Technical document generation method and device, equipment, medium and product
Through the intelligent technical briefing writing system assisted by large-scale language model, the problems of low efficiency, high professional requirements and high cost of writing traditional technical briefings are solved, high-quality and low-cost technical documents are achieved, and patent applications are improved. success rate and document accuracy.
Patent Information
- Application Number
- CN202411131070.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-16
- Publication Date
- 2024-11-22
AI Technical Summary
The writing of traditional technical briefings relies on manual labor, is inefficient, has high professional requirements, is prone to omissions or errors, and is costly, making it difficult to deal with complex technology descriptions and logical reasoning. The existing automation auxiliary tools lack intelligence and personalization.
Using an intelligent technical briefing writing assistance system based on large-scale language models (such as GPT-4), analyzing and processing the technical description provided by the inventors, generates technical background, technical problems, invention content, implementation methods, etc. The complete technical briefing text can be generated and modified repeatedly through question and answer methods, and then modified after generating the first draft.
It significantly improves the efficiency and accuracy of technical documents generation, reduces the difficulty and cost of writing, improves the success rate and quality of patent applications, and reduces the dependence on professionals.
Smart Images

Figure CN119003733A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of natural language processing technology, and in particular to a method, device, equipment, medium and product for generating technical documents. Background Art
[0002] As a technical document, the technical briefing is a very important part of the patent application process. It describes in detail the technical content, background, innovation and implementation methods of the invention. The writing of traditional technical briefings usually relies on manual work, which requires the inventor or patent agent to have a deep technical background and patent writing experience. However, writing high-quality technical briefings is a time-consuming and laborious task, especially when facing complex technical fields, it is easy to omit information or express inaccurately.
[0003] With the rapid development of artificial intelligence technology, especially the application of large-scale language models, more and more automated auxiliary tools have been introduced into the field of text generation and processing. These tools can reduce manual workload and potential errors, but such tools are usually based on simple templates or rules, lack intelligence and personalization, and are difficult to deal with complex technical descriptions and logical reasoning, and are still highly dependent on manual work.
[0004] Therefore, it is necessary to propose a solution to improve the writing quality and efficiency of technical documents.
[0005] The above contents are only used to assist in understanding the technical solution of the present application and do not constitute an admission that the above contents are prior art. Summary of the invention
[0006] The main purpose of this application is to provide a method, device, equipment, medium and product for generating technical documents, aiming to improve the writing quality and writing efficiency of technical documents.
[0007] To achieve the above object, the present application provides a method for generating a technical document, the method comprising:
[0008] Get user input;
[0009] Performing a full search based on the user input content to determine a first preset number of target reference contents, and analyzing the target reference contents to generate a technical content summary;
[0010] Based on the pre-trained large language model, determine the target technical document name, target technical document summary and target technical document full text according to the user input content and the technical content summary;
[0011] Generate a target technical document according to the target technical document name, the target technical document summary and the target technical document full text.
[0012] In one embodiment, the step of determining the target technical document name, the target technical document abstract and the target technical document full text according to the user input content and the technical content summary based on the pre-trained large language model includes:
[0013] Inputting the user input content and the technical content summary into the large language model to generate a second preset number of technical document names;
[0014] In response to receiving a first operation instruction from a user, selecting and / or modifying the second preset number of technical document names to obtain the target technical document name;
[0015] Inputting the user input content, the technical content summary and the target technical document name into the large language model to generate an initial technical document summary;
[0016] In response to receiving a second operation instruction from the user, regenerating or modifying the initial technical document abstract to obtain the target technical document abstract;
[0017] Inputting the user input content, the technical content summary and the target technical document summary into the large language model to generate the full text of the initial technical document;
[0018] In response to receiving the third operation instruction from the user, the full text of the initial technical document is regenerated or modified to obtain the full text of the target technical document.
[0019] In one embodiment, before the step of determining the target technical document name, the target technical document abstract and the target technical document full text according to the user input content and the technical content summary based on the pre-trained large language model, the following steps are further included:
[0020] Acquire target sample data, wherein the target sample data includes a training data set and a test data set;
[0021] Selecting a pre-trained model, and performing supervised learning training on the pre-trained model according to the training data set to obtain a trained model;
[0022] The trained model is performance tested according to the test data set to obtain a test result, and the trained model is iteratively optimized according to the test result to obtain the large language model.
[0023] In one embodiment, the step of obtaining target sample data includes:
[0024] Collecting original sample data, wherein the original sample data includes patent document data, technical literature data and historical input data;
[0025] Performing data cleaning on the patent document data, technical literature data, and historical input data to obtain cleaned sample data;
[0026] Performing data labeling on the cleaned sample data to obtain sample data labeling information;
[0027] The cleaned sample data and sample data annotation information are used as the target sample data.
[0028] In one embodiment, the patent document data includes patent application documents, examination opinion documents, authorized patent documents and related legal documents, and the step of performing data cleaning on the patent document data, technical literature data and historical input data to obtain cleaned sample data includes:
[0029] The irrelevant content, duplicate data and noise data in the patent application document, examination opinion document, authorized patent document, relevant legal documents, technical literature data and historical input data are removed to obtain the cleaned sample data.
[0030] In one embodiment, the step of performing data labeling on the cleaned sample data to obtain sample data labeling information includes:
[0031] Segmenting the cleaned sample data to obtain text processing units;
[0032] Identify the document structure to which each text processing unit belongs, wherein the document structure includes at least one of technical background, technical content, claims, technical field, key technical points and innovative points;
[0033] According to the document structure to which each text processing unit belongs, each text processing unit is labeled to obtain the sample data labeling information.
[0034] In one embodiment, the user input content includes a technical field, keywords, and creative description, and the step of performing a full search based on the user input content to determine a first preset number of target reference contents includes:
[0035] Extracting text content from the technical field, keywords, and creative description, and performing text preprocessing and vectorization processing on the text content to obtain vectorized text content;
[0036] Searching a preset patent database according to the vectorized text content, and calculating a similarity score between each patent document in the preset patent database and the vectorized text content;
[0037] The patent documents are sorted according to the similarity scores, and a first preset number of patent documents are selected as the target reference content.
