Patent innovation point detection method, device and equipment based on large model

Abstract

The invention provides a patent innovation point detection method, device and equipment based on a large model. Wherein after the to-be-evaluated text is obtained, based on a pre-trained large model in the patent innovation judgment field, retrieval word generation and retrieval are performed to obtain related prior art and innovation evaluation is performed, so that intelligent and precise detection of patent innovation points can be realized, the efficiency is improved, and the labor cost is reduced; and a powerful technical support is provided for patent examination and evaluation work.

Description

Technical Field

[0001] This application relates to a method, device and equipment for detecting patent innovation points based on a large model, and belongs to the field of artificial intelligence technology. Background Art

[0002] The current patent innovation assessment process usually includes the following carefully designed stages:

[0003] Phase 1: Accurate generation of search terms

[0004] During this phase, patent agent engineers draw on their deep expertise to thoroughly interpret the core technical solutions of the patents to be evaluated (or their technical disclosure documents, hereinafter referred to as the "target patents") and manually extract a series of highly relevant search terms. This process fully demonstrates the engineers' professionalism and keen ability to grasp technical details.

[0005] Stage 2: Comprehensive search of prior art

[0006] Using the search terms generated in Phase 1, we conduct an extensive and in-depth search for prior art related to or similar to the target patent's technical solution, leveraging diverse search platforms such as patent databases, academic paper libraries, and technical report libraries. This step aims to compile comprehensive technical information, laying a solid foundation for subsequent comparative analysis.

[0007] Stage 3: In-depth judgment of innovation points

[0008] After obtaining extensive prior art information, the evaluator or patent engineer, drawing on their deep technical background and extensive examination experience, must conduct a meticulous comparative analysis of the retrieved prior art and the technical solutions of the target patent. This process requires not only accurate identification of whether the prior art already discloses the key technical features of the target patent, but also a comprehensive consideration of the novelty and inventiveness of the technical solutions. Because this process involves a high degree of technical relevance and in-depth consideration, it must be personally conducted by a professionally qualified and experienced evaluator or patent engineer to ensure the accuracy and authority of the evaluation results.

[0009] The shortcoming of the above solution is that it requires manual participation in "search term generation" and "innovation point judgment", resulting in low efficiency and high cost of patent innovation evaluation. Summary of the Invention

[0010] This application provides a patent innovation point detection method, device and equipment based on a large model to solve the problem that patent innovation point detection in existing solutions requires manual participation, resulting in low efficiency and high cost.

[0011] In order to achieve the above objectives, this application provides the following technical solutions:

[0012] In a first aspect, the present application provides a method for detecting patent innovations based on a large model, which includes:

[0013] Get the text to be evaluated;

[0014] Based on the text to be evaluated, a pre-trained large model in the field of patent innovation judgment is called to generate search terms;

[0015] Based on the search terms, a large model in the field of patent innovation judgment is called to perform prior art search to obtain multiple prior art documents;

[0016] The large model in the field of patent innovation judgment is called to evaluate the innovation of the text to be evaluated relative to the multiple prior art documents to generate an innovation evaluation result.

[0017] Based on the above method, optionally, the step of generating search terms based on the text to be evaluated by calling a pre-trained large model in the field of patent innovation judgment includes:

[0018] Calling a pre-trained large model in the field of patent innovation judgment to generate keywords and categories based on the text to be evaluated;

[0019] Generate synonyms based on the generated keywords using the large model of the patent innovation judgment field;

[0020] By utilizing the large model of the patent innovation judgment field, reasonable search terms are combined based on keywords, synonyms of keywords and category words.

[0021] Based on the above method, optionally, calling the large model of the patent innovation judgment field to evaluate the innovation of the text to be evaluated relative to the multiple prior art documents to generate an innovation evaluation result includes:

[0022] Calling the big model in the field of patent innovation judgment to extract the technical features or innovative points of the text to be evaluated;

[0023] Compare and analyze the technical features or innovations with a certain existing technical document to determine whether the technical features or innovations are distinguishing technical features;

[0024] The innovation evaluation result of the text to be evaluated is generated by integrating the comparative analysis results of the text to be evaluated with all relevant prior art documents through the large model of the patent innovation judgment field.

