Patent retrieval type intelligent construction method and device and storage medium

Abstract

The invention relates to the technical field of database retrieval, in particular to a patent retrieval type intelligent construction method and device and a storage medium. The method provided by the invention comprises the following steps: acquiring a natural language of a patent queried by a user; inputting the natural language into a large language model for intention understanding to obtain intention information; the intention information at least comprises a retrieval technology intention, a time range intention, a region intention and an application subject intention; converting the intention information into an sql query statement; executing the sql query statement in a pre-established data table to obtain key information; generating structured information according to the key information; inputting the structured information and the mode information of the patent database into a cue word template, and inputting the cue word template into a large language model to obtain a retrieval formula output by the large language model; and performing retrieval in a patent database based on the retrieval formula. According to the method, an accurate retrieval formula is constructed by utilizing a large language model.

Description

Technical Field

[0001] This application relates to the field of database retrieval technology, and more specifically, to a patent retrieval-based intelligent construction method, device, and storage medium. Background Technology

[0002] Patent information retrieval is the cornerstone of patent examination, technological innovation, competitor analysis, and risk avoidance. However, existing patent retrieval systems and methods have long suffered from numerous inherent flaws, significantly limiting the efficiency and quality of retrieval.

[0003] First, the search threshold is too high. Users must be proficient in a complex and unintuitive search syntax, such as Boolean logical operators (AND, OR, NOT), proximity operators (e.g., W / n), and field qualifiers (e.g., TI, AB, CL, IC). This professional syntax system brings a high learning cost, excluding a large number of potential users, such as enterprise R&D personnel and researchers, who lack professional search training.

[0004] Secondly, the retrieval process is cumbersome and inefficient. Constructing an accurate and effective search query is not a simple task, but rather an iterative process requiring repeated trial and error. Users typically need to go through a lengthy cycle of "initial keyword selection → expansion with synonyms and near-synonyms → combination using logical operators → specific field limitation → initial result verification → adjusting the search query based on feedback." This process is highly dependent on manual operation, time-consuming, and labor-intensive, making it difficult to meet the need for quickly obtaining technical information.

[0005] Furthermore, the precision and recall of search results are difficult to balance, relying excessively on human experience. The quality of search results is closely related to the user's own search skills and depth of understanding of the specific technical field. Inexperienced users often cause two extreme situations due to incomplete keyword selection or improper logical relationship construction: one is that the search results are too broad, including a large number of irrelevant and noisy documents; the other is that the search results are too narrow, resulting in serious missed detections and creating potential legal risks.

[0006] To address these challenges, the industry has attempted to introduce intelligent technologies such as natural language processing. However, the core technologies of existing so-called "intelligent search" tools mostly remain at the level of keyword-based matching or simple vector space model calculations. These technologies cannot deeply analyze the real search needs of users expressed in natural language, which contain multiple layers of intent, and are even less able to accurately map such needs to the complex table relationships and strict search syntax rules of structured patent databases.

[0007] In view of the above, this application is hereby submitted. Summary of the Invention

[0008] The purpose of this application is to provide a method, device, and storage medium for intelligently constructing patent search formulas, so as to construct accurate search formulas using a large language model.

[0009] To achieve the above objectives, this application adopts the following technical solution: Firstly, this application provides an intelligent construction method for patent search, including: Obtain the natural language of user patent queries; The natural language is input into a large language model for intent understanding to obtain intent information; the intent information includes at least retrieval technology intent, time range intent, geographic intent, and applicant intent. Convert the intent information into an SQL query statement; The SQL query statement is executed in the pre-established data table to obtain key information; and structured information is generated based on the key information. The structured information and the pattern information of the patent database are input into the prompt word template, and the prompt word template is input into the large language model to obtain the retrieval expression output by the large language model; A search is performed in the patent database based on the search query.

