Distributed patent retrieval method and system based on domestic autonomous controllable environment

Abstract

The invention discloses a distributed patent retrieval method and system based on a domestic autonomous controllable environment. The method comprises the following steps: 1) establishing a distributed patent data architecture which comprises a relational database for storing patent structured data, an NoSQL database and a high-capacity storage server for storing patent unstructured data; a retrieval engine is arranged in the NoSQL database; 2) storing the patent data of the same core country into the same node, and storing the patent data of each non-core country into the remaining nodes; a core information main table for storing patents and a plurality of sub-tables are arranged in the node; the sub-table is used for storing patent information associated with the main table and a main key of the main table; the retrieval engine extracts information from the patent data of each country to construct indexes and dictionary items of patents; and 3) according to retrieval information and a retrieval mode input by a user, querying the matched patent, the node where the structured data of the patent is located and the position where the unstructured data is located in the index.

Description

Technical Field

[0001] The information technology involved in this patent mainly revolves around the construction and management of patent databases, especially a distributed patent retrieval method and system based on a domestically produced, autonomous and controllable environment. Background Art

[0002] As an important carrier of technological innovation, the amount of patent database data is growing rapidly. However, there is currently a lack of a standard and efficient patent data architecture to cope with this explosive growth in data volume. Patent data is not only huge in quantity and in various forms, but also has intricate associations, forming a huge network structure, which poses a huge challenge to data retrieval and analysis. Therefore, it has become an urgent need to build a patent data architecture that can support massive data processing, ensure data security, have high scalability, and support a domestically produced, autonomous and controllable environment.

[0003] Although there are many intelligent patent search methods on the market, they often face the dilemma of insufficient computing power or resources in domestically produced autonomous and controllable environments. In order to achieve efficient and correct patent searches under limited resources, it is necessary to combine new data architectures and adopt advanced algorithms and technical means to improve the accuracy and efficiency of searches while reducing resource consumption. Such patent search methods will be able to better adapt to the complex needs of domestically produced autonomous and controllable environments.

[0004] In view of the current lack of efficient patent retrieval systems, how to effectively manage, store and analyze these huge and complex data has become an important issue. It is urgent to develop a patent retrieval system based on domestically produced independent and controllable technology. Summary of the invention

[0005] The present invention proposes a distributed patent data retrieval method and system based on a domestically produced autonomous and controllable environment to solve the problems in the prior art in a targeted manner. The present invention achieves efficient and accurate patent data retrieval by optimizing the data architecture and retrieval method under limited resources (e.g., lack of GPU). This patent also involves the field of software engineering, especially when developing a domestically produced autonomous and controllable patent retrieval system, it is necessary to consider how to design a friendly user interface, optimize the operating process, and ensure the stability and security of the system.

[0006] In order to solve the above technical problems, the specific technical solutions of the present invention are as follows:

[0007] A distributed patent search method based on a domestically produced autonomous and controllable environment, comprising the following steps:

[0008] 1) Establish a distributed patent data architecture, including a relational database for storing patent structured data, a NoSQL database and a high-capacity storage server for storing patent unstructured data; the relational database uses distributed multi-nodes for information storage; the NoSQL database is provided with a search engine and a Hash data structure for caching search results;

[0009] 2) The patent data of the same core country is stored in the same node, and the patent data of each non-core country is stored in the remaining nodes; the node is provided with a main table and multiple sub-tables; the main table is used to store the core information of the patent, and the sub-table is used to store the patent information associated with the main table and the primary key of the main table; the search engine extracts the key fields of each patent, the node where the patent data is located and the primary key of the main table, and the primary key information of the secondary table from the patent data of each country, and constructs the index and dictionary items of the patent, wherein the fields used as dictionary items include the receiving country, patent type, legal status information, language, classification number, license type, transfer type, technical field, and unit type;

[0010] The unstructured data in each patent is stored in the high-capacity storage server and a multi-level index system is created.

[0011] Saving the index information of the patent unstructured data in the multi-level index system into the index;

[0012] 3) According to the search information and search mode input by the user, the index is searched for matching patents and the nodes where their structured data are located and the locations where their unstructured data are located.

[0013] Furthermore, the main table and sub-table are horizontally divided according to the publication date timestamp of the patent, the patent data within the same set time period are put into the same group of horizontal sub-tables, and the data volume of each horizontal sub-table is set.

[0014] Furthermore, a dynamic database table and index management mechanism is provided in the node to monitor the status and data growth of the node. When the data growth reaches a set threshold, a dynamic adjustment process is automatically triggered, including creating a new database table, updating the index or partitioning the table.

[0015] Furthermore, the method for constructing the multi-level index system is: first, multiple first-level folders are created by country, and then multiple second-level folders are rebuilt under the first-level folders by patent type, and third-level folders are organized under the second-level folders in the order of patent publication numbers, and each third-level folder is limited to storing a set amount of patent data; the third-level folders are used to store unstructured data of patents.

[0016] Furthermore, the core information includes patent title, category, type, receiving country, language, application number, application date, publication number, publication date, authorization number, authorization date, expiration date, expiration date, agent, agency, substantive examination announcement date, technical field, publication status, views, likes, collections, comments, legal status and payment status information; the patent information stored in the subtable includes abstract information, claim information, classification number cluster information, family member information, patent drawing attachment information, legal information, applicant group information, citation information and other information.

[0017] Furthermore, the key fields include fields involved in the full-text search function and information fields in the list display; the fields involved in the full-text search function include the associated primary key of the relational database, the unique identifier of the main table and the unique identifier of the subtable; the information fields in the list display include title information, abstract information, country information, classification information, full-text information, application information, attachment information, public information, claim information, and classification number information.

[0018] Furthermore, the search modes include general search, advanced search and semantic search.

[0019] Furthermore, the search results of each search mode are standardized so as to unify their formats.

[0020] A distributed patent search system based on a domestically produced autonomous and controllable environment, characterized in that it includes a corpus management module, a system management module and a search module;

[0021] The corpus management module adopts a distributed patent data architecture, including a relational database for storing patent structured data, a NoSQL database and a high-capacity storage server for storing patent unstructured data; the relational database adopts distributed multi-nodes for information storage; the NoSQL database is provided with a search engine and a Hash data structure for caching search results;

[0022] The system management module is used to store patent data of the same core country in the same node, and to store patent data of non-core countries in the remaining nodes; a main table and multiple sub-tables are provided in the node; the main table is used to store core information of patents, and the sub-table is used to store patent information associated with the main table and the primary key of the main table; the search engine extracts key fields of each patent, the node where the patent data is located and the primary key of the main table, and the primary key information of the secondary table from the patent data of each country, and constructs the index and dictionary items of the patent, wherein the fields used as dictionary items include the receiving country, patent type, legal status information, language, classification number, license type, transfer type, technical field, and unit type; the unstructured data in each patent is stored in the high-capacity storage server and a multi-level index system is created, and the index information of the unstructured data of the patent in the multi-level index system is saved in the index;

[0023] The search module is used to search for matching patents and nodes where structured data is located and locations where unstructured data is located in the index according to search information and search patterns input by the user.

[0024] The present invention provides a distributed patent data architecture based on a domestically produced, autonomous and controllable environment, including:

[0025] There are two core parts: structured data storage and unstructured data storage. In terms of structured data storage, the present invention cleverly combines the advantages of supporting domestically produced, independently controllable relational databases and NoSQL databases. This type of relational database is an ideal choice for storing patent information due to its structured, large storage capacity, strong stability, easy expansion and maintenance, and outstanding transaction processing capabilities. The NoSQL database, on the other hand, provides strong support for the retrieval of patent data with its high efficiency and flexibility in search and analysis. Through the combination of the two, the security and consistency of the data are guaranteed, and the efficiency of data retrieval is improved.

