Intelligent retrieval system and method based on ontology model and data fusion

Abstract

The invention provides an intelligent retrieval system and method based on an ontology model and data fusion, relates to the field of information retrieval and data processing, and solves the problems of low efficiency and poor user experience in the aspects of data cleaning, ontology design and result presentation in the traditional technology. The system comprises an administrator login and configuration module used for an administrator to log in and configure a data optimization processing tool; the domain expert ontology design module is used for designing and managing the ontology model by domain experts and storing ontology data; the user document uploading and analyzing module is used for receiving a document and analyzing the document into structured data; the graph analysis module is used for generating a knowledge graph; the common retrieval module is used for matching subject fields corresponding to user questions and obtaining retrieval results; the advanced retrieval and graph query module is used for matching corresponding entities from the knowledge graph on the basis of common retrieval and fusing to form an advanced retrieval result; the field specialty and accuracy of the intelligent retrieval system are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of information retrieval and data processing, and in particular to an intelligent retrieval system and method based on ontology model and data fusion. Background Art

[0002] With the rapid development of Internet technology, the amount of information has exploded. How to effectively extract valuable information from massive data has become a major challenge facing modern information technology. Among many information processing technologies, information retrieval technology is particularly important because it directly serves the daily needs of users. However, existing retrieval systems often show certain limitations when facing specific fields or specialized data. This is mainly due to the lack of in-depth understanding and processing capabilities of these systems for specific field knowledge.

[0003] Specifically, when users make query requests that contain professional terms or interdisciplinary concepts, traditional retrieval systems may not be able to accurately understand the query intent, thus affecting the relevance and accuracy of the retrieval results. For example, in professional fields such as healthcare and legal consulting, the accuracy and authority of information are crucial, but existing retrieval tools are difficult to meet such high-demand application scenarios. In addition, with the advent of the big data era, data sources are becoming increasingly diverse, and multiple types of data including text, images, and videos need to be integrated into the same retrieval framework, which places higher demands on the comprehensive processing capabilities of the retrieval system.

[0004] In addition to the lack of understanding of professional field data, existing retrieval systems also face many challenges in the data preprocessing stage. Data cleaning is an important part of ensuring retrieval quality, which involves a series of complex operations such as removing redundant information, correcting erroneous data, and standardizing heterogeneous data formats. However, most existing retrieval systems have a low degree of automation in data cleaning and rely on manual intervention, which not only increases costs but also may cause delays in data processing. At the same time, as a key bridge connecting user queries and backend databases, the quality and adaptability of ontology design directly affect the performance of the retrieval system. A well-designed ontology model can help the system better understand user needs and filter out the most relevant information from a large amount of data. However, there are currently few retrieval systems on the market that can flexibly and efficiently apply ontology models, making this process time-consuming and laborious.

[0005] Finally, the presentation of search results is also one of the key factors affecting user experience. The ideal result display should be intuitive and clear to users, so that they can quickly locate the required information. However, many existing search systems are too simplistic in the presentation of results, ignoring the friendliness and interactivity of the user interface, resulting in users being frustrated because they cannot find the required answers even if relevant information is retrieved. In summary, the development of an intelligent search system that can overcome related defects and has the ability to apply deep learning technology and powerful data processing has become an urgent problem to be solved in the current information technology field. Summary of the invention

[0006] The purpose of the present invention is to solve the problems of low efficiency and poor user experience in traditional retrieval systems in data cleaning, ontology design and result presentation, and therefore proposes an intelligent retrieval system and method based on ontology model and data fusion. The present invention effectively improves the field expertise and accuracy of the intelligent retrieval system by combining the ontology model with big data technology.