[0025] Based on the above method, optionally, the training method of the large model in the field of patent innovation judgment includes:

[0026] Obtain a preset amount of unannotated patent data;

[0027] Based on the unlabeled patent data, a continued pre-training method is used to train the general large model into a patent industry large model;

[0028] By using labeled search terms to generate training data, labeled document comparison training data, and labeled innovative conclusions to generate training data, the patent industry big model is supervised and fine-tuned to obtain a big model in the field of patent innovativeness judgment.

[0029] Based on the above method, optionally, based on the unlabeled patent data, a continued pre-training method is used to train the general large model into a patent industry large model, including:

[0030] Collect patent documents from major patent databases, clean and preprocess the collected patent document data, and extract the key parts of the patent documents; divide the preprocessed patent documents into training sets, validation sets, and test sets according to the set ratio;

[0031] Input the text in the training set into the general large model for further pre-training; during the training process, adjust the training strategy and introduce regularization or dropout technology to prevent overfitting based on the characteristics of the patent data; and regularly evaluate the performance of the trained large model on the validation set, and based on the evaluation results, tune the structure and / or parameters of the large model.

[0032] Based on the above method, optionally, the method generates training data using labeled search terms, labeled document comparison training data, and labeled innovative conclusion training data, and fine-tunes the patent industry macromodel through supervision to obtain a macromodel for patent innovativeness judgment, including:

[0033] Collect a preset number of patent documents and briefing documents as given patent data, and extract keywords from the given patent data;

[0034] The processed data is annotated, including keyword annotation, keyword synonym annotation, and category annotation; keywords, keyword synonyms, and categories are combined through "and" and "or" to form search term annotations, and the annotated search terms are obtained to generate training data;

[0035] Retrieve and collect relevant patents that are relevant to the given patent data, including relevance in technical fields and application scenarios; annotate the technical features of the given patent data and related patents to determine whether they are distinguishing technical features and the judgment process, and obtain annotated document comparison training data;

[0036] Combined with the document comparison and annotation results, annotate whether the given patent data is innovative and the judgment process, and obtain the annotated innovative conclusion to generate training data;

[0037] The training data generated by the above steps is used to supervise and fine-tune the large model of the patent industry to obtain a large model in the field of patent innovation judgment.

[0038] In a second aspect, the present application also provides a patent innovation point detection device based on a large model, which includes:

[0039] The acquisition module is used to obtain the text to be evaluated;

[0040] A search term generation module is used to generate search terms based on the text to be evaluated by calling a pre-trained large model in the field of patent innovation judgment;

[0041] A prior art search module is used to call the large model of the patent innovation judgment field to perform prior art search based on the search term to obtain multiple prior art documents;

[0042] The innovation evaluation module is used to call the large model of the patent innovation judgment field to evaluate the innovation of the text to be evaluated relative to the multiple prior art documents and generate an innovation evaluation result.

[0043] In a third aspect, an embodiment of the present application further provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor calls and executes the computer program, the method for detecting patent innovation points based on a large model as described in any one of the first aspects is implemented.

[0044] In the patent innovation point detection method, device and equipment based on the big model provided in this application, after obtaining the text to be evaluated, search terms are generated, relevant existing technologies are retrieved and innovation evaluation is performed based on the pre-trained big model in the field of patent innovation judgment. This can realize intelligent and precise detection of patent innovation points, improve efficiency, reduce labor costs, and provide strong technical support for patent examination and evaluation work. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The accompanying drawings are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present application, and together with the description, serve to explain the principles of the present application. In addition, these drawings and the description are not intended to limit the scope of the concept of the present application in any way, but rather to illustrate the concept of the present application for those skilled in the art by reference to specific embodiments.

[0046] Figure 1 A schematic diagram of a process for detecting patent innovation points based on a large model according to an embodiment of the present application;

[0047] Figure 2 A schematic diagram of a process for generating search terms based on a large model provided in one embodiment of the present application;

[0048] Figure 3 A schematic diagram of a process for generating innovative conclusions based on a large model, provided in one embodiment of the present application;

[0049] Figure 4 A schematic diagram of the overall process of patent innovation detection based on a large model provided in one embodiment of the present application;

[0050] Figure 5 A schematic diagram of the structure of a patented innovation point detection device based on a large model provided in one embodiment of the present application;

[0051] Figure 6 A schematic diagram of the structure of an electronic device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0052] To make the purpose, technical solutions and advantages of this application clearer, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the embodiments of this application. Obviously, the described embodiments are part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0053] Patent innovation testing is a specialized review process for the technical solutions in patent applications, designed to determine whether the solutions possess innovative qualities, specifically whether they meet the standards of novelty and inventiveness stipulated in the Patent Law. Furthermore, this process demonstrates significant importance in assessing patent value, assisting with patent searches, and conducting in-depth analysis.

