectual property and academic assistant systems in related technologies.

[0082] In an exemplary embodiment, if Figure 3 , step 230 may include the following steps:

[0083] Step 231, extract keywords from the question information to generate search term information.

[0084] Specifically, keywords related to metadata in patent documents in the database are extracted from question information through a large language model, and corresponding search terms are generated.

[0085] Step 233, inputting the search term information into the full-text search engine unit to perform a relevance search and obtain a predetermined first number of search results;

[0086] Specifically, by constructing keywords related to metadata, the metadata of patent documents is judged for relevance in the database, and patent documents with relevance are obtained until the number of documents obtained reaches a first number. For example, after the user enters question information, the patent search model quickly searches the database for the 100 patents most relevant to the question information.

[0087] In a possible implementation, the full-text search engine unit is an Elasticsearch search engine.

[0088] Step 235, reorder the search results based on the reordering unit to obtain patent search results corresponding to the question information.

[0089] Among them, during the re-ranking, a further and more accurate judgment is made on the relevance between the patent documents in the search results and the question information, thereby generating patent search results indicating the highest relevance to the question information.

[0090] In a possible implementation, the reordering unit is a BERT-based reordering model.

[0091] Through the above process, search terms for metadata are generated, which improves the accuracy and relevance of the search. Through the search and sorting of the full-text search engine unit and the re-ranking unit, the relevance of the patent documents can be accurately judged, thereby improving the search efficiency and the accuracy of the patent search results.

[0092] In an exemplary embodiment, if Figure 4 As shown, step 270 may include the following steps:

[0093] Step 271, input the question information into the knowledge graph unit to perform knowledge graph retrieval, and determine the background retrieval information corresponding to the question information.

[0094] Among them, the knowledge graph unit is a structured data model that represents patent knowledge in the form of a graph, which expresses entities and their connections through the relationship between nodes and edges.

[0095] In an exemplary embodiment, if Figure 5 As shown in the figure, the construction process of the knowledge graph can include the following steps:

[0096] Step 510, obtaining patent academic data.

[0097] Among them, patent academic data is consistent data that has undergone data cleaning and preprocessing.

[0098] For example, by pre-processing research papers, patent documents and other data in academic databases and literature resources such as data cleaning, deduplication, and format normalization, we can ultimately generate patent academic data in a unified format.

[0099] Step 530, perform entity association processing on the patent academic data to obtain key information entities and entity relationships between key information entities.

[0100] Among them, entity association processing includes word segmentation, named entity recognition and relationship extraction.

[0101] Specifically, the patent academic data is subjected to entity recognition through a large language model to obtain key information entities in the patent academic data, such as research topics, methods, conclusions, authors, institutions, etc., and to determine the entity relationships between entities, such as which topic the research method is applied to, the relationship between authors and institutions, etc.

[0102] Among them, the literature text in the academic database is decomposed into words or phrases through word segmentation, named entity recognition is used to identify the key information entities in the text, and relationship extraction is used to identify the relationship between entities. Finally, the domain-specific pre-trained model and knowledge base are combined to perform entity disambiguation and relationship verification to generate accurate key information entities and the entity relationships between key information entities.

[0103] In one possible implementation, the large language model is GPT-4.

[0104] Step 550, store the key information entities and the corresponding entity relationships in the graph database to obtain a knowledge graph unit.

[0105] It should be noted that in the knowledge graph, each key information entity and entity relationship exists in the form of nodes and edges. When storing in the graph database, the key information entities and corresponding entity relationships are stored in the graph database by adding attribute information corresponding to the key information entities and entity relationships to the nodes and edges in the graph database, such as research direction, publication year, impact factor, etc.

[0106] In a possible implementation, the graph database is a Neo4j database.

[0107] Through the above process, the generation of knowledge graph is realized. By performing entity association processing on patent academic data, the consistency and quality of the data, as well as the accuracy and relevance of entity recognition and relationship extraction, are ensured, thereby improving the accuracy of the intellectual property and academic assistant system. By storing the content of the enriched knowledge graph in the graph database, a multi-dimensional knowledge graph is obtained, which can support complex queries and analysis, thereby improving the efficiency of the intellectual property and academic assistant system.

[0108] Step 273, input the background search information and question information into a preset large language model unit to obtain an academic question-answering result corresponding to the question information.