[0007] The present invention adopts the following technical solutions to achieve the purpose:

[0008] An intelligent retrieval system based on ontology model and data fusion, the intelligent retrieval system includes the following modules:

[0009] Administrator login and configuration module, used to provide authentication for administrators when they log into the system and support system management authority allocation, and also configure data optimization processing tools;

[0010] The domain expert ontology design module is used to provide domain experts with support for designing, uploading and managing ontology models based on business needs, and store the corresponding ontology data;

[0011] User document upload and parsing module, used to receive and parse documents uploaded by users and store corresponding structured data;

[0012] The graph analysis module is used to generate a knowledge graph based on ontology data and structured data, and provide users with a visual interface of the knowledge graph;

[0013] The general search module is used to split and classify the user's query questions using a large model when the user performs a search query. It only searches the stored structured data and matches the subject field corresponding to the query question to obtain the general search results and return them to the user.

[0014] The advanced search and graph query module is used when a user performs a search query. First, the ordinary search module is used to match the subject field corresponding to the query question, and then the corresponding entity is matched from the knowledge graph based on the subject field to obtain the relationship between the entities, which is then integrated into an advanced search result and returned to the user.

[0015] Specifically, the administrator login and configuration module is used to enable the administrator to obtain system management permissions when logging into the system based on the identity authentication mechanism, and provide the administrator with configuration support for data optimization processing tools.

[0016] Furthermore, the data optimization processing tool includes a data collection tool and a data cleaning tool; the administrator login and configuration module is also used to: enable the administrator to migrate the structured data from the user document upload and parsing module to the retrieval database through the data collection tool and the data cleaning tool; and provide data cleaning rule setting support for the data cleaning tool.

[0017] Furthermore, the domain expert ontology design module is used to provide domain experts with support for the definition of concepts, attributes and relationships in the ontology model, and after the domain experts complete the design and uploading of the ontology model, parse and store the uploaded ontology model to obtain the corresponding ontology data; at the same time, the corresponding version of each uploaded ontology model is independently stored and management support is provided.

[0018] Specifically, the user document upload and parsing module is used to provide cloud storage services for users. After users upload documents in various formats through the cloud storage service, the module parses the documents, extracts key information from the documents, forms structured data, and stores it in the retrieval database. It also provides management functions for uploaded documents.

[0019] Furthermore, the structured data is stored in the Elasticsearch and / or Milvus retrieval database; the administrator login and configuration module is also used to migrate the data in the original database to the TIDB and / or Elasticsearch retrieval database.

[0020] Furthermore, the general search module is used to determine whether there is a subject field that matches the query question in the TIDB search database based on the split and classified query question; if the subject field exists in the TIDB search database, the large model is used again to optimize the description of the content contained in the subject field, obtain the general search results and return them to the user;

[0021] If the subject field does not exist in the TIDB search database, a hybrid search is performed on the Elasticsearch and / or Milvus search databases, and a weight calculation is used to determine whether the subject field exists. If so, the large model is used again to optimize the description of the content contained in the subject field, obtain normal search results, and return them to the user.

[0022] Furthermore, the advanced search and graph query module is used to find specific entities that match the subject field and query conditions from the knowledge graph after the subject field is obtained, based on the query conditions of the query question; then, based on the specific needs of the query question, the relationship of the specific entity is extracted from the knowledge graph to generate detailed query results; the advanced search and graph query module is also used to reuse the large model to encapsulate and optimize the detailed query results, merge them into advanced search results and return them to the user.

[0023] The present invention also provides an intelligent retrieval method based on ontology model and data fusion, the hardware basis of the method is the aforementioned intelligent retrieval system; the method comprises the following steps:

[0024] S1. The system administrator logs in to the system after identity authentication, obtains system authority allocation, and configures data optimization processing tools;

[0025] S2. Domain experts design and upload ontology models based on business needs, and the system stores the corresponding ontology data;

[0026] S3, obtains and parses the documents uploaded by the user, and stores the corresponding structured data;

[0027] S4. Generate a knowledge graph based on ontology data and structured data, and present a visual interface of the knowledge graph;

[0028] S5. When a user performs a general search, the large model is used to split and classify the user's query questions. By only searching the stored structured data, after matching the subject field corresponding to the query question, the general search results are obtained and returned to the user;

[0029] S6. When the user performs an advanced search, the user first uses the same method as the ordinary search in step S5 to match the subject field corresponding to the query question, then matches the corresponding entity from the knowledge graph based on the subject field, obtains the entity relationship, and integrates it into an advanced search result and returns it to the user.