[0054] The current patent innovation evaluation process requires deep human involvement and places high demands on the capabilities of relevant personnel, leading to problems such as low efficiency and high costs.

[0055] In order to solve the above problems, the present application provides a patent innovation point detection solution based on large model technology. First, the large model is used to deeply analyze the content of the patent to be evaluated (or related technical briefing documents, collectively referred to as "target patents"), and intelligently recommend a series of search terms closely related to the target patent. Subsequently, these search terms can be used to accurately retrieve a number of patent documents, academic papers or technical reports associated with the target patent in a patent database, academic paper library or technical report platform. Finally, the large model plays a key role again, and through a comprehensive comparative analysis of the target patent and the above search results, the innovativeness of the target patent is scientifically and objectively evaluated and revealed. In this way, manual participation can be effectively reduced, thereby improving efficiency and reducing costs. The following is a non-restrictive description of the specific implementation scheme through several examples or embodiments.

[0056] Some embodiments of the present application provide a patent innovation detection method based on a large model. In specific implementation, the method can be executed by an electronic device, such as a computer or a server.

[0057] Reference Figure 1 , Figure 1 The following is a flow chart of a method for detecting patent innovation points based on a large model provided in one embodiment of the present application. Figure 1 As shown, the patent innovation point detection method based on the large model of this embodiment includes the following steps:

[0058] Step S101: Obtain the text to be evaluated.

[0059] The text to be evaluated can be a completed patent text or a technical briefing document.

[0060] Step S102: Based on the text to be evaluated, a pre-trained large model in the field of patent innovation judgment is called to generate search terms.

[0061] Specifically, before executing this step, a specialized large-scale model dedicated to patent innovation assessment must be pre-trained. This model has been carefully designed and optimized to master two core skills: intelligent search term generation and accurate identification of innovative points. This lays a solid foundation for the efficient operation of subsequent processes.

[0062] In this step, you can directly call the trained big model and input the text to be evaluated into the big model. With its excellent natural language processing capabilities, the big model deeply understands the text content and intelligently generates a series of highly relevant and accurate search terms based on this, providing strong support for subsequent retrieval work.

[0063] More specifically, if Figure 2As shown, based on the text to be evaluated, the steps of calling a pre-trained large model in the field of patent innovation judgment to generate search terms may include:

[0064] Call the pre-trained big model in the field of patent innovation judgment to generate keywords and categories based on the text to be evaluated; use the big model in the field of patent innovation judgment to generate synonyms based on the generated keywords; use the big model in the field of patent innovation judgment to combine reasonable search terms based on keywords, synonyms of keywords and category words.

[0065] In this embodiment, the big model uses the COT (Chain of Thought) reasoning process to generate search terms. First, it generates the keywords and categories of the patent, then generates synonyms based on the generated keywords, and finally the big model combines reasonable search terms from the keywords, synonyms of the keywords and category words.

[0066] Through the above solution, search terms can be automatically and intelligently generated without the need for manual analysis of the text to be evaluated, thereby improving efficiency and reducing labor costs.

[0067] Step S103: Based on the search terms, a large model in the field of patent innovation judgment is called to perform prior art search to obtain multiple prior art documents.

[0068] Specifically, in this step, the search terms generated in step S102 can be used to conduct a broad and in-depth search for prior art related to or similar to the technical solution in the text to be evaluated, using authoritative search platforms such as patent databases, academic paper libraries, and technical report libraries. This process is intended to ensure that all potentially relevant technical information is obtained, providing comprehensive and rich data support for subsequent innovation determination.

[0069] Step S104: calling the large model in the field of patent innovation judgment to evaluate the innovation of the text to be evaluated relative to multiple existing technical documents, and generating an innovation evaluation result.