[0010] Secondly, this application provides an electronic device, comprising: At least one processor, and a memory communicatively connected to at least one of the processors; The memory stores instructions that can be executed by at least one of the processors, which enable the at least one processor to perform the aforementioned patent retrieval-based intelligent construction method.

[0011] Thirdly, this application provides a computer-readable storage medium storing computer instructions for causing a computer to execute the aforementioned patent search-based intelligent construction method.

[0012] Compared with the prior art, this application has the following significant advantages: 1) Extremely low barrier to entry: Users do not need to learn any search syntax; they can describe their needs using everyday language.

[0013] 2) Extremely high search efficiency: Reduces the professional search work that originally took several hours to seconds, completely eliminating manual trial and error.

[0014] 3) Accurate search results: Through deep understanding of intent by a large language model and reliable conversion of Text2SQL, the generated search query is logically rigorous, with accurate fields and comprehensive keywords, effectively avoiding missed and false detections.

[0015] 4) The generation of search queries has been transformed from a single, static step into a dynamic, feedback loop-based optimization process, which exceeds the conventional design thinking of those skilled in the art and strongly supports creativity. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the specific embodiments of this application or the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 This is a flowchart of a patent search-based intelligent construction method provided in an embodiment of this application; Figure 2 This is a schematic diagram of the structure of the electronic device provided in this application. Detailed Implementation

[0018] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of this application, including various details to aid understanding. These should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0019] The present application will be further described in detail below with reference to the embodiments.

[0020] Figure 1 This is a flowchart illustrating an intelligent patent search formula construction method provided in this application embodiment. This method can be executed by a computer program and integrated into an electronic device. This embodiment utilizes an intelligent patent search formula construction method integrated into an electronic device to construct accurate patent search formulas based on natural language. Figure 1 As shown, this embodiment provides a patent search-based intelligent construction method, including the following steps: S110, Obtain natural language from user patent queries.

[0021] For example, the natural language could be: "Please help me find relevant patents for heat pump air conditioning technology in new energy vehicles in the past five years, mainly from China and Japan, with a focus on the technologies of BYD and Toyota."

[0022] S120. Input natural language into a large language model for intent understanding to obtain intent information; the intent information includes at least retrieval technology intent, time range intent, geographical intent, and applicant intent.

[0023] Large language models perform intent understanding on natural language. For example, the natural language mentioned above is input into a large language model, along with the command "perform intent understanding on this natural language, including retrieval technology intent, time range intent, geographic intent, and applicant subject (or company, institution) intent".

[0024] The intent information includes a type field and a content field. The type field includes "search technology intent, time range intent, geographic intent, and applicant intent." The content field corresponds to the type field and consists of keywords under the specific type.

[0025] For example, the large language model outputs natural language at S110: Search intent: Technologies related to heat pump air conditioning in new energy vehicles; Intended timeframe: the past five years; Geographical intent: China and Japan; Intended purpose of the applicants: BYD and Toyota.

[0026] S130. Convert the intent information into an SQL query statement.

[0027] To facilitate the conversion of intent information into SQL query statements, this embodiment employs the text2sql technology. Text2SQL is a technology that automatically converts natural language questions (or instructions) into corresponding SQL (Structured Query Language) query statements.

[0028] Pre-construct corresponding data tables based on the type fields. For example, construct a keyword table, which includes multiple keywords and their corresponding extended keywords. Construct a classification number table, which includes multiple classification numbers and descriptive information for each classification number. Construct an applicant entity table, which includes multiple applicant entities and their corresponding related entities and former names. The related entities include the applicant entity's parent company, subsidiary company, and partner company, etc.

[0029] Optionally, the query address is first determined based on the type field of the intent information: the data table; that is, the retrieval technology intent corresponds to the keyword table and the classification number table, and the applicant intent corresponds to the applicant subject table.