[0026] Furthermore, a domestically produced, self-controlled distributed solution that supports the above database is adopted. The table structure is split, reorganized and designed for basic patent information, patent legal information, patent classification, patent attachments, patent abstracts, patent citation information, patent applicants, patent payment information, and basic user, role, menu and other information required by the system. Distributed multi-node and sub-table technical means are used for relational databases to store patent-related tables and basic tables required by the system to achieve physical isolation of different types of data.

[0027] Based on a detailed analysis of the global distribution of patents, we have optimized the patent data storage strategy and adjusted the country classification to focus on seven core countries that significantly dominate patent applications. For the remaining countries with fewer patents, they are unified into one distributed node to more effectively manage and highlight the main trends and leading countries.

[0028] In each storage node, in view of the numerous fields of patent data, vertical splitting is used to distribute data into multiple physical tables according to the attributes and access patterns of the data. Vertical splitting is to split a table with many columns into multiple tables. This splitting method mainly focuses on the division of data columns, aiming to reduce the number of fields in a single table, thereby optimizing the performance of the database. Vertical splitting usually stores uncommonly used field tables in a separate table, which can reduce the number of fields in common queries and improve query efficiency. In this system, the main table mainly stores the core information of patents, including patent title, category, type, receiving country, language, application number, application date, publication number, publication date, authorization number, authorization date, expiration date, agent, agency, substantive examination announcement date, technical field, publication status, views, likes, collections, comments, legal status and payment status. It is worth noting that although the main table may not cover comprehensive information, it is designed to quickly link to the subtable and associate foreign keys to ensure instant access to the full amount of patent data. Based on this efficient architecture, users can directly search and obtain the precise content required for the list in the main table, and gain insights into the global patent distribution and core country trends without tedious operations.

[0029] The sub-table is used to store other information associated with the main table, including several dimensions: summary information, claim information, classification number cluster information, family member information, patent figure attachment information, legal information, applicant group information, citation information and other information. For other single pieces of information that cannot be grouped or scaled, we integrate them into a comprehensive information table to comprehensively cover diversified content such as inventors, value scores, international application details, funding records, award information and confidentiality levels. This design ensures centralized management and efficient query of information, making it easy for users to quickly retrieve the specific items they need. The legal status of the patent will be updated during patent review and payment, and the data will be processed and displayed accordingly when viewing the patent details page. This splitting method not only reduces the complexity and data volume of a single table and improves the flexibility of data management, but also helps to optimize query performance and speed up data processing.

[0030] Furthermore, the above master and sub-tables are horizontally divided according to the patent publication date timestamps. For example, the patent data applied between 2015 and 2017 are placed in a group of horizontal sub-tables, and the amount of data in each table is strictly controlled within 20 million to optimize performance and management efficiency. In this way, the physical isolation of data is achieved. This isolation method has multiple advantages. First, it ensures that the privacy and security of data are strictly protected, thereby effectively preventing the risk of data leakage. At the same time, it can also significantly reduce the mutual influence and potential interference between different data, and ensure the accuracy and reliability of data processing. It is worth mentioning that in the face of downtime events that may occur under special circumstances, this isolation mechanism shows strong fault tolerance. Even if a part fails, other parts can still operate normally, ensuring the overall stability and availability of the system using this architecture.

[0031] The present invention also demonstrates outstanding innovation in NoSQL databases. On the one hand, a search engine is used to support the search method, and on the other hand, a cache library is used to store each search result, so that data can be quickly obtained with the same search conditions next time.

[0032] Furthermore, the same country plus timestamp format is used in the search engine for node distribution and sub-index establishment, the difference is that the index does not need to be split vertically. Different from the traditional practice of storing all fields in one index, the present invention adopts a more refined indexing strategy, which only extracts the key fields of the search requirements in the patent data, including the fields involved in the full-text search function such as: the associated primary key of the relational database, the unique identifier of the main table and sub-table of the patent data, and the information fields in the list display, such as: title information, abstract information, country information, classification information, full-text information, application information, attachment information, public information, claim information, classification number information, etc. The advantage of this indexing strategy is that when the user enters the query conditions, the system can quickly locate the relevant data in the index. If the full set of data is needed, it can be obtained by associating the stored primary key to the corresponding relational database storage node, which not only improves the search speed, but also can accurately match user needs, thereby greatly improving user satisfaction.

[0033] When the service is started, the dictionary items that are commonly used and basically unchanged in the database (such as: receiving country, patent type, legal status information, language, classification number, license type, transfer type, technical field, unit type, etc.) are stored in the search engine, and once the dictionary item is modified, the dictionary item in the index is updated synchronously, and the results of different search conditions of different users with timeliness are stored. When users search for the same condition in a very short time, feedback can be quickly given to users, improving the user experience.

[0034] Furthermore, we design an efficient caching strategy to use the Hash data structure to store and retrieve query results based on various conditions. Specifically, we convert the query conditions (such as keywords, time ranges, classification tags, etc.) initiated by users into keys through certain encoding rules (such as hashing, serialization, etc.), and then store the query results (usually data lists, object collections, or aggregated statistical results) as values ​​in the corresponding Hash. In order to control the life cycle of the cache and prevent the cached data from occupying memory resources indefinitely, we will set a specific expiration time for the cache item each time we cache data, such as one hour. In this way, when the cache item reaches its set expiration time, it will be automatically deleted from the memory to make room for new cached data. In this way, we can not only provide us with a fast and flexible caching solution, but also help us automatically manage the life cycle of cached data, ensure the validity and freshness of the cache, and thus optimize the performance and response speed of the application. At the same time, its distributed characteristics and high availability design also make this caching mechanism easy to cope with the needs of large-scale concurrent access and distributed deployment.

[0035] In the face of the severe challenge of the rapid growth of patent data, the present invention innovatively introduces a set of intelligent dynamic database table and index management mechanisms. This mechanism closely monitors the status of database nodes and data growth through real-time daemon threads or plug-in scripts. Once the data is found to be close to the preset threshold (such as 20 million records), the system will automatically trigger a dynamic adjustment process, including creating a new database table, adjusting existing indexes or partitioning tables, etc., to ensure that database performance is not affected.

[0036] In addition, in order to optimize the search efficiency and ensure comprehensive coverage of the data, the node status is monitored in real time, and the number of statistical tables or indexes is constructed to build a search map or search dictionary. Through the intuitive display of the search map, users can easily browse the data distribution, patent data volume, and the last update date of these data corresponding to each country on the page. This function greatly enhances the visualization of data distribution within the system, provides accurate navigation for search operations, and provides users with a convenient information overview and instant feedback on data freshness. In the subsequent search process, the system will first analyze the search formula through the search map to quickly locate the table or index that may contain the target data, thereby greatly improving the accuracy and efficiency of the search.

[0037] Through this innovative combination of dynamic expansion and intelligent retrieval, the present invention not only effectively addresses the storage and retrieval challenges brought about by data growth, but also significantly improves the adaptability and scalability of the system, ensuring that the system can still maintain efficient and stable operation when facing a continuously growing amount of data.