[0070] Specifically, to evaluate the innovativeness of a patent or briefing document relative to the prior art, it is necessary to compare and analyze the prior art of the patent or briefing document with relevant documents in search platforms such as patent databases, academic papers, and technical reports. In this step, the retrieved prior art and the technical solution of the document to be evaluated can be input into a large model for judging the innovativeness of patents. With its powerful comparative analysis and judgment capabilities, this large model accurately identifies whether the prior art has disclosed the key technical features of the target patent, and comprehensively considers the novelty and creativity of the technical solution, ultimately providing an objective and accurate evaluation result.

[0071] More specifically, if Figure 3As shown, the specific implementation method of step S104 may include:

[0072] Call the big model in the field of patent innovation judgment to extract the technical features or innovations of the text to be evaluated; compare and analyze the technical features or innovations with a certain existing technical document to determine whether the technical features or innovations are distinguishing technical features; use the big model in the field of patent innovation judgment to integrate the comparative analysis results of the text to be evaluated and all relevant existing technical documents to generate the innovation evaluation results of the text to be evaluated.

[0073] Through the above solution, large models can replace manual work in existing technology comparison and innovative analysis, greatly improving efficiency and reducing labor costs.

[0074] In this way, in the patent innovation point detection solution based on the big model provided by this application, after obtaining the text to be evaluated, search terms are generated, relevant existing technologies are retrieved, and innovation evaluation is performed based on the pre-trained big model in the field of patent innovation judgment. This can realize intelligent and precise detection of patent innovation points, improve efficiency, reduce labor costs, and provide strong technical support for patent review and evaluation work.

[0075] Furthermore, in some embodiments, a method for training a large model in the field of patent innovation judgment includes:

[0076] Obtain a preset amount of unlabeled patent data; based on the unlabeled patent data, use the continued pre-training method to train the general large model into a patent industry large model; use labeled search terms to generate training data, labeled document comparison training data, and labeled innovative conclusions to generate training data, supervise and fine-tune the patent industry large model, and obtain a large model in the field of patent innovation judgment.

[0077] Specifically, continued pretraining refers to further training an already trained model using new datasets to enhance its performance in specific tasks or fields. In this embodiment, a large amount of unlabeled patent data is used to train a general large model into a large patent industry model using the continued pretraining method, enabling it to be better applied to problem solving in the patent industry.

[0078] More specifically, in some embodiments, based on unlabeled patent data, a continuous pre-training method is used to train a general large model into a patent industry large model, which may specifically include:

[0079] Patent documents are collected from major patent databases, and the collected patent document data is cleaned and preprocessed to extract the key parts of the patent documents; the preprocessed patent documents are divided into training set, validation set and test set according to the set ratio; the text in the training set is input into the general large model for further pre-training; during the training process, the training strategy is adjusted according to the characteristics of the patent data and regularization or dropout technology is introduced to prevent overfitting; and the performance of the trained large model is regularly evaluated on the validation set, and the structure and / or parameters of the large model are tuned based on the evaluation results.

[0080] In this embodiment, the steps of continuing pre-training to obtain a large model of the patent industry can be summarized into the following stages:

[0081] 1) Data preparation stage

[0082] Patent data collection: We collect patent documents from major patent databases (such as the China National Intellectual Property Administration, the European Patent Office's Espacenet, and the World Intellectual Property Organization). We ensure the comprehensiveness and diversity of the data, covering patents in different technological fields and time periods.

[0083] Data cleaning and preprocessing: De-duplication, denoising, and formatting of patent documents. Extract key parts of patent documents, such as titles, abstracts, and claims.

[0084] Construct a training dataset: Divide the preprocessed patent documents into training set, validation set, and test set according to a certain ratio.

[0085] Ensure that the data set is segmented in a representative manner to reflect the overall situation of the patent industry.

[0086] 2) Continue pre-training phase

[0087] Domain data injection: Text from the patent dataset is fed into the large model for further pre-training. During the training process, the large model learns the specific terminology, concepts, and context of the patent industry.

[0088] Training strategy optimization: Adjust the training strategy, such as learning rate, batch size, number of training rounds, etc., based on the characteristics of the patent data.