[0030] Next, based on the content fields of the intent information, query content is generated in the data table. Specifically, keywords are extracted from the content fields of the retrieval technology intent: "new energy," "vehicle," and "heat pump air conditioner," generating query content in the keyword table, such as "search for extended keywords related to new energy vehicle heat pump air conditioners in the keyword table." Based on the content fields of the retrieval technology intent and the description information of the classification numbers, query content is generated in the classification number table, such as "search for classification numbers related to new energy vehicle heat pump air conditioners in the classification number table." Finally, query content is generated in the applicant entity table based on the content fields of the applicant entity intent, such as "search for all subsidiaries and former names of BYD and Toyota in the applicant entity table."

[0031] The following example illustrates the process of constructing an SL query statement: S131. Link the "content fields" in the intent information to the pre-built data tables. For example, semantically associate the content "new energy vehicle heat pump air conditioner" in the "retrieval technology intent" with the keyword field in the keyword table and the description content field in the classification number table; link the content "BYD" in the "applicant subject intent" with the standard name field in the applicant subject table.

[0032] S132. Determine the basic structure of the SQL query based on the "type field". For example, if it is recognized that "retrieval technology intent" and "applicant subject intent" need to be satisfied simultaneously, then generate an SQL skeleton containing WHERE conditions connected by AND; if there are multiple contents under the same type intent (such as multiple technical keywords), then connect them with OR.

[0033] S133. Fill the SQL skeleton with the specific values ​​after the link. During this process, query optimization strategies are introduced, such as: For keyword queries, semantic vector similarity is prioritized for matching, rather than just string matching. That is, keywords in the intent and keywords in the keyword table are converted into vectors, cosine similarity is calculated, and keywords with similarity above a threshold are selected to improve recall. For queries on the applicant entity table, recursive queries or graph database queries are used to obtain all levels of subsidiaries, parent companies, and former names of the applicant entity, ensuring the completeness of associated entities.

[0034] S140. Execute the SQL query statement in the pre-established data table to obtain key information; and generate structured information based on the key information.

[0035] Execute the SQL query in the pre-established keyword table to obtain extended keywords related to the retrieval technology; execute the SQL query in the pre-established classification number table to obtain the classification number of the field involved in the retrieval technology; execute the SQL query in the pre-established applicant subject table to obtain the applicant subject's associated subjects and former names.

[0036] For example, by querying the keyword table based on the extracted keywords, expanded keywords are obtained, thus expanding the keywords. Semantic matching is performed between the content field of the retrieval technical intent and the descriptive information in the classification number table to obtain classification numbers F28D9 and B60H1. Based on the query content obtained in the applicant entity table, a query is performed in the applicant entity table to obtain information on all subsidiaries of BYD and Toyota, as well as their former names.

[0037] Key information is categorized by type and value to generate structured information; structured information is also generated based on time-range and geographic intents. Since time-range and geographic intents are relatively clear and do not require expansion or understanding, structured information can be generated directly.

[0038] For example, the generated structured information is as follows: Keywords: New energy vehicles, electric vehicles, electric vehicles, heat pump air conditioning, refrigeration, etc. Classification numbers: B60H1 / 00, B60H1 / 32, F28D9, B60H1, F25B13 / 00, etc.; Time: >= 2019; Geographical scope: Chinese patent applications (CN), Japanese patent applications (JP); Applicants: BYD Co., Ltd., Shenzhen BYD Industrial Co., Ltd., etc.

[0039] In one specific implementation, the final structured information is generated based on an information fusion strategy. Specifically: S141. Execute SQL queries on different data tables to obtain an initial set of key information, such as: extended keyword set {K1, K2, ...}, related classification number set {IPC1, IPC2, ...}, and associated applicant entity set {A1, A2, ...}.

[0040] S142. Clean the initial key information obtained. For example, for the expanded keyword set, use the TF-IDF algorithm to sort the keywords by key and filter out redundant keywords with low relevance to the core intent; for the classification number set, summarize and deduplicate according to the IPC classification tree, and retain the most relevant main classification number or major group classification number.