[0109] In an exemplary embodiment, if Figure 6 As shown, step 273 may include the following steps:

[0110] Step 2731, performing multi-dimensional analysis on key information entities in the background search information based on the large language model unit, and generating user portrait information corresponding to the key information entities;

[0111] Among them, user portrait information is relevant data generated for a certain person in the key information entity.

[0112] For example, through the entity and relationship data in the knowledge graph, we can extract information such as the researcher's research direction, representative achievements, partners, and publication records, conduct multi-dimensional analysis on the researcher including research direction, representative achievements, partners, publication records, etc., and generate a corresponding user information portrait.

[0113] Step 2733, generate academic question and answer results based on the user portrait information corresponding to each key information entity.

[0114] Specifically, by querying key information entities related to the question information in the knowledge graph, the relevant major research results, important authors and institutions are determined, and then a review report including research trend charts, institutional contribution charts, etc. is generated for the corresponding user portrait information to generate academic question and answer results.

[0115] Through the above process, the generation of academic question and answer results is achieved, so that users can directly understand the comprehensive situation in the academic field based on the academic question and answer results. The large language model unit provides analysis and data support for the academic question and answer results, thereby improving the accuracy of the academic question and answer results.

[0116] In an exemplary embodiment, the intent classification of the question information is performed by an intent classification model, such as Figure 7 As shown in the figure, the process of building the intent classification model includes the following steps:

[0117] Step 710, obtaining a first training set, and training a preset classification pre-training model based on the first training set to obtain a classification prediction result generated by the classification pre-training model.

[0118] Among them, the first training set includes intellectual property question and answer data corresponding to the intellectual property question and answer category, patent search data corresponding to the patent search category, and patent academic data corresponding to the academic question and answer category.

[0119] For example, the first training set is a data set that includes 1 million intellectual property question and answer data and patent academic data and 150,000 patent search data. Among them, 500,000 of the intellectual property question and answer data and patent academic data are general field questions and answers from general instruction fine-tuning data, Baidu Knows Q&A, online Q&A, etc., 200,000 are patent questions and answers and legal questions and answers from Deli’s internal data, and 300,000 are academic question and answer data.

[0120] In one possible implementation, the classification pre-training model is a RoBERTa-base-chinese model.

[0121] Step 730 , performing a model performance evaluation on the classification prediction result based on the loss function, and obtaining an evaluation result indicating the prediction accuracy of the classification pre-training model for the data types of the first training set.

[0122] The loss function is used to calculate the difference between the classification prediction results predicted by the model and the actual reference labels, thereby quantifying the prediction ability of the pre-trained model as prediction accuracy, so that the pre-trained model can be trained in the direction of higher prediction accuracy during training.

[0123] In one possible implementation, the loss function is a cross entropy loss function.

[0124] Step 750, based on the evaluation results, adjust the parameters of the pre-trained model and repeat the training until the prediction accuracy of the classification pre-trained model reaches a preset threshold, thereby obtaining an intent classification model.

[0125] Specifically, the parameters of the classification pre-training model are continuously adjusted through the Adam optimizer, and the Dropout mechanism is added for training, so that the model is trained in the direction of being able to classify more difficult samples, thereby significantly improving the classification performance of the trained model in different categories.

[0126] In one possible implementation, a validation set is used to evaluate model performance and adjust parameters when the classification pre-training model is being trained. After the classification pre-training model is trained, the accuracy and performance of the model are evaluated through an independent test set to determine whether the classification pre-training model is trained.

[0127] Through the above process, the classification pre-training model is trained to generate an intent classification model, thereby ensuring the excellent feature extraction capability and classification performance of the intellectual property and academic assistant system, while preventing overfitting during the training process, ensuring the accuracy and classification performance of the intent classification model, and ensuring its reliability in practical applications.

[0128] In an exemplary embodiment, if Figure 8 As shown, the construction process of the intellectual property question-answering model may include the following steps:

[0129] Step 810: Obtain a second training set.

[0130] Among them, the second training set includes intellectual property question and answer data and general question and answer data.

[0131] In one possible implementation, the content of the intellectual property field covering various aspects such as patents, trademarks, and copyrights is cleaned and reconstructed, errors are automatically identified and corrected, and the corresponding answers in natural language are generated through GPT-4o to obtain intellectual property question and answer data. By obtaining encyclopedic knowledge and frequently asked questions, general question and answer data common to various fields is generated.

[0132] Step 830, fine-tune the model parameters of the preset question-and-answer pre-training model based on the second training set to obtain the intellectual property question-and-answer large model.