[0038] For unstructured data (such as patent drawings and attachments), we have designed an efficient and centralized storage management solution. The solution relies on a high-capacity storage server to ensure the secure and centralized storage of data. It uses advanced file management middleware, which not only supports domestically produced, self-controlled environments, but also optimizes file access efficiency. We have built a multi-level indexing system. First, the first-level folders are divided by country, then the second-level folders are subdivided by patent type, and finally the third-level folders are organized in the order of publication numbers. Each folder is limited to storing data of 10,000 patents. The folder contains all complete patent unstructured data. This strategy distributes files to different paths in an orderly manner. When a new file is uploaded, the system automatically assigns it a unique name in a format with a timestamp, and instantly updates the path information in the relational database and search engine node to ensure that the file has a unique identity and precise location in the system. This design not only simplifies the file management process, but also greatly improves the speed and efficiency of file access.

[0039] The present invention provides a distributed patent data retrieval method based on a domestically produced, autonomous and controllable environment, comprising:

[0040] Based on the application of the above patent data architecture, the retrieval method of the present invention can be divided into a general retrieval method, an advanced retrieval method and a semantic retrieval method. Regardless of the retrieval method, it depends on the index information, metadata, keywords and other information retained when the data is stored. In addition, the general retrieval method relies on keyword matching, the advanced retrieval relies on metadata and query syntax parsing, and the semantic retrieval relies on text semantic information and machine learning models. These retrieval methods all follow a set of efficient data acquisition logic. First, the target database is accurately locked through the data reduction rules preset by the data architecture, thereby avoiding the cumbersome process of comprehensive traversal of the huge distributed database, and realizing accurate retrieval on a specific small range of data sets. In addition, the present invention cleverly combines NoSQL database technology, pre-caches possible retrieval results, significantly shortens the subsequent response time under the same retrieval conditions, and effectively reduces resource consumption. Finally, in order to ensure the consistency and scalability of the system, the present invention unifies the input and output parameters of the three retrieval methods, ensures the consistency of data reduction, and lays a solid foundation for subsequent functional connection.

[0041] In the general search method, users are allowed to narrow the search scope according to specific conditions, thereby improving the accuracy and efficiency of the search results. The user's search history is recorded so that users can quickly re-search previous queries. When the user enters a keyword, relevant associative words or phrases are provided in real time based on the user's input to help users express their search intentions more accurately. Index optimization is performed on various types of information in the patent database, and technologies such as inverted indexing and distributed storage are used to improve the search speed. The query conditions are accurately matched with various parts of the patent data set, and the relevance evaluation model is used to calculate the relevance score of each match with the query condition. The search results are automatically classified according to the type, field, technical keywords, etc. of the patent, and statistical data such as quantity and distribution of each category are provided. This helps users to understand the search results more comprehensively and quickly locate the content of interest. Allow users to further narrow the search scope or change the search conditions based on the initial search results. In order to improve the efficiency of the secondary search, the system can save the user's initial search conditions and allow users to directly modify and adjust these conditions, such as adding / deleting keywords, adjusting weights, etc., to improve the flexibility and efficiency of the search. Support comprehensive search of the entire document or data set to ensure that users can find all relevant content. Allow users to filter and sort search results based on specific conditions. Highlight content related to the search keyword in the search results to help users quickly locate information of interest. In order to improve the visual effect and readability of the highlight prompts, the system can use eye-catching colors, fonts and styles to highlight relevant content, and provide an option to turn off the highlight prompts to meet the needs of different users.

[0042] In the advanced search method, a comprehensive patent search tool is provided to enable users to locate the required information efficiently and accurately. First of all, it supports the screening and statistical functions of patent major categories to ensure that users can quickly locate the patent field of interest and understand the patent distribution and trends in this field. In order to meet the user's more refined search needs, the ability to dynamically build Boolean type search formulas is introduced. Users can freely combine different logical operators and nest brackets according to their own needs to build complex and accurate search formulas. At the same time, a rich field option is provided to enable users to build search logic for different fields to achieve more accurate searches. In terms of matching mode, both precise matching and fuzzy matching are supported. Precise matching can help users quickly find patents that fully match the query keywords, while fuzzy matching can expand the search scope and find patents related to or similar to the query keywords. Users can flexibly switch matching modes according to actual needs to obtain the best search results. In addition, users are allowed to make more complex query combinations. Users can combine multiple search formulas together to form more complex query conditions, thereby more comprehensively covering the required patent information. The user operation process is as follows:

[0043] 1. The user inputs query conditions,

[0044] The user selects appropriate fields according to their own needs (such as title, abstract, claims, inventors, etc.), and uses logical operators (AND, OR, NOT) and parentheses to construct a Boolean retrieval expression.

[0045] 2. Construct the retrieval expression

[0046] The user can dynamically construct the retrieval expression through the tools provided by the interface. For example: title: "New energy vehicle" AND abstract: "Battery technology" (inventor: "Zhang San" OR inventor: "Li Si") AND classification: IPC / A61K.

[0047] 3. Processing of the retrieval system:

[0048] Parse the retrieval expression: The system first parses the retrieval expression input by the user to identify the fields, operators, and values therein. Split the retrieval expression: The system further splits each part of the expression and matches it with the stored data. Match the stored results: The system compares each query condition with the patent records in the database. The main stored results involved here include: Basic patent information: such as patent number, application date, publication date, etc. Patent details: including abstract, claims, specification, etc. Patent legal information: such as patent status, rights transfer records, etc. Patent classification: International Patent Classification (IPC), Cooperative Patent Classification (CPC), etc. Patent attachments: such as drawings, photos, etc. Patent abstract: Abstract text. Patent citation information: List of cited and citing patents. Patent applicants: Information such as applicant name, address, etc.

[0049] 4. Matching process

[0050] Exact match: If the user selects an exact match, the system only returns those patent records that exactly meet the query conditions.

[0051] Fuzzy match: If the user selects a fuzzy match, the system may use means such as synonym expansion, thesaurus, natural language processing technology, etc. to expand the search scope to capture more relevant patents.

[0052] 5. Combining multiple retrieval expressions The user can also create multiple retrieval expressions and use more complex logic to combine these expressions to further narrow or expand the search scope.

[0053] 6. Display the retrieval results Finally, the system will display the retrieval results according to the sorting method specified by the user (such as relevance, application date, publication date, etc.).

[0054] Users can also further filter, sort or analyze the search results. In this way, users can efficiently and accurately locate the patent information they are interested in.

[0055] In the semantic retrieval method, keyword search focuses on the direct matching of words, while semantic search has a deeper understanding and matching of the meaning of the query terms. Especially in the patent search scenario, although semantic search may not directly return the exact matches of the query terms, it can better capture the user's true intentions, thereby providing information that is more in line with the needs. Semantic search integrates natural language processing technologies such as error-tolerant spelling checking, vocabulary segmentation and normalization to ensure that valid results can be obtained even if the input is slightly deviated. Using advanced vector search technology, the system is able to identify and return documents that are highly relevant to the query meaning. In a domestically controlled environment, in view of computing power and resource constraints, an optimization strategy is adopted to improve retrieval efficiency. Specifically, by pre-generating vector representations of patent documents and storing them in the database, the need for real-time computing is reduced. During retrieval, the system instantly generates vectors corresponding to the user's query, quickly finds the most similar documents in the database, and evaluates semantic relevance by calculating similarity. Users can create personalized intelligent retrieval algorithms by custom configuring semantic computing and retrieval system components. In this process, users can train models and select the most suitable algorithm for semantic retrieval to generate highly personalized retrieval results. This design ensures that users can customize their search experience according to their needs.