[0030] Further, in step S1, the system administrator uses the configured data optimization processing tool to migrate the existing structured data in the system to the TIDB retrieval database; in step S3, the structured data corresponding to the document uploaded by the user is stored in the Elasticsearch and / or Milvus retrieval database;

[0031] In step S5, based on the split and classified query questions, it is determined whether there is a subject field that matches the query question in the TIDB search database; if the subject field exists in the TIDB search database, the large model is used again to optimize the description of the content contained in the subject field, obtain the general search results and return them to the user;

[0032] If the subject field does not exist in the TIDB search database, a hybrid search is performed on the Elasticsearch and / or Milvus search databases, and a weight calculation method is used to determine whether the subject field exists. If so, the large model is used again to optimize the description of the content contained in the subject field, obtain the general search results, and return them to the user;

[0033] In step S6, after the subject field is obtained through ordinary search matching, based on the query conditions of the query question, the specific entity that matches the subject field and the query conditions is found from the knowledge graph; then, based on the specific needs of the query question, the relationship of the specific entity is extracted from the knowledge graph to generate detailed query results; the large model is used again to encapsulate and optimize the detailed query results, merge them into advanced search results and return them to the user.

[0034] In summary, due to the adoption of this technical solution, the beneficial effects of the present invention are as follows:

[0035] The present invention effectively improves the domain expertise and accuracy of the retrieval system by combining the ontology model and big data technology. In a specific field, the ontology model, as a structured knowledge representation method, can effectively capture the concepts in the field and the relationships between them, thereby helping the system to more deeply understand the user's query intent. This deep understanding not only improves the relevance of the retrieval results, but also enables the system to perform better when processing complex and specialized data. In addition, by integrating big data processing technology, the present invention can efficiently manage large-scale data sets, ensuring good performance even when the amount of data surges.

[0036] In order to further improve the data quality and system flexibility, the present invention introduces data cleaning and configuration functions. The data cleaning function can automatically identify and correct errors in the data, standardize the data formats of different sources, and thus ensure the quality of input data. At the same time, the configuration function also allows users to customize search parameters according to their own needs, enhancing the adaptability and ease of use of the system. This means that whether in academic research or commercial applications, the present invention can provide more personalized and precise services.

[0037] In terms of data visualization and user interaction experience, the present invention adds a graph analysis function. This function can not only display complex search results in an intuitive graphical form, but also support users to explore the relationship between data through interactive operations. High-quality visual presentation and smooth interactive design can greatly improve user satisfaction, enable users to find the required information faster, and also provide a new perspective for data analysis.

[0038] The application of large model technology in the present invention also optimizes the retrieval process and result presentation. The large model has strong learning and generalization capabilities, and can automatically optimize retrieval-related algorithms with less human intervention, and continuously improve the retrieval effect. In this way, the present invention not only improves the intelligence level of the system, but also simplifies maintenance work and reduces operating costs. It can be seen that the intelligent retrieval system of the present invention effectively improves user experience and service quality through a combination of technological innovations. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a schematic diagram of the components of the intelligent retrieval system of the present invention;

[0040] Figure 2 It is a schematic diagram of the overall process of the intelligent retrieval method of the present invention. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0042] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0043] Example 1

[0044] An intelligent retrieval system based on ontology model and data fusion, the intelligent retrieval system includes the following modules, which can be viewed simultaneously Figure 1 The hint:

[0045] Administrator login and configuration module, used to provide authentication for administrators when they log into the system and support system management authority allocation, and also configure data optimization processing tools;

[0046] The domain expert ontology design module is used to provide domain experts with support for designing, uploading and managing ontology models based on business needs, and store the corresponding ontology data;

[0047] User document upload and parsing module, used to receive and parse documents uploaded by users and store corresponding structured data;

[0048] The graph analysis module is used to generate a knowledge graph based on ontology data and structured data, and provide users with a visual interface of the knowledge graph;

[0049] The general search module is used to split and classify the user's query questions using a large model when the user performs a search query. It only searches the stored structured data and matches the subject field corresponding to the query question to obtain the general search results and return them to the user.