[0133] The second training set is used to adaptively fine-tune the parameters of the question-answering pre-training model, so that the question-answering pre-training model continuously learns the answering method and answer content of the second training set during training, thereby generating a large intellectual property question-answering model that can generate answer content for question information.

[0134] In one possible implementation, the question-answering pre-training model is the Qwen2-7b-instruct model.

[0135] Through the above process, the generation of the intellectual property question and answer big model is realized, so that the intellectual property and academic assistant system can generate targeted answers to different fields of question information. At the same time, the generalization of the model is guaranteed through the common question and answer data, ensuring the robustness of the intellectual property question and answer big model in dealing with diversified problems. The intellectual property question and answer data enhances the question and answer ability of the intellectual property question and answer big model in the field of intellectual property, thereby improving the efficiency of the intellectual property and academic assistant system.

[0136] Fig. 9 This is a specific implementation diagram of an intellectual property and academic assistant system and its implementation method in an application scenario. In this application scenario, the user inputs question information into the intent classification model, and the intent classification is a three-classification model based on BERT. After identifying the user's intent through the question information, the intent classification result is generated.

[0137] Based on the intent classification results, the system determines that the question category of the question information is one of the intellectual property Q&A category, patent search category, and academic Q&A category, and thus decides to call the corresponding service, namely patent search service, intellectual property Q&A, and academic Q&A.

[0138] When the intent classification result indicates that the question category of the question information is a patent search category, the system calls the patent search service. First, the large language model accepts the question information and extracts the keywords for patent search from it. Then, the system uses Elasticsearch to search the database for the obtained keywords and recalls the 100 most relevant results. Secondly, the BERT-based reranking model is used to finely sort the relevance of these 100 patents. Finally, the n patents with the greatest relevance are selected according to the user's settings and returned to the user.

[0139] When the intent classification result indicates that the question information is classified as an intellectual property question and answer category, the question information is input into the intellectual property big model for processing. After the intellectual property big model generates corresponding answer information for the question information, the result is returned to the user.

[0140] When the intent classification result indicates that the question category of the question information is the academic question answering category, the system calls the academic question answering service. First, the system retrieves question-related information from the knowledge graph based on the user's input as the background information required for the answer. Then, the question information entered by the user and the retrieved background information are used as the input of the large language model to obtain the model's answer.

[0141] Finally, the system returns the retrieved patents or answer results generated by the large language model to the user interface or platform.

[0142] In this application scenario, the Intellectual Property and Academic Assistant System can answer questions about various aspects such as patents, trademarks, and copyrights. By collecting a large amount of high-quality intellectual property question-and-answer data and combining the powerful generation capabilities of the large language model, the Intellectual Property and Academic Assistant System can provide accurate and detailed professional intellectual property consulting services. It can effectively help users solve complex intellectual property issues and improve their work efficiency and professional level. Among them, the large language model can automatically generate literature abstracts and associate them with relevant information in the knowledge graph, thereby helping researchers to conduct literature reviews and research more efficiently. It not only improves the efficiency of literature processing, but also enhances the comprehensiveness and depth of literature reviews. By analyzing the data in the knowledge graph, a detailed portrait of the researcher can be quickly constructed, providing multi-dimensional research directions, results, partners and other information. The personalized user portrait function can help researchers better understand the research dynamics of themselves and their peers, thereby promoting academic cooperation and innovation.

[0143] The following is an embodiment of the device of the present application, which can be used to execute the intellectual property and academic assistant system and its implementation method involved in the present application. For details not disclosed in the embodiment of the device of the present application, please refer to the method embodiment of the intellectual property and academic assistant system and its implementation method involved in the present application.

[0144] See also Fig.10 In an embodiment of the present application, an intellectual property and academic assistant system device 1000 is provided, including but not limited to: an intention classification module 1010, a patent search module 1030, an intellectual property question and answer module 1050 and an academic question and answer module 1070.

[0145] The intent classification module 1010 is used to obtain question information, input the question information into the intent classification model for intent classification, and obtain an intent classification result indicating the question category of the question information, wherein the question category of the question information includes intellectual property question and answer category, patent search category, and academic question and answer category;

[0146] The patent search module 1030 is used to input the question information into a patent search model to generate a patent search result if the intent classification result indicates that the question category of the question information is a patent search category;

[0147] The intellectual property question and answer module 1050 is used to input the question information into the intellectual property big model to generate an intellectual property answer result if the intent classification result indicates that the question category of the question information is the intellectual property question and answer category;

[0148] The academic question and answer module 1070 is used to input the question information into the academic question and answer model to generate an academic question and answer result if the intention classification result indicates that the question category of the question information is the academic question and answer category.