[0038] In addition, to achieve the above-mentioned purpose, the present application also proposes a technical document generation device, the technical document generation device comprising:
[0039] The acquisition module is used to obtain user input content;
[0040] A retrieval module, configured to perform a full search based on the user input content, determine a first preset number of target reference contents, analyze the target reference contents, and generate a technical content summary;
[0041] A determination module, configured to determine the name of a target technical document, a target technical document summary, and a target technical document full text according to the user input content and the technical content summary based on a pre-trained large language model;
[0042] A generating module is used to generate a target technical document according to the target technical document name, the target technical document summary and the target technical document full text.
[0043] In addition, to achieve the above-mentioned purpose, the present application also proposes a technical document generation device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, and the computer program is configured to implement the steps of the technical document generation method described above.
[0044] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the technical document generation method described above are implemented.
[0045] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, the steps of the technical document generation method described above are implemented.
[0046] One or more technical solutions proposed in this application have at least the following technical effects:
[0047] By obtaining user input content; performing a full search based on the user input content, determining a first preset number of target reference content, and analyzing the target reference content to generate a technical content summary; based on a pre-trained large language model, determining the target technical document name, target technical document abstract and target technical document full text according to the user input content and the technical content summary; generating a target technical document based on the target technical document name, target technical document abstract and target technical document full text, by obtaining user input content, the content provided by the user can be analyzed and processed to fully understand the user's intentions and information provided; by performing a full search based on the user input content, By determining a first preset number of target reference contents, several reference documents that are most relevant to the user input contents can be determined from a massive database, and then the target reference contents are analyzed through a large language model to generate a technical content summary, and then the target technical document name, target technical document abstract and target technical document full text are determined according to the user input contents and the technical content summary, so as to generate a target technical document that meets the user's needs. In this process, the user only needs to perform simple input, selection or modification to quickly generate a target technical document with complete content that accurately expresses the user's intentions with the help of a large language model, which reduces the difficulty of users writing technical documents and improves the efficiency and quality of generating technical documents. BRIEF DESCRIPTION OF THE DRAWINGS
[0048] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.
[0049] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.
[0050] Figure 1 It is a flowchart of a method for generating a technical document according to a first embodiment;
[0051] Figure 2 is a flow chart of a method for generating a technical document according to a second embodiment;
[0052] Figure 3 is a schematic diagram of a first interface according to a second embodiment of the present application;
[0053] Figure 4 is a schematic diagram of a second interface according to a second embodiment of the present application;
[0054] Figure 5is a schematic diagram of a third interface according to the second embodiment of the present application;
[0055] Figure 6 is a schematic diagram of a fourth interface according to the second embodiment of the present application;
[0056] Figure 7 is a schematic diagram of a fifth interface according to the second embodiment of the present application;
[0057] Figure 8 is a flow chart of a method for generating a technical document according to a third embodiment;
[0058] Fig. 9 is a flowchart of a method for generating a technical document according to a fourth embodiment;
[0059] Fig.10 is a schematic diagram of an overall process according to a fourth embodiment;
[0060] Fig.11 is a flowchart of a method for generating a technical document according to a fifth embodiment;
[0061] Fig.12 This is a schematic diagram of the module structure of the technical document generating device according to the embodiment of the present application;
[0062] Fig.13 A schematic diagram of the device structure of the hardware operating environment involved in the technical document generation method in the embodiment of the present application.
[0063] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION
[0064] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.
[0065] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.
[0066] The main solution of the embodiment of the present application is: by obtaining user input content; performing a full search based on the user input content, determining a first preset number of target reference content, and analyzing the target reference content to generate a technical content summary; based on a pre-trained large language model, determining the target technical document name, target technical document abstract and target technical document full text according to the user input content and the technical content summary; generating a target technical document based on the target technical document name, target technical document abstract and target technical document full text, by obtaining user input content, the content provided by the user can be analyzed and processed to fully understand the user's intentions and the information provided; by The target technical document can be generated by searching the entire content of the database and determining a first preset number of target reference contents. The target reference contents can be determined from a massive database and then analyzed through a large language model to generate a technical content summary. The target technical document name, target technical document abstract and target technical document full text can be determined based on the user input contents and the technical content summary, thereby generating a target technical document that meets the needs of the user. In this process, the user only needs to perform simple input, selection or modification to quickly generate a target technical document with complete content that accurately expresses the user's intention with the help of a large language model, thereby reducing the difficulty of users in writing technical documents and improving the efficiency and quality of generating technical documents.
[0067] As a technical document, the technical briefing is a very important part of the patent application process. It describes in detail the technical content, background, innovation and implementation methods of the invention. The traditional writing of technical briefings usually relies on manual work, which requires the inventor or patent agent to have a deep technical background and patent writing experience. However, writing high-quality technical briefings is a time-consuming and laborious task. Especially when facing complex technical fields, it is easy to omit information or express inaccurately. Specifically, the existing technical briefing writing process has the following technical problems:
[0068] 1. Low efficiency: The traditional writing of technical briefing documents relies on manual work and takes a long time, especially in complex technical fields, which requires a lot of time for literature search, technical analysis and text writing;
[0069] 2. High professional requirements: Writing high-quality technical briefing documents requires the author to have a deep technical background and patent writing experience. Ordinary inventors find it difficult to complete it independently, or are unwilling to write it themselves and rely on professional patent agents;
[0070] 3. Information omission and error: Information omission or inaccurate expression may occur easily during manual writing, which may affect the quality and success rate of patent applications;
[0071] 4. High cost: Due to the need to rely on professionals, the cost of writing traditional technical briefing documents is relatively high, which is a heavy burden especially for small businesses and individual inventors.
[0072] With the rapid development of artificial intelligence technology, especially the application of large-scale language models, more and more automated auxiliary tools have been introduced into the field of text generation and processing. These tools can reduce manual workload and potential errors, but such tools are usually based on simple templates or rules, lack intelligence and personalization, and are difficult to deal with complex technical descriptions and logical reasoning. They are still highly dependent on manual work. Specifically, there are some automated auxiliary tools for assisting patent writing and text generation, but there are still certain limitations in using large-scale language models to assist in the writing of technical briefing documents, including:
[0073] 1. Text generation tools: Some existing text generation tools can help write certain parts of patent documents, such as title generation and abstract writing, but these tools are usually based on simple templates or rules, lack intelligence and personalization, and are difficult to handle complex technical descriptions and logical reasoning;
[0074] 2. Patent management software: Some patent management software provides document retrieval and management functions, which can help inventors organize and manage patent documents, but they still need to rely on manual work when actually writing the briefing documents;
[0075] 3. Machine translation and text proofreading tools: These tools can help process multilingual patent documents and perform text proofreading. However, in the specific writing process of the technical briefing document, it is still necessary to rely on professionals to accurately express the technical content.