[0050] The advanced search and graph query module is used when a user performs a search query. First, the ordinary search module is used to match the subject field corresponding to the query question, and then the corresponding entity is matched from the knowledge graph based on the subject field to obtain the relationship between the entities, which is then integrated into an advanced search result and returned to the user.

[0051] This embodiment will introduce each of the above modules in detail.

[0052] In the administrator login and configuration module, the intelligent retrieval system uses an authentication mechanism to ensure that administrators can log in safely and obtain system management permissions. At the same time, administrators can configure data optimization and processing tools, including data collection and data cleaning, such as Kettle tools, so that existing data in the original database can be efficiently and accurately migrated to TIDB and / or Elasticsearch retrieval databases.

[0053] TIDB is an open source distributed NewSQL database that is compatible with the MySQL protocol and is designed to provide high-performance solutions for mixed loads of online transaction processing and online analytical processing. TIDB is designed to achieve horizontal expansion, strong consistency, and high availability, and is particularly suitable for processing large-scale data sets and high-concurrency access scenarios.

[0054] In addition, this module also provides data cleaning rule settings for data cleaning tools, including data deduplication, format unification, and outlier processing, to improve data quality.

[0055] In the domain expert ontology design module, domain experts can design ontology models based on business requirements and define the concepts, attributes, and relationships therein. Domain experts are people with deep expertise and experience in a specific field. They have a deep understanding of the terminology, concepts, rules, and practices in the field. The ontology model represents the knowledge in the field in a structured way; it provides a clear knowledge framework by defining and organizing the concepts, attributes, and relationships in the field. The ontology model can help intelligent retrieval systems better understand and process data in the field.

[0056] Among them, concepts refer to basic entities or objects in a field. For example, in the medical field, concepts may include "disease", "symptoms", "drugs", etc. Attributes describe the specific characteristics or properties of concepts. For example, the concept of "disease" may have attributes such as "name", "cause", and "treatment method". Relationships represent the association between concepts. For example, there may be a "cause" or "accompany" relationship between "disease" and "symptoms", and there may be a "treatment" relationship between "disease" and "drug".

[0057] In this embodiment, after the domain experts complete the design and upload of the ontology model, the module can parse and store the uploaded ontology model to obtain the corresponding ontology data; at the same time, the corresponding version of each uploaded ontology model is independently stored and management support is provided, so as to facilitate the domain experts to iteratively optimize the ontology model.

[0058] In the user document upload and parsing module, users upload documents in various formats through the cloud storage service, which can be MinIO object storage service. This module parses the document, extracts key information to obtain the corresponding structured data, and stores the structured data in Elasticsearch and / or Milvus retrieval database. Elasticsearch itself is an open source distributed search and analysis engine, which is widely used in full-text search, log analysis, real-time data analysis and other scenarios. It is built on Apache Lucene and provides a RESTful API interface, making integration and use very convenient. Elasticsearch is designed to process large-scale data sets and provide high-performance search and analysis capabilities. Milvus is an open source vector similarity search engine that is specifically used to process similarity searches on large-scale vector data. Milvus supports a variety of vector indexing algorithms and can efficiently process high-dimensional vector data. It is widely used in recommendation systems, image recognition, natural language processing and other fields.

[0059] The user document upload and parsing module also provides management functions for the document itself, including classification, labeling and permission settings, etc. This function can be selectively used by system administrators and users.

[0060] In the graph analysis module, a knowledge graph is generated based on ontology data and structured data, and a visual interface is provided to help users intuitively view and analyze the relationship between data. It also supports the export and sharing of the knowledge graph.