[0149] It should be noted that the intellectual property and academic assistant system device provided in the above embodiment only uses the division of the above-mentioned functional modules as an example when implementing the intellectual property and academic assistant system. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the intellectual property and academic assistant system device will be divided into different functional modules to complete all or part of the functions described above.

[0150] In addition, the intellectual property and academic assistant system device and the intellectual property and academic assistant system and the implementation method embodiments provided in the above embodiments belong to the same concept, and the specific manner in which each module performs operations has been described in detail in the method embodiments and will not be repeated here.

[0151] Fig.11 A schematic diagram of the structure of a server is shown according to an exemplary embodiment. The server is suitable for Figure 1 A server 130 in an implementation environment is shown.

[0152] It should be noted that the server is only an example adapted to the present application and cannot be considered to provide any limitation on the scope of use of the present application. The server cannot be interpreted as requiring dependence on or having Fig.11 One or more components of exemplary server 2000 are shown.

[0153] The hardware structure of the server 2000 may vary greatly due to different configurations or performances, such as Fig.11 As shown, the server 2000 includes: a power supply 210 , an interface 230 , at least one memory 250 , and at least one central processing unit (CPU) 270 .

[0154] Specifically, the power supply 210 is used to provide operating voltage for each hardware device on the server 2000 .

[0155] The interface 230 includes at least one wired or wireless network interface 231 for interacting with external devices. Figure 1The interaction between the collection end 110 and the service end 130 in the implementation environment is shown.

[0156] Of course, in other examples adapted by the present application, the interface 230 may further include at least one serial-to-parallel conversion interface 233, at least one input-output interface 235, and at least one USB interface 237, etc. Fig.11 As shown, this is not intended to be a specific limitation.

[0157] The memory 250 is a carrier for storing resources, which may be a read-only memory, a random access memory, a disk or an optical disk, etc. The resources stored thereon include an operating system 251, an application 253 and data 255, etc. The storage method may be temporary storage or permanent storage.

[0158] Among them, the operating system 251 is used to manage and control the hardware devices and application programs 253 on the server 2000 to enable the central processor 270 to calculate and process the massive data 255 in the memory 250. It can be WindowsServerTM, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM, etc.

[0159] The application program 253 is a computer-readable instruction that performs at least one specific task based on the operating system 251, and may include at least one module ( Fig.11 (not shown), each module may respectively include computer-readable instructions for the server 2000. For example, the intellectual property and academic assistant system device may be regarded as an application 253 deployed on the server 2000.

[0160] The data 255 may be photos, pictures, etc. stored in a disk, or may be recommendation information, etc., stored in the memory 250 .

[0161] The central processor 270 may include one or more processors and is configured to communicate with the memory 250 through at least one communication bus to read the computer-readable instructions stored in the memory 250, thereby realizing the operation and processing of the mass data 255 in the memory 250. For example, the intellectual property and academic assistant system and its implementation method are completed by the central processor 270 reading a series of computer-readable instructions stored in the memory 250.

[0162] In addition, the present application can also be implemented through hardware circuits or hardware circuits combined with software. Therefore, the implementation of the present application is not limited to any specific hardware circuits, software, or a combination of the two.

[0163] See also Fig.12In an embodiment of the present application, an electronic device 4000 is provided, and the electronic device 400 may include: a desktop computer, a laptop computer, a server, etc.

[0164] exist Fig.12 In the embodiment, the electronic device 4000 includes at least one processor 4001 and at least one memory 4003.

[0165] The data interaction between the processor 4001 and the memory 4003 can be realized through at least one communication bus 4002. The communication bus 4002 may include a path for transmitting data between the processor 4001 and the memory 4003. The communication bus 4002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The communication bus 4002 can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Fig.12 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0166] Optionally, the electronic device 4000 may further include a transceiver 4004, which may be used for data interaction between the electronic device and other electronic devices, such as data transmission and / or data reception. It should be noted that in actual applications, the transceiver 4004 is not limited to one, and the structure of the electronic device 4000 does not constitute a limitation on the embodiments of the present application.