[0076] This application provides a solution. By using a large-scale language model (such as GPT-4, etc.), an intelligent technical briefing book writing assistance system is provided, which can automatically generate high-quality technical briefing books, improve writing efficiency and accuracy, reduce costs, and solve problems existing in the prior art. The system can analyze and process the technical description provided by the inventor to generate a complete technical briefing book text including technical background, technical problems, invention content, implementation methods, etc., significantly improving the success rate and quality of patent applications. In addition, the inventor can use the large model to repeatedly generate relevant content and modified content of the briefing book (including model regeneration + manual modification) through continuous question-and-answer methods, and finally complete the writing of the briefing book. The model generates a first draft, and the inventor only needs to complete the modification of the briefing book, which greatly improves the efficiency and quality of the briefing book generation, not only solving the difficulty of the inventor in writing the briefing book, but also saving individuals and enterprises the cost of paying others to write, reducing costs and increasing efficiency for enterprises.
[0077] It should be noted that the execution subject of this embodiment may be a computing service device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or an electronic device capable of realizing the above functions, a technical document generation device, etc. The following takes the technical document generation device as an example to illustrate this embodiment and the following embodiments.
[0078] Based on this, the present application embodiment provides a method for generating a technical document, referring to Figure 1 , Figure 1 It is a flowchart of a method for generating a technical document according to a first embodiment.
[0079] In this embodiment, the technical document generation method includes steps S10 to S40:
[0080] Step S10, obtaining user input content;
[0081] In the embodiment of the present application, a writing assistance system is provided. First, the user input content is received through the front-end interface of the writing assistance system, which is the basis of the entire technical document generation process. Among them, the front-end interface includes a display interface, an input box, a check box, a drop-down box, a display box, and a progress identifier, etc. The display interface can be used to display the title, summary, and content text generated by the system; the input box can be used to display the content input by the user; the check box can display the content selected by the user according to the user's operation; the drop-down box can fold and display different selection controls and / or text content, for example, the user can quickly select the relevant technical field by clicking the drop-down box, and the options displayed in the drop-down box can be arranged according to the preset priority, so that the user can quickly find the option to be selected; the progress identifier can display the full process nodes, such as the title, summary, and content nodes, and distinguish the nodes corresponding to the current page or the completed nodes to let the user know the progress. The user input content usually includes technical fields, keywords, and creative descriptions. In order to more accurately obtain the user's intention, the system may need to clarify and refine the user's input content through interactive dialogue. For example, the system can ask the user about the specific application scenarios of the technology, the problems solved by the technology, or specific technical requirements, so as to further perform a full search based on the user input content.
[0082] Step S20, performing a full search according to the user input content, determining a first preset number of target reference contents, and analyzing the target reference contents to generate a technical content summary;
[0083] Furthermore, after obtaining the user input content, a full search can be performed based on the user input content to determine a first preset number of target reference contents, and the target reference contents are analyzed to generate a technical content summary.
[0084] Optionally, after obtaining the user input content through the front-end interface, the user input content can be processed by the back-end, specifically by performing data analysis on the user input content, searching based on the analyzed data content, and further analyzing the search results, so as to generate relevant information, such as a technical content summary, the name of the target technical document, the target technical document abstract, and the full text of the target technical document, and through information feedback, present the relevant information generated by the back-end processing to the front-end interface for users to view, select, input, modify and output.
[0085] Optionally, the writing assistance system performs a full search based on the user input to determine the references such as patents and technical documents that are most relevant to the user's needs. The purpose of this step is to find the most valuable references to support the generation of technical documents. The search results will be analyzed and summarized into a technical content summary, which will help the system understand the technical background and limitations of the prior art.
[0086] Optionally, the method for generating a technical briefing document is used as an example in the embodiments of the present application to illustrate, so the database used for retrieval is a patent database. In other embodiments, other databases may also be used for retrieval, such as a paper database and online information. Similarly, the target reference content obtained may be a patent document, or may be literature content and other Internet information, which is not specifically limited in the embodiments of the present application.
[0087] Optionally, a full search is performed based on the user input content to determine the first preset number of target reference contents, which mainly includes steps such as text preprocessing, vectorization processing, similarity calculation, and document sorting and selection. Specifically, text preprocessing includes preprocessing steps such as cleaning, word segmentation, and removal of stop words for user questions and user documents; vectorization processing includes converting text into vector form using models such as TF-IDF, Word2Vec, or BERT; similarity calculation includes searching for documents with the highest similarity to the vectorized text content in a preset patent database; document sorting and selection includes sorting the search results according to the similarity score and selecting the top N documents with the highest scores.
[0088] Step S30, based on the pre-trained large language model, determine the target technical document name, target technical document abstract and target technical document full text according to the user input content and the technical content summary;
[0089] Furthermore, a full search is performed based on the user input content to determine a first preset number of target reference contents, and the target reference contents are analyzed to generate a technical content summary. Then, based on the pre-trained large language model, the target technical document name, the target technical document abstract and the full text of the target technical document can be determined according to the user input content and the technical content summary.
[0090] Optionally, a pre-trained large language model is used to combine user input and technical content summaries to determine the key elements of the technical document. This includes the name, summary, and full text of the document. The writing assistance system can generate multiple options for the user to choose from or further customize to ensure that the generated document meets the user's expectations and needs.
[0091] Step S40: generating a target technical document according to the target technical document name, the target technical document summary and the target technical document full text.
[0092] Furthermore, based on the pre-trained large language model, after determining the target technical document name, target technical document summary and target technical document full text according to the user input content and the technical content summary, the target technical document can be generated according to the target technical document name, target technical document summary and target technical document full text.
[0093] Optionally, the writing assistance system will generate the final target technical document based on the determined target technical document name, target technical document summary and target technical document full text. This step may include formatting, chart insertion and document typesetting to ensure the professionalism and readability of the document. In addition, the writing assistance system can also provide version control and document sharing functions to facilitate user collaboration and revision, and users can select or customize the typesetting or format of the exported target technical document. For example, the writing assistance system generates a complete technical document based on the document name selected by the user, the generated summary and the full text content. The document includes a detailed technical background, description of innovative points, embodiments and flow charts. After reviewing the document, the user can further propose modification requirements, and the writing assistance system can modify, polish, proofread or rewrite the full text or partial content of the generated target technical document according to the user's requirements.