[0061] In the general search module, based on the split and classified query questions, it is determined whether there is a subject field in the TIDB search database that matches the query question; if the subject field exists in the TIDB search database, the large model is used again to optimize the description of the content contained in the subject field, obtain the general search results and return them to the user;

[0062] If the subject field does not exist in the TIDB search database, a hybrid search is performed on the Elasticsearch and / or Milvus search databases, and a weight calculation is used to determine whether the subject field exists. If so, the large model is used again to optimize the description of the content contained in the subject field, obtain normal search results, and return them to the user.

[0063] In the advanced search and graph query module, when the module needs to be used, the intelligent search system first performs a general search. After obtaining the subject field, based on the query conditions of the query question, it finds the specific entity that matches the subject field and query conditions from the knowledge graph; then, based on the specific needs of the query question, it extracts the relationship of the specific entity from the knowledge graph to generate detailed query results. Subsequently, the module uses the large model again to encapsulate and optimize the detailed query results, merge them into advanced search results and return them to the user. The encapsulated and optimized advanced search results will be presented to the user in a more intuitive and easy-to-understand way, to maximize the satisfaction of the user's query request.

[0064] Example 2

[0065] Based on Example 1, this example introduces an intelligent retrieval method based on ontology model and data fusion. The hardware basis of this method is the intelligent retrieval system in Example 1. The overall process of this method can be found in Figure 2 The steps are summarized as follows:

[0066] S1. The system administrator logs in to the system after identity authentication, obtains system authority allocation, and configures data optimization processing tools;

[0067] S2. Domain experts design and upload ontology models based on business needs, and the system stores the corresponding ontology data;

[0068] S3, obtains and parses the documents uploaded by the user, and stores the corresponding structured data;

[0069] S4. Generate a knowledge graph based on ontology data and structured data, and present a visual interface of the knowledge graph;

[0070] S5. When a user performs a general search, the large model is used to split and classify the user's query questions. By only searching the stored structured data, after matching the subject field corresponding to the query question, the general search results are obtained and returned to the user;

[0071] S6. When the user performs an advanced search, the user first uses the same method as the ordinary search in step S5 to match the subject field corresponding to the query question, then matches the corresponding entity from the knowledge graph based on the subject field, obtains the entity relationship, and integrates it into an advanced search result and returns it to the user.

[0072] In step S1 of this embodiment, the system administrator uses the configured data optimization processing tool to migrate the existing structured data in the system to the TIDB retrieval database.

[0073] In step S3 of this embodiment, the structured data corresponding to the document uploaded by the user is stored in the Elasticsearch and / or Milvus retrieval database.

[0074] In step S5 of this embodiment, based on the split and classified query question, it is determined whether there is a subject field in the TIDB search database that matches the query question; if the subject field exists in the TIDB search database, the large model is used again to optimize the description of the content contained in the subject field, obtain the general search results and return them to the user;

[0075] If the subject field does not exist in the TIDB search database, a hybrid search is performed on the Elasticsearch and / or Milvus search databases, and a weight calculation is used to determine whether the subject field exists. If so, the large model is used again to optimize the description of the content contained in the subject field, obtain normal search results, and return them to the user.

[0076] In step S6 of this embodiment, after the subject field is obtained through ordinary search matching, based on the query conditions of the query question, the specific entity that matches the subject field and the query conditions is found from the knowledge graph; then, based on the specific needs of the query question, the relationship of the specific entity is extracted from the knowledge graph to generate detailed query results; the large model is used again to encapsulate and optimize the detailed query results, merge them into advanced search results and return them to the user.

[0077] Example 3

[0078] Based on any of the above embodiments, this embodiment uses the case of an intelligent retrieval system in the medical field to exemplify the situation of the system of the present invention. The background requirement of this example is: a large hospital hopes to improve the efficiency and accuracy of internal medical knowledge retrieval so that doctors and researchers can quickly obtain relevant cases, research reports and drug information. Therefore, in this example, the users are doctors and researchers.