[0056] The present invention provides a distributed patent data retrieval system based on a domestically produced, autonomous and controllable environment, comprising:

[0057] The corpus management module is used to manage patent data, including the management of full data and incremental data, and to implement data update operations such as adding, modifying, and deleting patent data. It provides an overview of the patent list and a detailed view of a single patent. It provides an interface service for an overview of the patent data list and a detailed view of each patent.

[0058] The corpus management module is a carefully designed management strategy specifically for patent data. It implements basic data update operations such as adding, modifying, and deleting, ensuring the real-time and accuracy of patent data, and uses the above-mentioned advanced storage structure to optimize the data processing flow. When adding new patent data, the system will accurately divide the complex patent data into multiple dimensions according to the preset storage logic, and perform refined data storage one by one. This step-by-step storage strategy not only improves data processing efficiency, but also effectively ensures the integrity and accessibility of the data, laying a solid foundation for subsequent query and analysis.

[0059] In order to better support the dynamic update of data, this module has specially carried out a refined data management design for the full data and incremental data of the patent corpus. Whether it is the overall data update or the incremental update of part of the data, the system can complete it quickly and accurately, ensuring the timeliness and accuracy of the data. When using this module, users can easily obtain the required data information through powerful query functions. The system supports displaying the key fields of the data in the form of a list. Users can also change the displayed fields through simple configuration operations according to actual needs, thereby meeting the diverse data viewing needs. In addition, the module also provides multiple filter items, and users can freely combine these filter items as needed to accurately filter the search results. This flexible filtering mechanism greatly improves the efficiency and accuracy of the search, allowing users to quickly locate the required information. Finally, in order to enhance the flexibility and controllability of data management, this module also supports the release and withdrawal of data. Users can publish or withdraw the entered data according to actual needs, providing flexible data display control for the front-end search end. This design not only meets the diverse needs of users for data management, but also improves the overall performance and user experience of the system.

[0060] The search module, based on the application of the above-mentioned patent data search method, provides functions such as search result display and browsing, patent precision push, patent cluster push, and search result scoring.

[0061] Furthermore, based on the application of the above-mentioned search method, the search module further integrates advanced functions such as associative search and multilingual search to provide users with a more intelligent and convenient patent search experience. This module supports users to search for patents by entering text or patent application number / publication number. Its unique semantic search technology can deeply analyze user input and output patent search results according to the degree of semantic similarity, ensuring that users can obtain patent information that is highly matched with their search intent. In terms of associative search, when a user enters part of a search term, the system will automatically display other words that are semantically related to the input term in a drop-down list. This function can not only help users express their search intent more accurately, but also effectively improve the efficiency and accuracy of retrieval.

[0062] In order to meet the needs of users of more languages, this module specifically supports the interconnected search function of the two major languages ​​of Chinese and English. Regardless of whether the user enters Chinese or English text, the system can automatically translate it and search with the search content in both languages, and finally return the bilingual search results. This function breaks the language barrier, allowing users to easily obtain more required patent information. In addition, this module also supports the classification and display of search results according to different business scenarios. Users can filter and sort the results according to dimensions such as patent type and label to quickly locate the content of interest. At the same time, the system also supports multiple sorting methods for search results, such as semantic relevance sorting, user click volume sorting, and latest release sorting, to meet the diverse search needs of users.

[0063] In order to improve the readability of the search results, the module also supports the highlighting function of the search terms in the result set. Keywords and word segments will be highlighted in eye-catching colors in specific fields of the result set, which is convenient for users to quickly identify and locate. At the same time, the highlighted fields and colors can be flexibly configured according to user needs. In order to further improve the user experience, the module also provides a search strategy subscription function. Users can personalize the patent information they are interested in during the search process and save the search rules as subscription policies. The system will regularly push relevant patent data according to the user's subscription strategy and display them in clusters to help users keep abreast of the latest patent trends. Finally, the module also introduces a search scoring mechanism. By scoring the degree of match between the user's search intent and the actual returned results, the system can continuously optimize the search algorithm and strategy to provide users with more accurate and efficient patent search services. At the same time, these scoring data can also provide strong support for subsequent system operation and maintenance and optimization.

[0064] The system management module provides user search log storage, dictionary configuration for optimizing search results such as proprietary words, synonyms, and sensitive words, policy tools such as top-blocking of search results, and the ability to maintain basic background information, including users, permissions, logs, data import and export, etc. (data import and export fields and formats are configurable).

[0065] Furthermore, the background management module plays a vital role, especially in the preservation and management of user search logs. The real-time stream processing technology Apache Flink is introduced to ensure that every user search behavior can be captured and transmitted to the log system in real time. At the same time, the log data is encrypted and stored by AES encryption to ensure user privacy and security. The user's search behavior is recorded, including core information such as search keywords, search timestamps, and user unique identifiers. The accumulation of these data enables the system to analyze the user's historical search behavior, click records, and browsing habits, and use machine learning algorithms to predict the user's next search intention and lay a data foundation. The machine learning algorithm will use real-time analysis technology based on the user's preferences and behavior patterns to monitor the user's interaction with the search results (such as clicks, dwell time, etc.), dynamically adjust the order of search results, ensure that the user's most interested or most relevant content is displayed first, and realize context-aware intelligent recommendations. It helps to better understand user needs and improve user experience. In addition, in order to meet the personalized needs of different users, the background management module supports flexible configuration of dictionaries such as proprietary words, synonyms, and sensitive words. Users can easily add, edit, or delete relevant words according to their own business characteristics, thereby significantly improving the accuracy and efficiency of retrieval. This highly customizable dictionary configuration function ensures the accuracy and pertinence of search results.

[0066] The background management module provides powerful policy configuration functions for the display and control of search results. Users can customize the top and blocking policies of search results according to their needs. It also supports users to set dynamic weights for different search result fields (such as title, abstract, keywords, etc.) in the background. The system adjusts the search result sorting in real time according to these weights, so that users can more flexibly control which information is more prominent in the results. In this way, important information that users are concerned about can be displayed to users first, and unnecessary content can be effectively filtered, which greatly improves the effectiveness of search results and user satisfaction. In terms of patent basic data management, the background management module also performs well. It provides configurable dictionary management functions, allowing users to configure different versions of classification numbers and other basic patent metadata according to actual needs. This flexible configuration method enables the system to adapt to search needs in different scenarios and ensure the comprehensiveness and accuracy of search results.

[0067] In addition to the above functions, the backend management module also provides general functions of the backend system, such as user registration, login, and permission management. Users can easily enter the platform through the registration function, and the specific operation permissions are dynamically assigned by the system administrator according to the user's role and responsibilities. This refined permission management ensures the security and stability of the system. Finally, in order to facilitate users to process large data sets, the backend management module also provides data import and export functions. Users can import large amounts of data into the system for analysis and processing as needed, or export the processed data to files with specified formats and field configurations to adapt to different usage scenarios. This efficient data processing capability greatly improves the user's work efficiency.

[0068] Compared with the prior art, the present invention has the following positive effects:

[0069] The present invention not only has efficient and accurate data retrieval functions, but also fully considers the security, compliance and scalability of data, and can achieve fast and accurate patent retrieval and analysis by applying the above data architecture and retrieval method. At the same time, the system also provides a friendly user interface and operation process to facilitate users to perform patent retrieval and data analysis. The successful development and application of this system will provide strong support for scientific and technological innovation in a domestically-developed and controllable environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 This is a system data architecture diagram of the present invention.

[0071] Figure 2 This is a module diagram of the patent search system of the present invention.

[0072] Figure 3 This is a diagram of the distributed relational database architecture of the present invention.

[0073] Figure 4 This is a diagram of the distributed NoSql database architecture of the present invention.