[0094] This embodiment adopts the above scheme, specifically by obtaining user input content; performing a full search based on the user input content to determine a first preset number of target reference content, and analyzing the target reference content to generate a technical content summary; based on a pre-trained large language model, determining the target technical document name, target technical document abstract and target technical document full text according to the user input content and the technical content summary; generating a target technical document based on the target technical document name, target technical document abstract and target technical document full text, by obtaining user input content, the content provided by the user can be analyzed and processed to fully understand the user's intention and information provided; by performing a search based on the user input content A full search is performed to determine the first preset number of target reference contents. Several reference documents that are most relevant to the user input contents can be determined from a massive database, and then the target reference contents are analyzed through a large language model to generate a technical content summary. Then, the target technical document name, target technical document abstract and target technical document full text are determined according to the user input contents and the technical content summary, so as to generate the target technical document in accordance with the user's needs. In this process, the user only needs to perform simple input, selection or modification to quickly generate a target technical document with complete content that accurately expresses the user's intentions with the help of a large language model, which reduces the difficulty of users writing technical documents and improves the efficiency and quality of technical document generation.
[0095] Reference Figure 2 , Figure 2 This is a flowchart of a method for generating technical documents according to the second embodiment. This embodiment proposes a second embodiment of the present application based on the aforementioned first embodiment. In the second embodiment of the present application, the same or similar contents as those in the above embodiment can be referred to the above introduction and will not be described in detail later. Figure 2 , step S30 includes steps S301-S306:
[0096] Step S301: inputting the user input content and the technical content summary into the large language model to generate a second preset number of technical document names;
[0097] Reference Figure 3 , Figure 3 is a schematic diagram of a first interface according to the second embodiment of the present application, as shown in Figure 3As shown, the writing assistance system takes the user input content and the technical content summary as input, and uses the large language model to generate a certain number of technical document names. The purpose of this step is to provide one or more suitable titles for the technical document, which should accurately reflect the core content and technical characteristics of the document. Generate technical document names related to user input and technical content summary, and provide multiple options to increase the flexibility of user selection. Users can choose the title that best meets their needs from the generated names, and the generated names can inspire users to think further about the content of the technical document. In the embodiment of the present application, natural language processing technology is used to analyze user input and summary content, extract keywords and concepts, and use the creative writing ability of the large language model to generate attractive and descriptive document names.
[0098] Step S302: in response to receiving a first operation instruction from a user, selecting and / or modifying the second preset number of technical document names to obtain the target technical document name;
[0099] Reference Figure 4 , Figure 4 is a schematic diagram of a second interface according to the second embodiment of the present application, as shown in Figure 4 As shown, the system responds to the user's instructions and allows the user to select or modify one of the generated document names to ensure that the final document name fully meets the user's expectations and needs. Ensuring that the document name accurately reflects the user's intentions and technical features, the user experience can be enhanced by providing user customization options. In the embodiment of the present application, a user interface is provided to allow the user to select or edit the document name, and the writing assistance system can further optimize the name generation algorithm based on the user's modification feedback.
[0100] Step S303: inputting the user input content, the technical content summary and the target technical document name into the large language model to generate an initial technical document summary;
[0101] Reference Figure 5 , Figure 5 is a schematic diagram of a third interface according to the second embodiment of the present application, as shown in Figure 5 As shown in Figure 1, the writing assistance system uses user input, technical content summary, and selected document name as input to generate an initial summary of the technical document. The summary should briefly and clearly summarize the main content and technical innovations of the document. By analyzing the document name and technical content summary, key information is extracted, and a coherent and logical summary text is generated using a large language model.
[0102] Step S304: in response to receiving the second operation instruction of the user, regenerating or modifying the initial technical document abstract to obtain the target technical document abstract;
[0103] Reference Figure 6 , Figure 6 is a schematic diagram of a fourth interface according to the second embodiment of the present application, as shown in Figure 6 As shown in the figure, the writing assistance system regenerates or modifies the generated summary based on user feedback to ensure the accuracy of the summary content and user satisfaction. Provide a user feedback mechanism to collect user opinions on the summary, adjust the summary content based on the feedback, or regenerate the summary to ensure that the summary fully meets the user's expectations for the technical document.
[0104] Step S305: inputting the user input content, the technical content summary and the target technical document abstract into the large language model to generate the full text of the initial technical document;
[0105] Reference Figure 7 , Figure 7 is a schematic diagram of a fifth interface according to the second embodiment of the present application, as shown in Figure 7 As shown, the full text of the technical document is generated based on user input, technical content summary and abstract. The full text should describe the technical background, problem statement, technical solution, implementation example and potential application scenarios in detail. By combining the abstract and technical content summary, a large language model is used to generate a structured and information-rich full text of the document, ensuring the logic and orderliness of the document content, thereby generating a comprehensive, detailed and complete technical document.
[0106] Step S306: in response to receiving the third operation instruction from the user, regenerate or modify the full text of the initial technical document to obtain the full text of the target technical document.
[0107] Optionally, the generated full text of the document is regenerated or modified based on further feedback from the user, thereby ensuring the final quality of the document content and meeting the specific needs of the user. A detailed user feedback interface is provided in the embodiment of the present application, allowing the user to make modification suggestions for each part of the document. Based on the user's feedback, the writing assistance system can make detailed adjustments or regenerate the document content, improve the quality and depth of the document content based on user feedback, and ensure that each part of the document accurately conveys the user's intentions and technical details.
[0108] Through the above scheme, this embodiment specifically generates a second preset number of technical document names by inputting the user input content and the technical content summary into the large language model; in response to receiving the user's first operation instruction, the second preset number of technical document names are selected and / or modified to obtain the target technical document name; the user input content, the technical content summary and the target technical document name are input into the large language model to generate an initial technical document summary; in response to receiving the user's second operation instruction, the initial technical document summary is regenerated or modified to obtain the target technical document summary; the user input content, the technical content summary and the target technical document summary are input into the large language model to generate the initial technical document full text; in response to receiving the user's third operation instruction, the initial technical document full text is regenerated or modified to obtain the target technical document full text, which can ensure that each part of the generated target technical document accurately conveys the user's intention and provides comprehensive, complete and accurate technical details.