[0079] First, for administrator login and configuration: the system administrator logs in to the system through identity authentication, configures data collection tools such as Kettle, and then migrates case data, research reports and other data in the hospital's internal database to TIDB and / or Elasticsearch retrieval databases. The administrator also sets data cleaning rules, such as removing duplicate case records and unifying medical record formats, to ensure data quality.

[0080] Medical experts then design the ontology model based on the hospital's business needs, defining concepts such as diseases, symptoms, examinations, treatments, and their attributes and relationships. The intelligent retrieval system automatically parses and stores the ontology model, and domain experts can manage and optimize the ontology version at any time.

[0081] In the user document upload and analysis part, doctors and researchers upload medical records, research reports and other documents through MinIO object storage service. The intelligent retrieval system then parses the documents, extracts key information such as disease name, treatment plan, etc., forms structured data and stores it in Elasticsearch and Milvus retrieval databases.

[0082] The intelligent retrieval system then performs knowledge graph analysis, generates a medical knowledge graph based on ontological data and structured data, and displays the relationship between diseases, symptoms, examinations, and treatments through a visual interface. Doctors and researchers can quickly understand disease-related knowledge and research progress through the graph analysis function.

[0083] The normal search is performed as follows: when a doctor uses the intelligent search system to query the treatment plan for a certain disease, the system uses the big model to split and classify the question and determine whether it corresponds to the subject field in TIDB. If there is relevant data in TIDB, the system directly returns the treatment plan; otherwise, the system performs a hybrid search from Elasticsearch and Milvus and uses the big model to optimize the response content.

[0084] Advanced search and graph query are performed as follows: Doctors perform advanced search, such as querying all relevant cases and research reports of a certain disease. The system first performs a general search, then queries related entities and relationships from the graph to generate detailed query results. Finally, the detailed query results are encapsulated through the big model and presented to the doctor in an intuitive and easy-to-understand way.

[0085] Through the above scheme, the intelligent retrieval system can play an important role in the medical field, improve the work efficiency of medical staff, and provide patients with better medical services.

Claims

1. An intelligent retrieval system based on ontology model and data fusion, characterized in that: The intelligent retrieval system includes the following modules: Administrator login and configuration module, used to provide authentication for administrators when they log into the system and support system management authority allocation, and also configure data optimization processing tools; The domain expert ontology design module is used to provide domain experts with support for designing, uploading and managing ontology models based on business needs, and store the corresponding ontology data; User document upload and parsing module, used to receive and parse documents uploaded by users and store corresponding structured data; The graph analysis module is used to generate a knowledge graph based on ontology data and structured data, and provide users with a visual interface of the knowledge graph; The general search module is used to split and classify the user's query questions using a large model when the user performs a search query. It only searches the stored structured data and matches the subject field corresponding to the query question to obtain the general search results and return them to the user. The advanced search and graph query module is used when a user performs a search query. First, the ordinary search module is used to match the subject field corresponding to the query question, and then the corresponding entity is matched from the knowledge graph based on the subject field to obtain the relationship between the entities, which is then integrated into an advanced search result and returned to the user.

2. The intelligent retrieval system according to claim 1, characterized in that: The administrator login and configuration module is used to enable the administrator to obtain system management permissions when logging into the system based on the identity authentication mechanism, and provide the administrator with configuration support for data optimization processing tools.

3. The intelligent retrieval system according to claim 2, characterized in that: Data optimization and processing tools include data collection tools and data cleaning tools; the administrator login and configuration module is also used to: enable the administrator to migrate structured data from the user document upload and parsing module to the retrieval database through the data collection tool and the data cleaning tool; and provide data cleaning rule setting support for the data cleaning tool.

4. The intelligent retrieval system according to claim 1, characterized in that: The domain expert ontology design module is used to provide domain experts with support for the definition of concepts, attributes and relationships in the ontology model. After the domain experts complete the design and upload of the ontology model, the uploaded ontology model is parsed and stored to obtain the corresponding ontology data. At the same time, the corresponding version of each uploaded ontology model is independently stored and management support is provided.