[0041] S143. Assign different weights to different types of key information to determine the fields for retrieving patents in subsequent search queries. For example: Assign high weight to core technical keywords (such as "heat pump air conditioner"), and it is recommended to search in the title TI and abstract ABST fields. Assign medium weight to extended technical keywords (such as "refrigeration system"), and it is recommended to search primarily in the abstract ABST field. Assign high weight to the core classification number (such as B60H1 / 00) for exact matching. Assign medium weight to the higher-level classification number (such as B60H1) for range matching. This weighting information will be incorporated into the structured information as metadata for use in subsequent steps.

[0042] S150. Input the structured information and the pattern information of the patent database into the prompt word template, input the prompt word template into the large language model, and obtain the retrieval expression output by the large language model.

[0043] The schema information of the patent database includes: field names, field types, and search syntax rules. For example, the schema information includes "database field rules: TI: title, ABST: abstract, PD: publication date, PPA: main applicant, IPC: IPC classification number, using AND / OR / NOT logical operators".

[0044] Input the prompt word template into the large language model, and simultaneously input the target "You are a patent search expert. Please generate an executable search query in the database based on the following structured information," to obtain the search query: ((TI,ABST,CLM+=(New Energy Vehicle or Electric Vehicle or Electric Vehicle or Pure Electric or New Energy Vehicle or New Energy Car or Electric Vehicle or New Energy Vehicle?) and (Heat Pump Air Conditioner or Heat Pump or Air Conditioner or Thermal Management System or Thermal Management or Refrigeration System or Heating System or Refrigerant or Heating or Refrigeration or Heat Exchange or Heat Exchange)) or (IPC=(B60H1 / 00 or B60H1 / 32 or F25B13 / 00 or F25B30 / 02 or F25B30 / 06 or F25B41 / 00 or F25B41 / 06 or F25B41 / 20 or F25B49 / 02))) and (ADY>=2019)and (pa=("Shenzhen BYD Industrial Co., Ltd.","BYD Corporation","BYDGOSKIE B PBADAW","BYD","BYD COMPANY") LIMITED"...) and (ppacc=CN or ppacc=JP ) Optionally, the search query directly output by the large language model may be incorrect. The search query can be grammatically checked using preset rules or a lightweight model. Preset rules include: the number of left and right parentheses must be the same, and keywords must not be repeated.

[0045] The retrieval query is parsed using a lightweight syntax rule model (e.g., based on a decision tree or a predefined context-free grammar). This syntax rule model checks: 1) Operator usage validity: Check whether the use of AND, OR, and NOT operators conforms to the database's search syntax, such as avoiding consecutive occurrences of AND and OR operators.

[0046] 2) Field validity: Verify that the field names used in the search query (such as TI, ABST, IPC) exist in the pattern information of the patent database.

[0047] If an error is detected, the syntax rule model will automatically correct it, such as by adding missing parentheses or adjusting the order of operators.

[0048] Optionally, the grammatically checked search query can be sent to a simulated search environment (mirror or sampled database) of a patent database for a small-scale trial search, and the core indicators of the trial search results can be analyzed: result set size and result set relevance.

[0049] If the number of results in the result set is too large (e.g., >10,000), the search query is deemed too broad, and search query optimization is automatically triggered. For example, some keywords connected by OR are changed to connected by AND, or the weight of the core classification number is increased.

[0050] Based on semantic similarity calculation methods, the first N (e.g., 10) results in the result set are compared with natural language. For example, the semantic similarity of the abstracts of the first N patents is calculated with the natural language entered by the user during the search. If the average semantic similarity of the N patents is lower than a threshold, it is determined that the search query may deviate from the core intent. The large language model is then prompted to regenerate key information and new structured information, and subsequent steps are performed.