[0109] Reference Figure 8 , Figure 8 This is a flowchart of a method for generating technical documents according to the third embodiment. This embodiment proposes a third embodiment of the present application based on any of the above embodiments. In the third embodiment of the present application, the same or similar contents as those in the above embodiments can be referred to the above introduction and will not be described in detail later. Figure 8 , before step S30, steps S01-S03 are also included:
[0110] Step S01: Acquire target sample data, wherein the target sample data includes a training data set and a test data set;
[0111] Step S02: selecting a pre-trained model, and performing supervised learning training on the pre-trained model according to the training data set to obtain a trained model;
[0112] Step S03: Perform a performance test on the trained model according to the test data set to obtain a test result, and iteratively optimize the trained model according to the test result to obtain the large language model.
[0113] Optionally, the step of acquiring target sample data includes:
[0114] Collecting original sample data, wherein the original sample data includes patent document data, technical literature data and historical input data;
[0115] Performing data cleaning on the patent document data, technical literature data, and historical input data to obtain cleaned sample data;
[0116] Performing data labeling on the cleaned sample data to obtain sample data labeling information;
[0117] The cleaned sample data and sample data annotation information are used as the target sample data.
[0118] Optionally, the patent document data includes patent application documents, examination opinion documents, authorized patent documents and related legal documents.
[0119] Optionally, in the embodiment of the present application, patent document data, technical literature data and user input data are used to fine-tune the pre-trained model. Specifically, the patent document data can be obtained from a public patent database, and a large number of patent documents are collected, including patent applications, examination opinions, authorized patents and related legal documents. By collecting technical literature, research papers and industry reports related to patents, the technical background knowledge of the model can be enriched. By collecting user input data and feedback when using patent writing tools, user needs and common problems can be understood.
[0120] Optionally, the step of performing data cleaning on the patent document data, technical literature data and historical input data to obtain cleaned sample data includes:
[0121] The irrelevant content, duplicate data and noise data in the patent application document, examination opinion document, authorized patent document, relevant legal documents, technical literature data and historical input data are removed to obtain the cleaned sample data.
[0122] Optionally, in the embodiment of the present application, irrelevant content, duplicate data and noise data are removed through data cleaning to ensure data quality.
[0123] Optionally, the step of performing data labeling on the cleaned sample data to obtain sample data labeling information includes:
[0124] Segmenting the cleaned sample data to obtain text processing units;
[0125] Identify the document structure to which each text processing unit belongs, wherein the document structure includes at least one of technical background, technical content, claims, technical field, key technical points and innovative points;
[0126] According to the document structure to which each text processing unit belongs, each text processing unit is labeled to obtain the sample data labeling information.
[0127] Optionally, the document is segmented into processable units, such as sentences or paragraphs. The data is annotated, including annotating the structure of the patent document (such as background technology, invention content, claims), technical fields, key technical points and innovation points.
[0128] Optionally, the pre-trained model selected in the embodiment of the present application is a pre-trained model suitable for text processing, such as BERT, GPT or XLNet. These models have been trained on a large amount of text and have good language understanding ability. According to the characteristics of the patent document, the model architecture may need to be adjusted, for example, increasing or modifying the number of layers of the model, adjusting the attention mechanism, etc. The model is trained using a labeled data set, and through supervised learning, the model learns how to identify and process professional terms and expressions in patent documents. During the training process, the validation set can be used to monitor the performance of the model and adjust hyperparameters such as learning rate, batch size, etc. to optimize the model performance. The training model writes patent documents, including generating claims, technical field descriptions, etc. The training model parses patent documents, including extracting key information, understanding document structure, etc. The performance of the model is evaluated on an independent test set. Indicators such as accuracy, recall, and F1 score are used to measure the writing and parsing capabilities of the model. According to the test results, the model is iteratively improved, and the data collection or model architecture adjustment step can be returned to further improve performance. Then the trained model is deployed to practical applications, the performance of the model in actual use is monitored, and adjustments are made based on feedback.
[0129] This embodiment adopts the above scheme, specifically by acquiring target sample data, wherein the target sample data includes a training data set and a test data set; selecting a pre-trained model, and performing supervised learning training on the pre-trained model according to the training data set to obtain a trained model; performing a performance test on the trained model according to the test data set to obtain a test result, and iteratively optimizing the trained model according to the test result to obtain the large language model, and further fine-tuning the pre-trained model with good language comprehension ability by using data such as patent documents and related technical literature to obtain a large language model suitable for generating technical documents, which is applied to an intelligent writing assistance system, and can automatically generate high-quality technical documents, improve writing efficiency and accuracy, and reduce labor costs.
[0130] Reference Fig. 9 , Fig. 9 This is a flowchart of a method for generating technical documents according to the fourth embodiment. This embodiment is based on any of the above embodiments and proposes the fourth embodiment of this application. In the fourth embodiment of this application, the same or similar contents as any of the above embodiments can be referred to the above introduction and will not be repeated later. Fig. 9 , step S20 includes steps S201-S203:
[0131] Step S201: extracting text content from the technical field, keywords, and creative description, and performing text preprocessing and vectorization processing on the text content to obtain vectorized text content;
[0132] Reference Fig.10 , Fig.10 FIG. 4 is a schematic diagram of an overall process according to a fourth embodiment. Fig.10 As shown in the figure, extracting text content and performing vectorization processing is the basis of the entire document generation process. It involves extracting text content from user input content including technical fields, keywords and creative descriptions, and preprocessing and vectorizing it so that subsequent steps can be more effectively retrieved and analyzed. Its purpose is to convert unstructured text content into a structured data form for machine processing and analysis. Through vectorization processing, the accuracy and efficiency of text similarity calculation are improved. Through preprocessing and vectorization processing, a standardized text representation can be obtained, which lays the foundation for subsequent retrieval and similarity calculation, while ensuring that the semantic information of the text content is effectively retained and expressed.
[0133] Optionally, extracting user input content includes extracting plain text information from the user input content; performing text preprocessing on the text content includes removing noise (such as HTML tags), word segmentation, removing stop words, stemming, etc., and vectorization processing includes applying NLP technology, such as TF-IDF, Word2Vec or BERT model, to convert text into a numerical vector.