5. The intelligent retrieval system according to claim 1, characterized in that: The user document upload and parsing module is used to provide cloud storage services for users. After users upload documents in various formats through the cloud storage service, the module parses the documents, extracts key information from the documents, forms structured data, and stores it in the retrieval database. It also provides management functions for uploaded documents.

6. The intelligent retrieval system according to claim 5, characterized in that: Structured data is stored in Elasticsearch and / or Milvus retrieval databases; the administrator login and configuration module is also used to migrate data from the original database to TIDB and / or Elasticsearch retrieval databases.

7. The intelligent retrieval system according to claim 6, characterized in that: The general search module is used to determine whether there is a subject field that matches the query question in the TIDB search database based on the split and classified query question; if the subject field exists in the TIDB search database, the large model is used again to optimize the description of the content contained in the subject field, obtain the general search results and return them to the user; If the subject field does not exist in the TIDB search database, a hybrid search is performed on the Elasticsearch and / or Milvus search databases, and a weight calculation is used to determine whether the subject field exists. If so, the large model is used again to optimize the description of the content contained in the subject field, obtain normal search results, and return them to the user.

8. The intelligent retrieval system according to claim 7, characterized in that: The advanced search and graph query module is used to find the specific entities that match the subject field and query conditions from the knowledge graph based on the query conditions of the query question after the subject field is obtained; then, based on the specific needs of the query question, the relationship of the specific entity is extracted from the knowledge graph to generate detailed query results; the advanced search and graph query module is also used to reuse the large model to encapsulate and optimize the detailed query results, merge them into advanced search results and return them to the user.

9. An intelligent retrieval method based on ontology model and data fusion, characterized in that: The hardware basis of the method is the intelligent retrieval system according to any one of claims 1 to 8; the method comprises the following steps: S1. The system administrator logs in to the system after identity authentication, obtains system authority allocation, and configures data optimization processing tools; S2. Domain experts design and upload ontology models based on business needs, and the system stores the corresponding ontology data; S3, obtains and parses the documents uploaded by the user, and stores the corresponding structured data; S4. Generate a knowledge graph based on ontology data and structured data, and present a visual interface of the knowledge graph; S5. When a user performs a general search, the large model is used to split and classify the user's query questions. By only searching the stored structured data, after matching the subject field corresponding to the query question, the general search results are obtained and returned to the user; S6. When the user performs an advanced search, the user first uses the same method as the ordinary search in step S5 to match the subject field corresponding to the query question, then matches the corresponding entity from the knowledge graph based on the subject field, obtains the entity relationship, and integrates it into an advanced search result and returns it to the user.

10. The intelligent retrieval method according to claim 9, characterized in that: In step S1, the system administrator uses the configured data optimization processing tool to migrate the existing structured data in the system to the TIDB retrieval database; in step S3, the structured data corresponding to the document uploaded by the user is stored in the Elasticsearch and / or Milvus retrieval database; In step S5, based on the split and classified query questions, it is determined whether there is a subject field that matches the query question in the TIDB search database; if the subject field exists in the TIDB search database, the large model is used again to optimize the description of the content contained in the subject field, obtain the general search results and return them to the user; If the subject field does not exist in the TIDB search database, a hybrid search is performed on the Elasticsearch and / or Milvus search databases, and a weight calculation method is used to determine whether the subject field exists. If so, the large model is used again to optimize the description of the content contained in the subject field, obtain the general search results, and return them to the user; In step S6, after the subject field is obtained through ordinary search matching, based on the query conditions of the query question, the specific entity that matches the subject field and the query conditions is found from the knowledge graph; then, based on the specific needs of the query question, the relationship of the specific entity is extracted from the knowledge graph to generate detailed query results; the large model is used again to encapsulate and optimize the detailed query results, merge them into advanced search results and return them to the user.

Images