[0051] Optionally, before inputting the prompt word template into the large language model to obtain the retrieval expression output by the large language model, it is necessary to train the large language model so that it can fully understand the pattern information of the patent database and how to construct the correct retrieval expression.

[0052] Specifically, multiple training samples and labels are collected; wherein, the training samples are prompt word templates for input structured information and pattern information, and the labels are correct retrieval expressions; a large language model is trained based on multiple training samples and labels.

[0053] S160. Perform a search in the patent database based on the search query.

[0054] Enter the search query into the search box of the patent database and click the search button to perform the search.

[0055] Specifically, the generated search query is automatically displayed on the front end for users to view. Users can also modify the corresponding fields. Then, clicking "Search" will forward the automatically generated search query to the patent database's search interface for searching, and the returned patent list will be returned to the user. Users can also make further modifications based on this, greatly improving the efficiency of the initial search.

[0056] Compared with the prior art, this application has the following significant advantages: 1) Extremely low barrier to entry: Users do not need to learn any search syntax; they can describe their needs using everyday language.

[0057] 2) Extremely high search efficiency: Reduces the professional search work that originally took several hours to seconds, completely eliminating manual trial and error.

[0058] 3) Accurate search results: Through deep understanding of intent by a large language model and reliable conversion of Text2SQL, the generated search query is logically rigorous, with accurate fields and comprehensive keywords, effectively avoiding missed and false detections.

[0059] 4) The generation of search queries has been transformed from a single, static step into a dynamic, feedback loop-based optimization process, which exceeds the conventional design thinking of those skilled in the art and strongly supports creativity.

[0060] like Figure 2 As shown, this embodiment provides an electronic device, including: At least one processor; and A memory that is communicatively connected to at least one processor; wherein, The memory stores instructions executable by at least one processor, which, when executed, enables the processor to perform the described method. Since at least one processor in this electronic device is capable of performing the described method, it possesses at least the same advantages as the described method.

[0061] Optionally, the electronic device also includes interfaces for connecting the various components, including high-speed interfaces and low-speed interfaces. The components are interconnected using different buses and can be mounted on a common motherboard or otherwise installed as needed. The processor can process instructions executed within the electronic device, including instructions stored in or on memory to display graphical information of a GUI (Graphical User Interface) on an external input / output device (such as a display device coupled to the interface). In other embodiments, multiple processors can be used with multiple memories, and / or multiple buses can be used with multiple memories, if desired. Similarly, multiple electronic devices (e.g., as a server array, a group of blade servers, or a multiprocessor system) can be connected, each providing some of the necessary operations. Figure 2 Take processor 301 as an example.

[0062] The memory 302, as a computer-readable storage medium, can be used to store software programs, computer-executable programs, and modules, such as the program instructions / modules corresponding to the patent-search-based intelligent construction method in this application embodiment. The processor 301 executes various functional applications and data processing of the device by running the software programs, instructions, and modules stored in the memory 302, thereby realizing the aforementioned patent-search-based intelligent construction method.

[0063] The memory 302 may primarily include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on terminal usage. Furthermore, the memory 302 may include high-speed random access memory and non-volatile memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state storage device. In some instances, the memory 302 may further include memory remotely located relative to the processor 301, which can be connected to the device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0064] The electronic device may further include an input device 303 and an output device 304. The processor 301, memory 302, input device 303, and output device 304 can be connected via a bus or other means. Figure 2 Taking the example of a connection between China and Israel via a bus.

[0065] Input device 303 can receive input digital or character information, and output device 304 may include a display device, an auxiliary lighting device (e.g., an LED), and a haptic feedback device (e.g., a vibration motor). The display device may include, but is not limited to, a liquid crystal display (LCD), a light-emitting diode (LED) display, and a plasma display. In some embodiments, the display device may be a touchscreen.

[0066] This embodiment provides a computer-readable storage medium storing computer instructions for causing a computer to perform the methods described above. The computer instructions on this computer-readable storage medium, used to cause a computer to perform the methods described above, thus have at least the same advantages as the methods described above.