[0134] Step S202: searching a preset patent database according to the vectorized text content, and calculating a similarity score between each patent document in the preset patent database and the vectorized text content;
[0135] Furthermore, based on the vectorized text, the preset patent database is searched and the similarity score with the vectorized text content is calculated, thereby quickly locating the patent document most relevant to the user input content, providing a basis for the subsequent selection of target reference content, and obtaining a list of patent documents that are highly relevant to user needs. The correlation between documents can be quantified through the similarity score.
[0136] Optionally, the vectorized text content is used as a query to search the patent database, and for each retrieved patent document, its similarity score with the vectorized text content is calculated. In this process, methods such as cosine similarity and Euclidean distance can be used to measure the similarity.
[0137] Step S203: Sort the patent documents according to the similarity scores, and select the first first preset number of patent documents as the target reference content.
[0138] Furthermore, the search results are sorted, and the most relevant patent documents are selected as target reference content based on the similarity score, so as to obtain a set of the most relevant patent documents as the basis for generating technical documents; through sorting, users can quickly understand the relevance level of the documents, thereby ensuring that the selected documents are highly relevant to user needs and providing support for generating high-quality technical documents. By setting a limit on the number of reference documents, the number of reference documents can be controlled for easy subsequent processing.
[0139] Optionally, the search results are sorted in descending order according to the similarity score, and the top N patent documents with the highest scores are selected according to a preset number, and patent documents from different angles or aspects are selected considering the diversity and coverage of the documents. It should be noted that the preset number N can be set by the user according to actual conditions, or can be automatically determined by the writing assistance system according to experience values, and this is not specifically limited in the embodiments of the present application.
[0140] This embodiment adopts the above scheme, specifically by extracting the text content in the technical field, keywords and creative description, and performing text preprocessing and vectorization processing on the text content to obtain vectorized text content, converting unstructured text content into structured data form, which is convenient for machine processing and analysis, and improving the accuracy and efficiency of text similarity calculation through vectorization processing; by searching a preset patent database according to the vectorized text content, calculating the similarity score between each patent document in the preset patent database and the vectorized text content, obtaining a list of patent documents highly relevant to user needs, and quantifying the correlation between documents through similarity scores; by sorting the patent documents according to the similarity scores, and selecting the first preset number of patent documents as the target reference content, it is ensured that the selected documents are highly relevant to user needs, providing support for generating high-quality technical documents, and controlling the number of reference documents through preset number restrictions, which is convenient for subsequent processing.
[0141] Reference Fig.11 , Fig.11 This is a flowchart of a method for generating technical documents according to the fifth embodiment. This embodiment is based on any of the above embodiments and proposes the fifth embodiment of this application. In the fifth embodiment of this application, the same or similar contents as any of the above embodiments can be referred to the above introduction and will not be repeated later. Fig.11 In the application scenario of using the writing assistance system to generate a technical briefing document, the method for generating the technical briefing document includes:
[0142] Step 1: Automatically generate title:
[0143] Specifically, the large model generates a title based on the user's input content. The input is: user input {query}, patent search content {patent search}, and the output is: patent title.
[0144] The specific implementation logic includes:
[0145] 1) All user input (including technical fields, keywords, creative descriptions, etc.) is used as the search input {query}, and a full search is performed in the patent database, and the top N semantically related patents are retrieved (the most relevant N patents are taken as a reference for the large model to write the patent briefing document).
[0146] 2) After retrieving the top N patents, the big model automatically interprets and analyzes the N patents, learns and understands the contents of the N patents, and summarizes them into a summary of the patent search content {patent search};
[0147] 3) The user input {query} and the patent search content {patent search} are used as the input of the big model to generate several titles of patent briefing documents;
[0148] 4) Title selection: If the user is not satisfied with the title generated by the large model, he / she can regenerate the title, or edit and modify the title, and finally select a title to proceed to the next step.
[0149] Step 2: Automatically generate patent abstract
[0150] Specifically, the big model generates a summary based on the user's input content. The input is: user input {query}, patent search content {patent search}, user-selected title {title}, and the output is: patent briefing summary.
[0151] The specific implementation logic includes:
[0152] 1) The user input {query}, patent search content {patent search} and patent title {title} are used as the input of the big model to generate the summary of the patent briefing document;
[0153] 2) Modification of the summary: Supports regeneration of summary content (users can enter their own requirements to regenerate the summary); or users can manually edit and modify the summary content.
[0154] Step 3: Automatically generate the full text of the patent briefing
[0155] Specifically, the big model generates an abstract based on the user's input content. The input is: user input {query}, patent search content {patent search}, patent abstract {abstract}, and the output is: the full text of the patent briefing document.
[0156] The specific implementation logic includes:
[0157] 1) The user's input {query}, patent search content {patent search} and patent abstract {abstract} are used as the input of the big model to generate the full text of the patent briefing, including the patent technology field, patent technology background, patent technology problems, invention steps, core advantages and flow chart generation;
[0158] 2) Modification of the full text of the briefing book:
[0159] The embodiment of the present application supports the regeneration of the full text content, and supports the rewriting / proofreading / continuation / polishing of partial content of the briefing book. For example, to regenerate the "invention steps", you can enter the user requirement (prompt), and after clicking regenerate, the "invention steps" will be rewritten; manual editing and modification of the full text content is supported.
[0160] This embodiment, through the above scheme, specifically by using a large-scale language model (such as GPT-4, etc.), provides an intelligent patent briefing book writing assistance system, which can automatically generate high-quality patent briefing books, improve writing efficiency and accuracy, reduce costs, and solve problems existing in the prior art. The system can analyze and process the technical description provided by the inventor to generate a complete patent briefing book text including technical background, technical problems, invention content, implementation methods, etc., significantly improving the success rate and quality of patent applications. In addition, the user can make repeated modifications (including model regeneration + manual modification) based on the first draft of the briefing book generated by the model, and finally complete the writing of the briefing book. The first draft is generated by the model, and the user only needs to complete the modification of the briefing book, thereby greatly improving the efficiency and quality of the briefing book generation, which not only solves the difficulty of users in writing briefing books, but also saves individuals and enterprises the cost of paying others to write, thereby reducing costs and increasing efficiency for enterprises.