[0074] Figure 5 It is a general search flow chart of the present invention.

[0075] Figure 6 It is the advanced search flow chart of the present invention.

[0076] Figure 7 This is a flowchart of the semantic retrieval of the present invention. DETAILED DESCRIPTION

[0077] The present invention is further described in detail below in conjunction with the accompanying drawings. The examples given are only used to explain the present invention but not to limit the scope of the present invention.

[0078] The present invention proposes a distributed patent data architecture based on a domestically produced autonomous and controllable environment, which aims to store and manage a large amount of patent data in an efficient and flexible manner. In the domestically produced autonomous and controllable environment, analysis can be performed from two dimensions: hardware and software. However, the present invention mainly focuses on the development at the software level. For the domestically produced hardware environment used at the bottom layer, generally speaking, these hardware will not have a significant impact on the core functions and performance of the present invention. In order to ensure the stable operation of the software on different hardware platforms, compatibility issues are fully considered during the design and development process, and reasonable adaptation is made through necessary minor adjustments to ensure that the software can perform well in various domestically produced autonomous and controllable hardware environments.

[0079] like Figure 1 System data architecture diagram. In terms of technology selection, the present invention adheres to the principles of open source, high compatibility, easy to obtain and fast construction. At the operating system level, the widely used Galaxy Kylin system is selected. In the selection of relational databases, the widely recognized Mysql is used as the core data storage system. In order to further improve the efficiency and accuracy of data retrieval, Elastic search is introduced as the retrieval engine, and its powerful full-text search and distributed characteristics can meet complex data query requirements. At the same time, in order to alleviate the pressure of database access, Redis is selected as the cache database, and the number of direct accesses to the database is reduced by caching hot data, thereby improving the response speed of the system. In addition, in order to meet the storage requirements of unstructured data, the present invention introduces MinIO as an efficient and reliable file management system, which can easily cope with the storage challenges of massive patent data with its excellent scalability, security and ease of use. In addition, in response to the needs of large-scale data processing and distributed databases, the ShardingSphere distributed database middleware solution is introduced. It provides functions such as data sharding, read-write separation, and distributed transactions, which can ensure balanced distribution and efficient access of data between multiple database nodes, thereby meeting the requirements of the present invention for high performance and high availability. It should be clear that the current technology examples selected in this data architecture design are not the only options, but are representative examples listed based on the existing technology ecosystem and actual needs. In practical applications, in addition to the currently listed technologies, the present invention also encompasses and considers various other technology options.

[0080] When dealing with large amounts of structured data, especially data stored in relational databases, such as Figure 3As shown in the figure, a distributed database middleware is introduced to implement refined node management. The main table has multiple distributed nodes. In order to ensure the uniqueness and scalability of the primary key, the snowflake ID strategy is used in these distributed nodes to generate the primary key. According to the actual situation of the number of patent publications, an efficient data segmentation strategy is formulated, that is, segmentation according to the country dimension, so as to build multiple distributed database nodes. Specifically, an independent data node is set up for each selected core country, and a comprehensive node of "other countries" is set up to accommodate patent data from other countries. In this distributed architecture, a vertical splitting strategy is further implemented according to the specific application requirements of the patent data business. The vertical splitting covers nine key business dimensions, including basic information, abstract, claims, patent families, legal information, citation information, attachments, applicants (patent holders), and classification numbers, aiming to ensure the integrity of the data and fully meet the business needs. In order to achieve fast association and query of data, ensure that the primary key information of the main table is saved in the table. At the same time, based on business needs and data access patterns, the sharding key and corresponding sharding strategy are carefully set to achieve efficient storage and retrieval of data. This optimization measure not only improves data processing efficiency, but also ensures the accuracy of business logic and the reliability of data.

[0081] Given that patents are technological achievements with a strong timeliness, the frequency of data updates and query efficiency are crucial. Therefore, on the basis of vertical splitting, the split forms are copied on each data node, and the single-table data is organized according to a certain time range (such as quarter, year). Doing so not only helps to ensure that the amount of data in a single table is controlled within 20 million to avoid reaching the storage limit, but also maintains a high efficiency when querying. This multi-dimensional splitting and replication strategy not only significantly reduces the amount of data in a single data table and improves query performance, but also manages data in an orderly manner according to the timeline, providing strong support for subsequent data analysis and mining. Through such optimization, patent data can be processed and analyzed more efficiently and accurately, providing strong support for technological innovation and intellectual property management.

[0082] For NoSQL databases, it consists of two major components: the Elasticsearch search engine and the Redis cache. For the index setting of the Elasticsearch search engine, the logic similar to that of the structured database is followed to build the index according to the country and time range. The key fields in the patent data, such as patent number, publication number, title, receiving country, abstract, specification, claims, etc., are screened and synchronized to meet the core needs of the retrieval application. In addition, in order to ensure the integrity of the data and the convenience of subsequent analysis, the primary key ID in the main table of the relational database is also saved in the index. In this way, when it is necessary to obtain detailed information on a certain patent, all fields and data of the relevant records can be easily identified and retrieved through the primary key ID. This innovative indexing strategy enables the system to quickly locate the required data in the Elasticsearch index when facing the user's query request, and return it to the user efficiently. Compared with the traditional practice of storing all fields in one index, this strategy significantly improves the speed and accuracy of retrieval.

[0083] For the Redis cache part, we make full use of its efficient data caching function. Whenever a search is performed, the system will first check whether there is corresponding cache data in Redis. If matching cache information is found, the system will directly return this data, thus avoiding unnecessary database queries and computing overhead. Only when relevant information cannot be found in the cache will the system call the Elasticsearch search engine for logical queries. Through the collaborative work of this cache and search engine, the goal of fast response and efficient search is achieved, providing users with a smoother and more convenient data query experience.

[0084] With the rapid growth of patent data, the present invention has shown excellent flexibility. It has the ability to dynamically create database tables and Elasticsearch indexes. Through the automation mechanism of scripts, daemon threads or integrated system functions, the present invention can detect and accurately count the distribution of data in the current database in real time, and dynamically create new database tables and Elasticsearch indexes according to the above configuration rules, so as to ensure the optimization of data structure and the improvement of query performance. Taking the external script as an example, first build a highly modular and extensible interface service. It is necessary to have a clear interface exposure capability, and it is also necessary to encapsulate dynamic database operations, especially the creation and management of database tables and indexes, into easy-to-call methods. Such a design ensures the flexibility and reusability of the script and lays a solid foundation for the data processing process. When new data is imported or added externally, the script first starts the detection mechanism to accurately capture the data release timestamp and country information to determine the data destination. Further quickly locate the database partition or index where the data should be inserted. When facing data of a new country or time range, the script will automatically trigger the middleware command to create the corresponding data form and index and import the data. For the case where it needs to be inserted into an existing node, the script will first perform a detailed data volume statistics task. If it is detected that the amount of data in the target form or index exceeds the preset threshold (such as 20 million records), the creation process of a new form and index is automatically triggered and the data is introduced. This strategy effectively avoids the excessive expansion of a single database object and ensures the query performance and convenience of data maintenance. At the same time, the script will also synchronously update the retrieval method file to record the information of the newly added forms and indexes in detail. This step is crucial because it ensures that the query engine can instantly identify and utilize the latest data architecture, realizing the automation and intelligence of patent data processing. This dynamic expansion feature enables the present invention to continuously adapt to the needs of data growth, and maintain an efficient and stable operating state regardless of whether it is facing sudden data surges or long-term stable growth. It can not only effectively manage massive data, but also provide solid technical support for the analysis, retrieval and application of patent data.