[0067] The medium in this application may be any combination of one or more computer-readable media. The medium may be a computer-readable signal medium or a computer-readable storage medium. The medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of the medium (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, the medium may be any tangible medium containing or storing a program that may be used by or in connection with an instruction execution system, apparatus, or device.

[0068] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of sending, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0069] Program code contained on a computer-readable medium may be transmitted using any suitable medium, including but not limited to wireless, wire, optical fiber, RF (Radio Frequency), or any suitable combination thereof.

[0070] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof. Programming languages ​​include object-oriented programming languages—such as Java, Smalltalk, and C++—as well as conventional procedural programming languages—such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0071] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired means, such as coaxial cable, optical fiber, digital subscriber line (DSL), or wireless means, such as infrared, wireless, microwave, etc. The computer-readable storage medium can be any available medium that a computer can access, or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium, or a semiconductor medium, etc. It is worth noting that the computer-readable storage medium mentioned in the embodiments of this application can be a non-volatile storage medium, in other words, it can be a non-transient storage medium.

[0072] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this application can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this application can be achieved, and this is not limited herein.

[0073] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A patent retrieval-based intelligent construction method, characterized in that, include: Obtain the natural language of user patent queries; The natural language is input into a large language model for intent understanding to obtain intent information; The intent information includes at least the intent of the retrieval technology, the intent of the time range, the intent of the geographical location, and the intent of the applicant. Convert the intent information into an SQL query statement; Execute the SQL query statement in the pre-created data table to obtain key information; And generate structured information based on the key information; The structured information and the pattern information of the patent database are input into the prompt word template, and the prompt word template is input into the large language model to obtain the retrieval expression output by the large language model; A search is performed in the patent database based on the search query.

2. The intelligent construction method for patent retrieval according to claim 1, characterized in that, The intent information includes a type field and a content field; The intent information is converted into an SQL query statement, including: Based on the type field of the intent information, determine the query address: data table; Based on the content field of the intent information, generate the query content in the data table.

3. The intelligent construction method for patent retrieval according to claim 1, characterized in that, Execute the SQL query statement in the pre-established data table to obtain key information, including: Execute the SQL query statement in the pre-established keyword table to obtain extended keywords related to the retrieval technology; Execute the SQL query statement in the pre-established classification number table to obtain the classification number of the field involved in the retrieval technology; Execute the SQL query statement in the pre-established application subject table to obtain the related subjects and former names of the application subject.

4. The intelligent construction method for patent retrieval according to claim 1, characterized in that, Structured information is generated based on the key information, including: Generate structured information based on key information according to type and value; Structured information is generated based on the stated time range and geographic intent.

5. The intelligent construction method for patent retrieval according to claim 1, characterized in that, The schema information of the patent database includes: field names, field types, and search syntax rules.

6. The intelligent construction method for patent retrieval according to claim 1, characterized in that, After obtaining the retrieval expression output by the large language model, the following is also included: The retrieval expression is grammatically checked using preset rules or a lightweight model.

7. The intelligent construction method for patent retrieval according to any one of claims 1-6, characterized in that, Before inputting the prompt word template into the large language model to obtain the search query output by the large language model, the process also includes: Collect multiple training samples and labels; wherein, the training samples are prompt word templates for input structured information and pattern information, and the labels are correct search terms; A large language model is trained based on the multiple training samples and labels.

8. An electronic device, characterized in that, include: At least one processor, and a memory communicatively connected to at least one of the processors; The memory stores instructions executable by at least one of the processors, which are executed to enable the at least one processor to perform the patent retrieval-based intelligent construction method according to any one of claims 1-7.

9. A computer-readable storage medium, characterized in that, The medium stores computer instructions for causing the computer to execute the intelligent construction method of the patent retrieval formula as described in any one of claims 1-7.

Images