[0161] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the method for generating technical documents of the present application. More simple transformations based on this technical concept are all within the scope of protection of the present application.
[0162] This application also provides a technical document generation device, please refer to Fig.12 , the technical document generating device comprises:
[0163] The acquisition module is used to obtain user input content;
[0164] A retrieval module, configured to perform a full search based on the user input content, determine a first preset number of target reference contents, analyze the target reference contents, and generate a technical content summary;
[0165] A determination module, configured to determine the name of a target technical document, a target technical document summary, and a target technical document full text according to the user input content and the technical content summary based on a pre-trained large language model;
[0166] A generating module is used to generate a target technical document according to the target technical document name, the target technical document summary and the target technical document full text.
[0167] In one embodiment, the step of determining the target technical document name, the target technical document abstract and the target technical document full text according to the user input content and the technical content summary based on the pre-trained large language model includes:
[0168] Inputting the user input content and the technical content summary into the large language model to generate a second preset number of technical document names;
[0169] In response to receiving a first operation instruction from a user, selecting and / or modifying the second preset number of technical document names to obtain the target technical document name;
[0170] Inputting the user input content, the technical content summary and the target technical document name into the large language model to generate an initial technical document summary;
[0171] In response to receiving a second operation instruction from the user, regenerating or modifying the initial technical document abstract to obtain the target technical document abstract;
[0172] Inputting the user input content, the technical content summary and the target technical document summary into the large language model to generate the full text of the initial technical document;
[0173] In response to receiving the third operation instruction from the user, the full text of the initial technical document is regenerated or modified to obtain the full text of the target technical document.
[0174] In one embodiment, before the step of determining the target technical document name, the target technical document abstract and the target technical document full text according to the user input content and the technical content summary based on the pre-trained large language model, the following steps are further included:
[0175] Acquire target sample data, wherein the target sample data includes a training data set and a test data set;
[0176] Selecting a pre-trained model, and performing supervised learning training on the pre-trained model according to the training data set to obtain a trained model;
[0177] The trained model is performance tested according to the test data set to obtain a test result, and the trained model is iteratively optimized according to the test result to obtain the large language model.
[0178] In one embodiment, the step of obtaining target sample data includes:
[0179] Collecting original sample data, wherein the original sample data includes patent document data, technical literature data and historical input data;
[0180] Performing data cleaning on the patent document data, technical literature data, and historical input data to obtain cleaned sample data;
[0181] Performing data labeling on the cleaned sample data to obtain sample data labeling information;
[0182] The cleaned sample data and sample data annotation information are used as the target sample data.
[0183] In one embodiment, the patent document data includes patent application documents, examination opinion documents, authorized patent documents and related legal documents, and the step of performing data cleaning on the patent document data, technical literature data and historical input data to obtain cleaned sample data includes:
[0184] The irrelevant content, duplicate data and noise data in the patent application document, examination opinion document, authorized patent document, relevant legal documents, technical literature data and historical input data are removed to obtain the cleaned sample data.
[0185] In one embodiment, the step of performing data labeling on the cleaned sample data to obtain sample data labeling information includes:
[0186] Segmenting the cleaned sample data to obtain text processing units;
[0187] Identify the document structure to which each text processing unit belongs, wherein the document structure includes at least one of technical background, technical content, claims, technical field, key technical points and innovative points;
[0188] According to the document structure to which each text processing unit belongs, each text processing unit is labeled to obtain the sample data labeling information.
[0189] In one embodiment, the user input content includes a technical field, keywords, and creative description, and the step of performing a full search based on the user input content to determine a first preset number of target reference contents includes:
[0190] Extracting text content from the technical field, keywords, and creative description, and performing text preprocessing and vectorization processing on the text content to obtain vectorized text content;
[0191] Searching a preset patent database according to the vectorized text content, and calculating a similarity score between each patent document in the preset patent database and the vectorized text content;
[0192] The patent documents are sorted according to the similarity scores, and a first preset number of patent documents are selected as the target reference content.
[0193] The technical document generation device provided by the present application adopts the technical document generation method in the above embodiment, which can solve the technical problem of technical document generation. Compared with the prior art, the beneficial effects of the technical document generation device provided by the present application are the same as the beneficial effects of the technical document generation method provided by the above embodiment, and other technical features in the technical document generation device are the same as the features disclosed in the above embodiment method, which will not be repeated here.
[0194] The present application provides a technical document generation device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the technical document generation method in the above-mentioned embodiment one.
[0195] Reference below Fig.13 , which shows a schematic diagram of the structure of a technical document generation device suitable for implementing the embodiment of the present application. The technical document generation device in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Fig.13 The technical document generating device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.
[0196] like Fig.13 As shown, the technical document generation device may include a processing device 1001 (e.g., a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM: Read Only Memory) 1002 or a program loaded from a storage device 1003 to a random access memory (RAM: Random Access Memory) 1004. In RAM1004, various programs and data required for the operation of the technical document generation device are also stored. The processing device 1001, ROM1002, and RAM1004 are connected to each other through a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Generally, the following systems can be connected to the I / O interface 1006: an input device 1007 including, for example, a touch screen, a touch pad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output device 1008 including, for example, a liquid crystal display (LCD: Liquid Crystal Display), a speaker, a vibrator, etc.; a storage device 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1009. The communication device 1009 can allow the technical document generation device to communicate with other devices wirelessly or by wire to exchange data. Although the figure shows a technical document generation device with various systems, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems can be implemented or provided instead.
[0197] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.
[0198] The technical document generation device provided by the present application adopts the technical document generation method in the above embodiment, which can solve the technical problem of technical document generation. Compared with the prior art, the beneficial effects of the technical document generation device provided by the present application are the same as the beneficial effects of the technical document generation method provided by the above embodiment, and other technical features in the technical document generation device are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.
[0199] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.
[0200] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.
[0201] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, computer programs) stored thereon, and the computer-readable program instructions are used to execute the technical document generating method in the above-mentioned embodiment.
[0202] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.
[0203] The computer-readable storage medium may be included in the technical document generation device; or may exist independently without being assembled into the technical document generation device.