[0085] For unstructured data, such as the drawings and attachments of patents, the present invention introduces an advanced file management system for storage. When a new drawing or attachment is uploaded, the system will automatically generate a file path named according to the timeline and file format. And store the corresponding records in a structured database. This design allows each file to have a unique identifier and location in the file system, thereby simplifying the file management and access process. Users can easily access and download these files through the interface provided by the system.

[0086] The present invention innovatively integrates a number of cutting-edge technologies such as relational databases, NoSQL databases, distributed database middleware, and file management systems, and carefully creates an efficient, flexible, secure and reliable distributed patent data architecture. While meeting the domestic independent and controllable environment, it also demonstrates wide adaptability to other environments, aiming to meet the growing demand for patent data storage and retrieval. This architecture not only significantly improves the performance of data storage and retrieval, and effectively shortens the response time, but also strictly adheres to the principles of data security and privacy protection, providing solid technical support for the management and utilization of patent data. Through this innovative architecture, it can provide a full range of services for the storage, query, analysis and application of patent data, and contribute to the protection and utilization of intellectual property rights.

[0087] This invention proposes a distributed patent search method based on a domestically-developed autonomous and controllable environment, which aims to overcome challenges that may be encountered in this environment, such as insufficient computing power, resource constraints, and slow processing speed. Through innovative distributed architecture design and algorithm optimization, this method can make full use of existing resources to achieve efficient and accurate patent search, and provide strong support for patent data management and utilization in a domestically-developed autonomous and controllable environment.

[0088] Based on the application of the above patent data architecture, three search strategies are designed to meet the needs of users at different levels. First, the general search function is tailored for ordinary patent search users. Through simple query operations, users can quickly find patent data that matches the query conditions and quickly obtain the required information without complex operations. Secondly, in order to meet more professional patent search needs, an advanced search function has been developed. This function is aimed at users who have higher requirements for patent search, such as technical analysts, market researchers, etc. By setting specific technical fields, applicants, legal status and other advanced screening conditions, more accurate and in-depth patent information can be obtained, so as to conduct technical trend analysis, competitor analysis, etc. Finally, cutting-edge semantic search technology is introduced. Semantic search aims to capture the user's query intention, accurately understand the user's query needs through natural language processing technology, and return patent information that is highly relevant to it. This function not only avoids search errors caused by synonyms and near-synonyms, but also helps users quickly find patent data closely related to their field of interest, and realize in-depth data mining and achievement discovery.

[0089] In view of the huge capacity of the patent database, if the search operation is not reasonably restricted, it will directly lead to a long search response time, which will seriously affect the user experience. To this end, the present invention specifically designs three search forms and incorporates a configurable range screening function, which covers core parameters such as the receiving country, time range, and patent type. Under the default setting, this method intelligently limits the search scope to China as the receiving country and focuses on patent data within the past five years. This strategy not only accurately defines the boundaries of the search data, but also effectively controls the amount of data, thereby greatly improving the search efficiency. In addition, in view of the highly flexible dynamic form and index construction mechanism integrated into the data architecture, in order to maintain the integrity and efficiency of data retrieval, we implemented an innovative data distribution mapping strategy that directly maps the physical layout of the data to the search file structure. Under this framework, the search operation is no longer a static query process, but dynamically adjusts its search logic according to the real distribution of data recorded in real time in the search file. Taking Elasticsearch as an example, when querying patent data in China at a certain time in 2024, ideally, the system only needs to locate the index corresponding to the 2024 time period under the Chinese node for retrieval. However, given the surge in patent data that year, our dynamic allocation algorithm made an intelligent decision to create a dual index for the 2024 data to effectively distribute the data load and optimize query performance. This decision was then accurately recorded in the search file as the basis for building the search logic. Therefore, in the actual search process, the system first identifies the dual index layout for the 2024 patent data based on the index allocation scheme recorded in detail in the search file. Subsequently, the search logic is dynamically constructed, using these two indexes as the benchmark for parallel searches, thereby ensuring that all relevant data can be efficiently covered while effectively managing the challenges brought by large-scale data sets. This strategy not only improves the accuracy and speed of retrieval, but also reflects the flexibility and adaptability of the data architecture in dealing with complex data growth patterns.

[0090] After the search is completed, the system efficiently stores the search results in the Redis database, and builds a key-value pair association between the search strategy and the search results. This design enables the system to quickly obtain data directly from Redis when the same search strategy is executed later, greatly shortening the search time and further improving the overall work efficiency. In order to ensure the consistency of the results returned by the three search methods, the present invention also implements a result specification processing mechanism by defining a unified data structure: defining a unified data structure for all search results, such as using JSON format to represent the search results. Ensure that each field has a clear meaning and format. Standardize field names: All search results should use the same field name to represent the same information. Through data type conversion, ensure that the same fields in all search results have the same data type. For numeric fields, specify unified precision and formatting rules. Ensure that the result formats returned by various methods are unified, which facilitates subsequent data display and analysis. The search results include two parts: one part is dynamic statistical information, which calculates the number of relevant data under each category by selecting key fields such as authorization year, acceptance country, application year, legal status, publication year, expiration year, patent type, etc. Users can further filter the results based on this information; the other part is patent-specific field data output based on dynamic custom search headers. The former is implemented through elasticsearch's bucket aggregation technology, while the latter uses a relational database to record the dynamic header structure configured by each user, thereby accurately controlling the final return data of elasticsearch. Next, the specific differences and advantages between these three search methods will be elaborated in detail.

[0091] In general search methods, rapid search and summary search are generally emphasized to meet users' needs for rapid acquisition of key information. However, due to the large number and complexity of patent data fields, it is not realistic to traverse all fields in practice. Therefore, a fuzzy positioning keyword strategy is adopted, which deeply integrates multiple key information fields such as publication number, title, keyword, abstract, etc., and searches and analyzes through these fields. The core advantage of this strategy lies in its excellent search scope control ability, which can quickly and accurately lock the core data that users are most concerned about, and significantly improve the search efficiency and relevance of results. However, its limitations are also obvious. That is, it is impossible to fully cover all data related to the search topic. At the same time, the complex combined search logic may occasionally cause "search illusion", that is, although some search results are actually relevant, the search terms are not directly displayed on the first-view interface of the search results (such as the title), but are hidden in the abstract or original text, affecting the intuitiveness.

[0092] To make up for this shortcoming and further optimize the user experience, a number of intelligent enhancement functions have been innovatively integrated into the search process. Among them, the most notable is the introduction of a field scoring mechanism in the combined search process. For example, for elasticsearch, the title is considered the highest scoring field because it directly reflects the core content of the document. It aims to give priority to displaying results containing search terms in the title, thereby effectively reducing search illusions and allowing users to focus on core information more quickly. Following closely is the summary field, which has a moderate score that takes into account both the generality of the content and the comprehensiveness of the search. Fields such as keywords and publication numbers are given relatively low but indispensable scores based on their specific functions to ensure the integrity and flexibility of the search logic.

[0093] Secondly, before the user enters the search text, the system will automatically display the historical search records to help the user quickly locate the previous search field, saving the user's time and energy. When the user starts to enter the search text, the system will intelligently recommend synonyms and similar words based on the carefully maintained vocabulary in the background, thereby broadening the user's search ideas and increasing the diversity of search results. In addition, an optional translation plug-in function is also provided for users. If the user checks this function, the system will automatically translate the entered search terms into English and display bilingual mixed content in the search results. This function not only meets the user's needs for multilingual retrieval, but also provides users with a richer source of information. Finally, in order to facilitate users to track and manage their own search records, the current search record will be saved after each call to the search interface. In this way, the user can easily view the historical search records and quickly locate the previous search results when searching next time.