[0167] Processor 4001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It may implement or execute various exemplary logic blocks, modules and circuits described in conjunction with the disclosure of this application. Processor 4001 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0168] The memory 4003 can be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compressed optical disk, laser disk, optical disk, digital versatile disk, Blu-ray disk, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program instructions or codes in the form of instructions or data structures and can be accessed by the electronic device 400, but is not limited to this.

[0169] The memory 4003 stores computer-readable instructions, and the processor 4001 can read the computer-readable instructions stored in the memory 4003 through the communication bus 4002 .

[0170] The computer-readable instructions are executed by one or more processors 4001 to implement the intellectual property and academic assistant system and its implementation method in the above-mentioned embodiments.

[0171] In addition, a storage medium is provided in an embodiment of the present application, on which computer-readable instructions are stored, and the computer-readable instructions are executed by one or more processors to implement the above-mentioned intellectual property and academic assistant system and its implementation method.

[0172] A computer program product is provided in an embodiment of the present application. The computer program product includes computer-readable instructions. The computer-readable instructions are stored in a storage medium. One or more processors of an electronic device read the computer-readable instructions from the storage medium, load and execute the computer-readable instructions, so that the electronic device implements the above-mentioned intellectual property and academic assistant system and its implementation method.

[0173] Compared with the related art, this application uses the intent classification model to classify the intent, and then inputs the question information into the corresponding model for processing, thereby improving the pertinence of the intellectual property and academic assistant system and the efficiency of each model in processing question information, implementing natural language processing of users, being able to understand the natural language questions of users, and providing accurate and useful answers based on large language models and knowledge graphs. It greatly improves the efficiency and accuracy of users in the knowledge retrieval and decision-making process, and can automatically generate comprehensive relevant results according to user needs, improving the efficiency of users in obtaining information based on the results, thereby effectively solving the problem of low efficiency of the intellectual property and academic assistant system in related technologies. It realizes the generation of search terms for metadata, improves the accuracy and relevance of the retrieval, and realizes accurate judgment of the relevance of patent documents through retrieval and sorting of the full-text search engine unit and the re-ranking unit, thereby improving the retrieval efficiency and the accuracy of patent retrieval results. The generation of knowledge graphs is realized. By processing the entity association of patent academic data, the consistency and quality of data are ensured, and the accuracy and relevance of entity recognition and relationship extraction are improved, so as to improve the accuracy of the intellectual property and academic assistant system. By storing the content of the rich knowledge graph through the graph database, a multi-dimensional knowledge graph is obtained, which can support complex queries and analysis, and improve the efficiency of the intellectual property and academic assistant system. The generation of academic question and answer results is realized, so that users can directly understand the comprehensive situation in the academic field based on the academic question and answer results. The large language model unit provides analysis and data support for the academic question and answer results, and improves the accuracy of the academic question and answer results. The training of the classification pre-training model generates an intention classification model, thereby ensuring the excellent feature extraction ability and classification performance of the intellectual property and academic assistant system, while preventing overfitting during the training process, ensuring the accuracy and classification performance of the intention classification model, and ensuring reliability in practical applications. The intellectual property and academic assistant system can answer questions about various aspects such as patents, trademarks, and copyrights. By collecting a large amount of high-quality intellectual property question and answer data and combining the powerful generation ability of the large language model, the intellectual property and academic assistant system can provide accurate and detailed professional intellectual property consulting services. It can effectively help users solve complex intellectual property issues and improve their work efficiency and professional level. Among them, the large language model can automatically generate literature abstracts and associate them with relevant information in the knowledge graph, thereby helping researchers to conduct literature reviews and research more efficiently. It not only improves the efficiency of literature processing, but also enhances the comprehensiveness and depth of literature reviews. By analyzing the data in the knowledge graph, a detailed portrait of the researcher can be quickly constructed, providing multi-dimensional research directions, results, partners and other information. The personalized user portrait function can help researchers better understand the research dynamics of themselves and their peers, thereby promoting academic cooperation and innovation.

[0174] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0175] The above are only some implementation methods of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. An intellectual property and academic assistant system and its implementation method, characterized in that: include: Acquire question information, and input the question information into an intent classification model for intent classification, to obtain an intent classification result indicating a question category of the question information, wherein the question category of the question information includes an intellectual property question and answer category, a patent search category, and an academic question and answer category; If the intent classification result indicates that the question category of the question information is a patent search category, inputting the question information into a patent search model to generate a patent search result; If the intent classification result indicates that the question category of the question information is an intellectual property question and answer category, inputting the question information into the intellectual property big model to generate an intellectual property answer result; If the intention classification result indicates that the question category of the question information is an academic question and answer category, the question information is input into the academic question and answer model to generate an academic question and answer result.