[0204] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the technical document generating device, the technical document generating device: obtains user input content; performs a full search based on the user input content, determines a first preset number of target reference content, and analyzes the target reference content to generate a technical content summary; based on a pre-trained large language model, determines the target technical document name, target technical document abstract and target technical document full text according to the user input content and the technical content summary; generates a target technical document based on the target technical document name, target technical document abstract and target technical document full text. By obtaining the user input content, the content provided by the user can be analyzed and processed to fully understand the user. The user's intention and information provided; by performing a full search based on the user input content, a first preset number of target reference contents can be determined, and several reference documents most relevant to the user input content can be determined from a massive database, and then the target reference content is analyzed through a large language model to generate a technical content summary, and then the target technical document name, target technical document abstract and target technical document full text are determined based on the user input content and the technical content summary, so as to generate a target technical document that meets the user's needs. In this process, the user only needs to perform simple input, selection or modification, and the large language model can be used to quickly generate a target technical document with complete content that accurately expresses the user's intention, which reduces the difficulty of users writing technical documents and improves the efficiency and quality of technical document generation.
[0205] Computer program code for performing the operations of the present application may be written in one or more programming languages or a combination thereof, including object-oriented programming languages such as Java, Smalltalk, C++, and conventional procedural programming languages such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).
[0206] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.
[0207] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.
[0208] The readable storage medium provided in the present application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned method for generating technical documents, and can solve the technical problem of generating technical documents. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in the present application are the same as the beneficial effects of the method for generating technical documents provided in the above-mentioned embodiment, and will not be described in detail here.
[0209] The present application also provides a computer program product, including a computer program, which implements the steps of the above-mentioned technical document generation method when executed by a processor.
[0210] The computer program product provided by this application can solve the technical problem of generating technical documents. Compared with the prior art, the beneficial effects of the computer program product provided by this application are the same as the beneficial effects of the technical document generating method provided by the above embodiment, which will not be repeated here.
[0211] The above descriptions are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect applications in other related technical fields are included in the patent protection scope of the present application.
Claims
1. A method for generating a technical document, characterized in that: The method includes: Get user input; Performing a full search based on the user input content to determine a first preset number of target reference contents, and analyzing the target reference contents to generate a technical content summary; Based on the pre-trained large language model, determine the target technical document name, target technical document summary and target technical document full text according to the user input content and the technical content summary; Generate a target technical document according to the target technical document name, the target technical document summary and the target technical document full text.
2. The method according to claim 1, characterized in that The step of determining the target technical document name, the target technical document abstract and the target technical document full text according to the user input content and the technical content summary based on the pre-trained large language model comprises: Inputting the user input content and the technical content summary into the large language model to generate a second preset number of technical document names; In response to receiving a first operation instruction from a user, selecting and / or modifying the second preset number of technical document names to obtain the target technical document name; Inputting the user input content, the technical content summary and the target technical document name into the large language model to generate an initial technical document summary; In response to receiving a second operation instruction from the user, regenerating or modifying the initial technical document abstract to obtain the target technical document abstract; Inputting the user input content, the technical content summary and the target technical document summary into the large language model to generate the full text of the initial technical document; In response to receiving the third operation instruction from the user, the full text of the initial technical document is regenerated or modified to obtain the full text of the target technical document.
3. The method according to claim 1, characterized in that Before the step of determining the target technical document name, the target technical document abstract and the target technical document full text according to the user input content and the technical content summary based on the pre-trained large language model, the following steps are further included: Acquire target sample data, wherein the target sample data includes a training data set and a test data set; Selecting a pre-trained model, and performing supervised learning training on the pre-trained model according to the training data set to obtain a trained model; The trained model is performance tested according to the test data set to obtain a test result, and the trained model is iteratively optimized according to the test result to obtain the large language model.
4. The method according to claim 3, characterized in that The step of obtaining target sample data comprises: Collecting original sample data, wherein the original sample data includes patent document data, technical literature data and historical input data; Performing data cleaning on the patent document data, technical literature data, and historical input data to obtain cleaned sample data; Performing data labeling on the cleaned sample data to obtain sample data labeling information; The cleaned sample data and sample data annotation information are used as the target sample data.
5. The method according to claim 4, characterized in that The patent document data includes patent application documents, examination opinion documents, authorized patent documents and related legal documents. The step of performing data cleaning on the patent document data, technical literature data and historical input data to obtain cleaned sample data includes: Removing irrelevant content, duplicate data, and noise data from the patent application document, examination opinion document, authorized patent document, relevant legal documents, technical literature data, and historical input data to obtain the cleaned sample data; The step of labeling the cleaned sample data to obtain sample data labeling information includes: Segmenting the cleaned sample data to obtain text processing units; Identify the document structure to which each text processing unit belongs, wherein the document structure includes at least one of technical background, technical content, claims, technical field, key technical points and innovative points; According to the document structure to which each text processing unit belongs, each text processing unit is labeled to obtain the sample data labeling information.
6. The method according to claim 1, characterized in that The user input content includes a technical field, keywords, and creative description, and the step of performing a full search based on the user input content to determine a first preset number of target reference contents includes: Extracting text content from the technical field, keywords, and creative description, and performing text preprocessing and vectorization processing on the text content to obtain vectorized text content; Searching a preset patent database according to the vectorized text content, and calculating a similarity score between each patent document in the preset patent database and the vectorized text content; The patent documents are sorted according to the similarity scores, and a first preset number of patent documents are selected as the target reference content.
7. A technical document generation device, characterized in that: The device comprises: The acquisition module is used to obtain user input content; A retrieval module, configured to perform a full search based on the user input content, determine a first preset number of target reference contents, analyze the target reference contents, and generate a technical content summary; A determination module, configured to determine the name of a target technical document, a target technical document summary, and a target technical document full text according to the user input content and the technical content summary based on a pre-trained large language model; A generating module is used to generate a target technical document according to the target technical document name, the target technical document summary and the target technical document full text.
8. A technical document generation device, characterized in that: The device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the technical document generation method according to any one of claims 1 to 6.
9. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the technical document generation method according to any one of claims 1 to 6 are implemented.
10. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the steps of the technical document generation method according to any one of claims 1 to 6 are implemented.
Citation Information
Cited By
Dynamic document generation method and system based on LLM (Language Language Model)
CN119577127A
Bridge steel structure technical file intelligent auxiliary generation method and device based on large model
CN119761317A
Document generation method, device and equipment
CN120197597A
Approximate patent retrieval method and device, electronic equipment and storage medium
CN120804294A
Patent document similarity comparison system and method
TWI935909B