[0089] Regularization, dropout and other techniques are introduced to prevent overfitting.

[0090] Performance monitoring and tuning: Regularly evaluate the model's performance on the validation set. Based on the evaluation results, fine-tune the model, such as adjusting the model structure and optimizing algorithm parameters.

[0091] After obtaining the patent industry model, the labeled data is used to perform supervised fine-tuning (SFT) on the patent industry model so that it can better handle specific tasks, namely search term generation, prior art comparative retrieval, and innovation evaluation.

[0092] In some embodiments, the patent industry macromodel is supervised and fine-tuned using training data generated by labeled search terms, labeled document comparison training data, and labeled innovative conclusions to obtain a macromodel for patent innovativeness judgment. Specifically, the macromodel may include:

[0093] Collect a preset number of patent documents and briefing data as given patent data, and extract keywords from the given patent data; keywords include patent titles, abstracts, claims, etc., and ensure the comprehensiveness and diversity of the data, covering patents in different technical fields and different time periods.

[0094] The processed data is labeled, including keyword labeling, keyword synonym labeling and category annotation; keywords, keyword synonyms and categories are combined through "And" and "Or" to form search term labeling, and the labeled search terms are obtained to generate training data; the labeling work is carried out by domain experts to ensure the accuracy and effectiveness of the labeling.

[0095] Retrieve and collect relevant patents that are relevant to the given patent data, including relevance in technical fields and application scenarios; for the technical features of the given patent data and related patents, mark whether they are distinguishing technical features and the judgment process, and obtain labeled file comparison training data; here, multiple field experts are required to participate in the labeling, and vote to ensure the accuracy and consistency of the labeling.

[0096] Combined with the document comparison and annotation results, the innovativeness of the given patent data and the judgment process are annotated to obtain the annotated innovative conclusions and generate training data. Similarly, multiple domain experts are required to participate in the annotation to ensure the accuracy and consistency of the annotation.

[0097] The training data generated by the above steps is used to supervise and fine-tune the large model of the patent industry to obtain a large model in the field of patent innovation judgment.

[0098] Through the above scheme, a large model with good performance in the field of patent innovation judgment can be obtained, so that it can be applied in practice to achieve intelligent and accurate detection of patent innovation points. The overall process of patent innovation point detection based on the large model is as follows: Figure 4 shown.

[0099] In addition, the embodiment of the present application provides a patent innovation point detection device based on a large model, referring to Figure 5 , the device comprises:

[0100] An acquisition module 51 is used to acquire the text to be evaluated;

[0101] A search term generation module 52 is used to generate search terms based on the text to be evaluated by calling a pre-trained large model in the field of patent innovation judgment;

[0102] Prior art search module 53, for performing prior art search based on the search terms and calling a large model in the field of patent innovation judgment to obtain multiple prior art documents;

[0103] The innovation evaluation module 54 is used to call a large model in the field of patent innovation judgment to evaluate the innovation of the text to be evaluated relative to multiple existing technical documents and generate an innovation evaluation result.

[0104] Among them, regarding the specific implementation methods of each module of the above-mentioned large-model-based patent innovation point detection device, reference can be made to the corresponding contents in the aforementioned method embodiment, which will not be repeated here.

[0105] In addition, an embodiment of the present application provides an electronic device, such as Figure 6 As shown, the electronic device includes a memory 61 and a processor 62; wherein the memory 61 stores a computer program, and when the processor 62 calls and executes the computer program, the patent innovation point detection method based on the large model in any of the above embodiments is implemented.

[0106] It can be understood that the same or similar parts of the above embodiments can be referenced to each other, and the contents not described in detail in some embodiments can refer to the same or similar contents in other embodiments.

[0107] It should be noted that, in the description of the present invention, the terms "first", "second", etc. are used for descriptive purposes only and should not be understood as indicating or implying relative importance. In addition, in the description of the present invention, unless otherwise specified, the meaning of "plurality" is at least two.