2. The method according to claim 1, characterized in that The patent search model includes a full-text search engine unit and a re-ranking unit; The step of inputting the question information into a patent search model to generate patent search results includes: Extract keywords from the question information to generate search term information; Inputting the search term information into a full-text search engine unit to perform a relevance search to obtain a predetermined first number of search results; The search results are reordered based on the reordering unit to obtain patent search results corresponding to the question information.

3. The method according to claim 1, characterized in that The academic question answering model includes a knowledge graph unit and a large language model unit; If the intention classification result indicates that the question category of the question information is an academic question and answer category, the question information is input into an academic question and answer model to generate an academic question and answer result, including: Inputting the question information into the knowledge graph unit to perform knowledge graph retrieval, and determining background retrieval information corresponding to the question information; The background search information and the question information are input into a preset large language model unit to obtain an academic question-answering result corresponding to the question information.

4. The method according to claim 3, characterized in that Before inputting the question information into the knowledge graph unit to perform knowledge graph retrieval and generate background retrieval information, the method further includes: Access to patent academic data; Performing entity association processing on the patent academic data to obtain key information entities and entity relationships between the key information entities, wherein the entity association processing includes word segmentation, named entity recognition and relationship extraction; The key information entities and the corresponding entity relationships are stored in a graph database to obtain a knowledge graph unit.

5. The method according to claim 3, characterized in that The step of inputting the background search information and the question information into a preset large language model unit to obtain an academic question answering result corresponding to the question information includes: Performing multi-dimensional analysis on key information entities in the background search information based on the large language model unit to generate user portrait information corresponding to the key information entities; Based on the user portrait information corresponding to each key information entity, academic question and answer results are generated.

6. The method according to claim 1, characterized in that Before inputting the question information into the intent classification model for intent classification, the method further includes: Obtaining a first training set, and training a preset classification pre-training model based on the first training set to obtain a classification prediction result generated by the classification pre-training model, wherein the first training set includes intellectual property question and answer data corresponding to the intellectual property question and answer category, patent search data corresponding to the patent search category, and patent academic data corresponding to the academic question and answer category; Performing a model performance evaluation on the classification prediction result based on a loss function to obtain an evaluation result indicating the prediction accuracy of the classification pre-training model for the data types of the first training set; Based on the evaluation result, the parameters of the pre-trained model are adjusted and the training is repeated until the prediction accuracy of the classification pre-trained model reaches a preset threshold, thereby obtaining an intent classification model.

7. The method according to claim 1, characterized in that Before inputting the question information into the intellectual property big model to generate the intellectual property answer result, the method further includes: Acquire a second training set, wherein the second training set includes intellectual property question and answer data and general question and answer data; Based on the second training set, the model parameters of the preset question and answer pre-training model are fine-tuned to obtain the intellectual property question and answer large model.

8. An intellectual property and academic assistant system device, characterized in that: include: An intent classification module is used to obtain question information, and input the question information into an intent classification model for intent classification, to obtain an intent classification result indicating a question category of the question information, wherein the question category of the question information includes an intellectual property question and answer category, a patent search category, and an academic question and answer category; A patent search module, for inputting the question information into a patent search model to generate a patent search result if the intent classification result indicates that the question category of the question information is a patent search category; an intellectual property question-answering module, configured to input the question information into an intellectual property big model to generate an intellectual property answer result if the intent classification result indicates that the question category of the question information is an intellectual property question-answering category; The academic question-answering module is used to input the question information into the academic question-answering model to generate an academic question-answering result if the intention classification result indicates that the question category of the question information is an academic question-answering category.

9. An electronic device, characterized in that: include: at least one processor and at least one memory, wherein: The memory has computer-readable instructions stored thereon; The computer-readable instructions are executed by one or more of the processors, so that the electronic device implements the intellectual property and academic assistant system and its implementation method as described in any one of claims 1 to 7.

10. A storage medium having computer-readable instructions stored thereon, characterized in that: The computer-readable instructions are executed by one or more processors to implement the intellectual property and academic assistant system and its implementation method as described in any one of claims 1 to 7.

Images