[0094] In the advanced search method, unlike the general search method, the conventional restrictions are broken and the practice of searching only based on limited dimensions of terms and fields is abandoned. First, by introducing a wide range of screening options, including authorization year, receiving country, application year, legal status, publication year, expiration year, and patent type, the user's search scope customization rights are greatly expanded. This innovation not only enhances the flexibility of the search, but also significantly improves the relevance and accuracy of the search. Subsequently, a comprehensive and optional field drop-down box is designed, from which users can select specific fields according to their needs and perform fuzzy or exact matching on the contents of these fields. This design allows users to locate the required information more accurately, thereby avoiding the interference of a large amount of irrelevant information and improving the search efficiency. More importantly, it supports users to flexibly link multiple search terms using Boolean operators (such as AND, OR, NOT), and supports configurable fuzzy or exact searches, so as to build a more professional, complex and accurate search strategy. At the implementation level, the method of building entity arrays is adopted, and the Boolean operator recognition and traversal in the code logic are used to build an efficient and accurate search engine logic. This strategy not only improves the accuracy of the search, but also increases the depth and breadth of the search, allowing users to easily obtain more comprehensive and in-depth search results. Overall, by providing a wide range of screening options, flexible field selection, and powerful Boolean operator support, we have successfully built a powerful and flexible advanced search system. The system can fully meet the diverse needs of users in patent searches and provide users with more efficient, accurate, and in-depth search services. Whether in academic research, technological innovation, or business decision-making, the system will play an important role, helping users quickly obtain the information they need and achieve efficient information value conversion.

[0095] In the semantic retrieval method, an intuitive and concise operation experience is provided for users. Users only need to enter the relevant publication number / application number or search text, and the system can quickly return patent search results with high semantic similarity. When the user enters the publication number or application number, the relevant record is located in the patent database. The key fields such as the title and abstract of the patent are deeply analyzed and integrated, and used as the input of the semantic search. In this way, patents similar to the user's query intention are intelligently matched. For the search text input method, the semantic vector is calculated in real time and compared with other data in the database, and the semantic similarity is determined by spatial distance. Considering the possible computing power limitations in the domestic independent and controllable environment, two optimization strategies are proposed. First, for the existing patent data, its semantic vector is generated in advance and stored in the Elasticsearch index. In this way, when searching, the system can directly read the generated vector from the index, which greatly reduces the amount of calculation. Secondly, for the newly added patent data, the semantic vector is generated by enabling asynchronous threads in the background and updated to the index in real time. This strategy not only ensures the timely inclusion of new data, but also avoids excessive occupation of system performance.

[0096] The present invention proposes a distributed patent search system based on a domestically produced autonomous and controllable environment. The system integrates a corpus management module, a system management module and a search module. Figure 2 , providing users with efficient, accurate and flexible patent data retrieval services. The system supports the management of full and incremental patent data, provides a variety of retrieval methods and result display functions, and supports user-defined retrieval optimization strategies to meet the diverse needs of different users. Adopt identity authentication and permission management mechanisms to ensure that only authorized users can access and operate patent data. Encrypt sensitive data for storage and transmission to prevent data leakage and illegal access. Improve the system's response speed and data processing capabilities through load balancing, caching technology, index optimization and other means. Monitor the system's operating status and performance indicators to promptly discover and resolve potential performance issues.

[0097] The corpus module is the core component of the patent data retrieval system, which is responsible for the maintenance, update, query and display of patent data. This module ensures the real-time, accuracy and ease of use of patent data through carefully designed management strategies, providing a solid foundation for subsequent query and analysis. The corpus management module adopts a distributed architecture and achieves high-performance data storage and access through database clustering and caching technology. The system supports multiple storage engines and selects appropriate storage structures for optimization based on the characteristics of patent data and business needs.

[0098] 1. Data update and preservation

[0099] The system provides basic data update operation interfaces such as adding, modifying, and deleting, supports batch processing and transaction control, and ensures data integrity and consistency. Through triggers or scheduled tasks, the changes of patent data sources are monitored in real time, and the update operations of full data and incremental data are automatically completed. When adding new patent data, the system will accurately divide the complex patent data into multiple dimensions, such as title, abstract, inventor, classification, etc. according to the preset storage logic. The data of each dimension is refined, including multimodal data such as text, pictures, and sounds. Distributed storage technology is used to store data in different tables on the same node to improve data processing efficiency and accessibility. During the data storage process, the system will check and verify the data to ensure the accuracy and integrity of the data. A data backup and recovery mechanism is used to back up data regularly to prevent data loss, and a data recovery function is provided to deal with unexpected situations.

[0100] 2. Data Query and Display

[0101] The system supports powerful query functions. Users can perform precise queries by keywords, categories, time and other conditions. It provides a list format to display key fields of data. Users can customize the display fields according to actual needs to meet diverse data viewing needs. The system provides multiple filter items. Users can freely combine these filter items as needed to accurately filter the search results. It supports advanced filtering functions such as fuzzy query and logical operations to improve the flexibility and accuracy of retrieval. It provides a friendly user interface and interactive experience to display key information of patent data in the form of charts, tables and other forms. It supports data export function. Users can export query results to CSV, Excel and other formats for offline analysis.

[0102] 3. Data Release and Withdrawal

[0103] Users can publish the entered data according to actual needs, making it visible on the front-end search side. The system supports batch publishing and scheduled publishing of data to meet the diverse needs of users. Users can withdraw the published data at any time to make it invisible on the front-end search side. The system will record the publishing and withdrawal history of data to facilitate user tracing and management.

[0104] The search module applies the above search methods to meet the diverse needs of users for patent search. Based on the basic search method, this patent search module integrates advanced functions such as search strategy subscription and search scoring mechanism, aiming to provide users with a more intelligent and convenient patent search experience. For details of the search method in this section, please refer to the previous section.

[0105] 1. Search strategy subscription function

[0106] Users can customize the patent information they are interested in and save the search rules as subscription policies. The system regularly pushes relevant patent data based on the user's subscription policy and displays them in clusters.

[0107] 2. Search Scoring Mechanism

[0108] Introduce a search scoring mechanism to score the degree of match between the user's search intent and the actual returned results. Optimize search algorithms and strategies based on scoring data to improve search accuracy and efficiency. Scoring data provides strong support for subsequent system operation and optimization.

[0109] As the core component of the patent search system, the backend management module is responsible for many key tasks, including user search log storage and management, dictionary configuration, search result display and control, patent basic data management, and routine user management.

[0110] 1. User search log storage and management

[0111] Whenever a user performs a search operation, the backend management module will automatically record the user's search behavior, including search keywords, search timestamps, and core information such as the user's unique identifier. The log data will be saved in the system's database for subsequent data analysis and search algorithm optimization. The backend management module will provide data analysis tools that allow administrators to conduct in-depth analysis of user search logs. The analysis results will be used to optimize the search algorithm and improve the accuracy and relevance of the search results.

[0112] 2. Dictionary Configuration

[0113] The backend management module fully integrates diversified dictionary configuration capabilities, covering key categories such as proper noun libraries, synonym sets, and sensitive word lists, aiming to meet the refined needs of different business scenarios. Each dictionary is equipped with an independent, user-friendly configuration interface, allowing users to easily add, fine-tune, and delete dictionary content with one click, greatly enhancing the flexibility and convenience of configuration. Users can flexibly adjust the dictionary content according to their own business characteristics and needs to ensure that the retrieval system can accurately capture business key points.