[0108] Throughout this specification, references to terms such as "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" indicate that the specific features, structures, materials, or characteristics described in conjunction with that embodiment or example are included in at least one embodiment or example of the present invention. In this specification, schematic representations of these terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0109] Although the embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A patent innovation point detection method based on a large model, characterized in that: include: Get the text to be evaluated; Based on the text to be evaluated, a pre-trained large model in the field of patent innovation judgment is called to generate search terms; Based on the search terms, a large model in the field of patent innovation judgment is called to perform prior art search to obtain multiple prior art documents; The large model of the patent innovation judgment field is called to evaluate the innovation of the text to be evaluated relative to the multiple prior art documents to generate an innovation evaluation result.

2. The method according to claim 1, characterized in that Based on the text to be evaluated, calling a pre-trained large model in the field of patent innovation judgment to generate search terms includes: Calling a pre-trained large model in the field of patent innovation judgment to generate keywords and categories based on the text to be evaluated; Generate synonyms based on the generated keywords using the large model of the patent innovation judgment field; By using the large model of the patent innovation judgment field, reasonable search terms are combined based on keywords, synonyms of keywords and category words.

3. The method according to claim 1, characterized in that The step of calling the large model of the patent innovation judgment field to evaluate the innovation of the text to be evaluated relative to the multiple prior art documents and generating an innovation evaluation result includes: Calling the big model in the field of patent innovation judgment to extract the technical features or innovative points of the text to be evaluated; Compare and analyze the technical features or innovations with a certain existing technical document to determine whether the technical features or innovations are distinguishing technical features; The large model in the field of patent innovation judgment is used to integrate the comparative analysis results of the text to be evaluated and all relevant prior art documents to generate an innovation evaluation result of the text to be evaluated.

4. The method according to claim 1, characterized in that: The training method of the large model in the field of patent innovation judgment includes: Obtain a preset amount of unannotated patent data; Based on the unlabeled patent data, a continued pre-training method is used to train the general large model into a patent industry large model; The training data are generated by using labeled search terms, labeled document comparison training data and labeled innovative conclusions, and the patent industry big model is supervised and fine-tuned to obtain a big model in the field of patent innovativeness judgment.

5. The method according to claim 4, characterized in that Based on the unlabeled patent data, the general large model is trained into a patent industry large model by using a continuous pre-training method, including: Collect patent documents from major patent databases, clean and preprocess the collected patent document data, and extract the key parts of the patent documents; divide the preprocessed patent documents into training sets, validation sets, and test sets according to the set ratio; Input the text in the training set into the general large model for further pre-training; during the training process, adjust the training strategy and introduce regularization or dropout technology to prevent overfitting according to the characteristics of the patent data; and regularly evaluate the performance of the trained large model on the validation set, and optimize the structure and / or parameters of the large model based on the evaluation results.

6. The method according to claim 4, characterized in that The aforementioned method uses the labeled search terms to generate training data, the labeled document comparison training data, and the labeled innovative conclusions to generate training data, and fine-tunes the patent industry big model to obtain a big model in the field of patent innovativeness judgment, including: Collect a preset number of patent documents and briefing data as given patent data, and extract keywords from the given patent data; The processed data is annotated, including keyword annotation, keyword synonym annotation and category annotation; keywords, keyword synonyms and categories are combined through "And" and "or" to form search term annotations, and the annotated search terms are obtained to generate training data; Retrieve and collect relevant patents that are relevant to the given patent data, including relevance in technical fields and application scenarios; for the given patent data and the technical features of the related patents, mark whether they are distinguishing technical features and the judgment process, and obtain marked file comparison training data; Combined with the document comparison and annotation results, annotate whether the given patent data is innovative and its judgment process, and obtain the annotated innovative conclusion to generate training data; The training data generated by the above steps is used to supervise and fine-tune the patent industry big model to obtain a big model in the field of patent innovation judgment.

7. A patent innovation point detection device based on a large model, characterized in that: include: An acquisition module is used to obtain the text to be evaluated; A search term generation module, used to generate search terms based on the text to be evaluated by calling a pre-trained large model in the field of patent innovation judgment; A prior art search module, for performing a prior art search based on the search terms and calling a large model of the patent innovation judgment field to obtain a plurality of prior art documents; The innovation evaluation module is used to call the large model of the patent innovation judgment field to evaluate the innovation of the text to be evaluated relative to the multiple prior art documents and generate an innovation evaluation result.

8. An electronic device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and when the processor calls and executes the computer program, the method according to any one of claims 1 to 6 is implemented.

Images