[0114] The system has a built-in intelligent integration mechanism that can automatically and seamlessly integrate the configured dictionary into the core of the search method. By deeply optimizing the algorithm logic, it can significantly improve the accuracy and efficiency of the search results, bringing users a more accurate and efficient search experience. In addition, the system has also innovatively introduced dictionary pinning and blocking strategies to cope with complex and changing search needs. For blocked words, the system will automatically filter out search results containing these words, effectively avoiding interference from unexpected or sensitive information. For pinned words, the system takes special processing measures to give preferential treatment in the score distribution of search results, especially optimizing the score between the title and the abstract of the results containing pinned words, ensuring that these important information can be displayed first in the search results, further improving the intelligence level of the search service and user experience.

[0115] 3. Search Results Display and Control

[0116] Administrators can customize the top-ranking strategy for search results based on business needs. The system will sort the search results based on the top-ranking strategy to ensure that important information is displayed to users first. Administrators can set blocking rules to filter search results that do not meet the requirements. Blocking rules can be configured based on multiple conditions such as keywords and patent types.

[0117] IV. Patent Basic Data Management

[0118] The backend management module provides a configurable dictionary management function, allowing users to configure different versions of classification codes and other basic patent metadata according to actual needs. Users can customize dictionary fields, values, and version information. The system will automatically record dictionary version information, making it convenient for users to view and switch between different versions of dictionaries. Administrators can publish new dictionary versions as needed and control the scope of use of each version.

[0119] 5. General User Management

[0120] Users can enter the platform through the registration function and set up their personal account and password. Users need to authenticate their identities when logging in to ensure system security. System administrators can dynamically assign operation permissions based on user roles and responsibilities. Permission management will strictly control user access to system resources and the scope of operations to ensure system security and stability.

[0121] 6. Data import and export function

[0122] The backend management module provides data import function and supports the import of multiple data formats (such as CSV, Excel, etc.). Users can select the imported data fields and formats as needed and import the data into the system for analysis and processing. Users can export the processed data to a file with a specified format and field configuration. The system supports multiple data export formats (such as PDF, Word, etc.) to meet the needs of different usage scenarios.

[0123] 7. System Security and Maintenance

[0124] The system will encrypt and store key data to ensure data security. The system will back up data regularly to prevent data loss or damage. The backend management module will record system operation logs, including key operations such as user login and data modification. Administrators can monitor the system operation status through the log audit function to find and solve problems in a timely manner. The system will be regularly upgraded and maintained to fix potential security vulnerabilities and performance issues. Online help documents and customer service support are provided to help users solve problems during use.

[0125] Although the specific embodiments of the present invention are disclosed for the purpose of illustration, the purpose is to help understand the content of the present invention and implement it accordingly, those skilled in the art will understand that various substitutions, changes and modifications are possible without departing from the spirit and scope of the present invention and the appended claims. Therefore, the present invention should not be limited to the content disclosed in the best embodiment, and the scope of the present invention is subject to the scope defined in the claims.

Claims

1. A distributed patent search method based on a domestically produced autonomous and controllable environment, comprising the following steps: 1) Establish a distributed patent data architecture, including a relational database for storing patent structured data, a NoSQL database and a high-capacity storage server for storing patent unstructured data; the relational database uses distributed multi-nodes for information storage; the NoSQL database is provided with a search engine and a Hash data structure for caching search results; 2) The patent data of the same core country is stored in the same node, and the patent data of each non-core country is stored in the remaining nodes; the node is provided with a main table and multiple sub-tables; the main table is used to store the core information of the patent, The subtable is used to store patent information associated with the main table and the primary key of the main table; the search engine extracts the key fields of each patent, the node where the patent data is located and the primary key of the main table, and the primary key information of the secondary table from the patent data of each country, and constructs the index and dictionary items of the patent, wherein the fields used as dictionary items include the receiving country, patent type, legal status information, language, classification number, license type, transfer type, technical field, and unit type; the unstructured data in each patent is stored in the high-capacity storage server and a multi-level index system is created, Saving the index information of the patent unstructured data in the multi-level index system into the index; 3) According to the search information and search mode input by the user, the index is searched for matching patents and the nodes where their structured data are located and the locations where their unstructured data are located.

2. The method according to claim 1, characterized in that The main table and sub-table are horizontally divided according to the publication date timestamp of the patent, the patent data within the same set time period are put into the same group of horizontal sub-tables, and the data volume of each horizontal sub-table is set.

3. The method according to claim 1, characterized in that The node is provided with a dynamic database table and index management mechanism for monitoring the status and data growth of the node. When the data growth reaches a set threshold, a dynamic adjustment process is automatically triggered, including creating a new database table, updating the index or performing table partitioning.

4. The method according to claim 1, 2 or 3, characterized in that: The method for constructing the multi-level index system is as follows: first, multiple first-level folders are created by country, and then multiple second-level folders are rebuilt under the first-level folders by patent type, and third-level folders are organized under the second-level folders in the order of patent publication numbers, and each third-level folder is limited to store a set amount of patent data; The third-level folder is used to store unstructured data of patents.

5. The method according to claim 1, 2 or 3, characterized in that: The core information includes patent title, category, type, receiving country, language, application number, application date, publication number, publication date, authorization number, authorization date, due date, expiration date, agent, agency, substantive examination announcement date, technical field, publication status, views, likes, collections, comments, legal status and payment status information; the patent information stored in the subtable includes abstract information, claim information, classification number cluster information, family member information, patent drawing attachment information, legal information, applicant group information, citation information and other information.

6. The method according to claim 1, 2 or 3, characterized in that: The key fields include the fields involved in the full-text search function and the information fields in the list display; the fields involved in the full-text search function include the associated primary key of the relational database, the unique identifier of the main table and the unique identifier of the subtable; the information fields in the list display include title information, abstract information, country information, classification information, full-text information, application information, attachment information, public information, claim information, and classification number information.

7. The method according to claim 1, 2 or 3, characterized in that: The retrieval modes include general retrieval, advanced retrieval and semantic retrieval.

8. The method according to claim 7, characterized in that The search results of each search mode are standardized to unify their formats.

9. A distributed patent search system based on a domestically produced autonomous and controllable environment, characterized in that: It includes corpus management module, system management module and retrieval module; The corpus management module adopts a distributed patent data architecture, including a relational database for storing patent structured data, a NoSQL database and a high-capacity storage server for storing patent unstructured data; the relational database adopts distributed multi-nodes for information storage; the NoSQL database is provided with a search engine and a Hash data structure for caching search results; The system management module is used to store patent data of the same core country in the same node, and to store patent data of non-core countries in the remaining nodes; a main table and multiple sub-tables are provided in the node; the main table is used to store core information of patents, and the sub-table is used to store patent information associated with the main table and the primary key of the main table; the search engine extracts key fields of each patent, the node where the patent data is located and the primary key of the main table, and the primary key information of the secondary table from the patent data of each country, and constructs the index and dictionary items of the patent, wherein the fields used as dictionary items include the receiving country, patent type, legal status information, language, classification number, license type, transfer type, technical field, and unit type; the unstructured data in each patent is stored in the high-capacity storage server and a multi-level index system is created, and the index information of the unstructured data of the patent in the multi-level index system is saved in the index; The search module is used to search for matching patents and nodes where structured data is located and locations where unstructured data is located in the index according to search information and search patterns input by the user.

Images