Tobacco agriculture standard knowledge graph application system and visualization method

Abstract

The invention relates to the technical field of agricultural information, in particular to a tobacco agricultural standard knowledge graph application system and a visualization method. A standard management and query module is used for classified storage and management of tobacco agricultural standards; the standard knowledge graph module is used for dynamically displaying entity nodes and relationships thereof through a visual tool and supporting dynamic addition of the nodes and the relationships; the standard knowledge graph construction module is used for automatically extracting an entity-relation-entity triple from a standard text, storing triple data in a Neo4j graph database and generating a knowledge graph; the standard entity relationship query module is used for generating a Cyber query statement according to a head entity and a relationship type input by a user to retrieve from the graph database; the standard knowledge question and answer module is used for receiving natural language questions input by a user and generating professional answers in combination with a large language model and a graph database query result. According to the method, efficient retrieval, dynamic updating and accurate knowledge service of the tobacco agricultural standard are realized.

Description

Technical Field

[0001] The present invention relates to the field of agricultural information technology, and in particular to a tobacco agricultural standard knowledge graph application system and a visualization method. Background Art

[0002] In the tobacco agriculture sector, the management and application of standards are of great significance for ensuring tobacco quality and regulating production processes. However, existing standard management technologies have the following shortcomings:

[0003] 1. Low data storage and retrieval efficiency: Traditional standard information is usually stored in relational databases in the form of tables or documents. This storage method makes it difficult to intuitively display the complex relationships between standards, resulting in low retrieval efficiency and inability to meet the needs of fast queries.

[0004] 2. Insufficient knowledge relevance: Existing technologies lack structured extraction and association analysis of entities in standards (such as technical elements, clause elements, etc.) and their relationships, resulting in the use of standard knowledge remaining superficial and unable to explore deep-level implicit association rules.

[0005] 3. Poor dynamic expansion capability: With the continuous updating and expansion of tobacco agricultural standards, traditional databases need to frequently modify table structures, resulting in high maintenance costs and difficulty in flexibly responding to dynamic changes in standard content.

[0006] 4. Limited natural language processing capabilities: Existing standard question-answering systems usually rely on a single natural language processing technology, which is prone to knowledge hallucination problems (i.e., fabricating non-existent knowledge) and cannot simultaneously guarantee the accuracy of answers and the professionalism of the language. Summary of the Invention

[0007] The present invention aims to solve the problem that traditional standard management methods rely on tables or relational databases to store data, which makes it difficult to intuitively display complex relationships and lacks dynamic expansion capabilities and efficient natural language question-answering functions. A tobacco agricultural standard knowledge graph application system and visualization method are proposed. Through an intelligent question-answering system based on Neo4j knowledge graph storage, multi-technology collaborative information extraction and LLM fusion, efficient retrieval, dynamic update and precise knowledge services of tobacco agricultural standards are achieved.

[0008] In order to achieve the above purpose, the technical solutions adopted are:

[0009] The present invention provides a tobacco agricultural standard knowledge graph application system, which includes the following modules:

[0010] The standards management and query module is used to classify, store and manage tobacco agricultural standards according to national standards, industry standards, local standards and enterprise standards, and supports viewing and downloading standard PDF documents;

[0011] Standard knowledge graph module, used to dynamically display entity nodes and their relationships through visualization tools, and supports dynamic addition of nodes and relationships;

[0012] A standard knowledge graph construction module is used to automatically extract "entity-relationship-entity" triples from standard texts, store the structured triple data of tobacco agricultural standards in a Neo4j graph database, and generate a knowledge graph;

[0013] Standard entity-relationship query module, which generates Cypher query statements based on the head entity and relationship type input by the user, and retrieves and visualizes related triples from the graph database;

[0014] The standard knowledge question-and-answer module is used to receive natural language questions input by users, generate professional answers by combining the large language model and graph database query results, and display relevant knowledge graphs.

[0015] According to the tobacco agricultural standards knowledge graph application system of the present invention, further, the standard management and query module also includes:

[0016] The standard information table is stored in the MySQL database and contains the standard number, standard name, responsible department, status, and the URL of the PDF document stored in MinIO;

[0017] The standard search function is implemented based on database LIKE query, allowing users to quickly locate target standards by standard number or standard name.

[0018] According to the tobacco agriculture standard knowledge graph application system of the present invention, further, the standard knowledge graph module supports the dynamic addition of nodes and relationships in the following ways:

[0019] The user enters the node name and type through the interface, and the system automatically generates a Cypher statement to create a new node;

[0020] The user selects the source node, target node, and relationship type, and the system automatically generates a Cypher statement to create a new relationship;

[0021] Neo4j's APOC extension library is used to support the creation and modification of dynamic nodes and relationships.

[0022] According to the tobacco agricultural standard knowledge graph application system of the present invention, further, the standard knowledge graph construction module embeds a standard information element extraction algorithm, a standard technical element extraction algorithm and a standard article element extraction algorithm. The standard information element extraction algorithm is implemented based on a large language model, the standard technical element extraction algorithm is implemented based on regular matching, and the standard article element extraction algorithm implements named entity recognition and relationship extraction based on a deep learning model.

[0023] According to the tobacco agricultural standard knowledge graph application system of the present invention, further, the standard entity relationship query module implements the query function through the following steps:

[0024] The user enters the header entity and relationship type, and the system converts it into a Cypher query statement;

[0025] The Cypher query statement is sent to the Neo4j graph database through the py2neo library for execution, and the Neo4j graph database returns triple data;

[0026] Use cytoscape.js to display the query results in the form of a visual map.

[0027] According to the tobacco agricultural standard knowledge graph application system of the present invention, further, the workflow of the standard knowledge question-answering module includes:

[0028] The user selects the question type and enters a natural language question;

[0029] The system uses LLM to identify entities and intents in the question and generates corresponding Cypher query statements;

[0030] Retrieve relevant triples from the graph database, then use LLM to sort out the knowledge of the triples retrieved from the graph database and generate structured answers;

[0031] The answer and related knowledge graph are displayed at the same time.

[0032] According to the tobacco agriculture standard knowledge graph application system of the present invention, further, the entities and intentions in the standard knowledge question-answering module are recognized and implemented in the following two technical categories:

[0033] For standard information and standard technical elements, the entities and intentions in user questions are directly identified through LLM;

[0034] For standard clause elements, deep learning models are used to identify entities in the clauses, and PromptEngineering is used to guide LLM to parse the intent.

[0035] Furthermore, the present invention also provides a visualization method for a tobacco agricultural standard knowledge graph application system, comprising:

[0036] Extract information elements, technical elements and clause elements from the tobacco agricultural standard text to form structured triples;

[0037] Import triple data into Neo4j graph database to build knowledge graph;

[0038] Dynamically display the knowledge graph through visualization tools, and support dynamic addition of nodes and relationships;

[0039] Retrieve and return relevant knowledge from the knowledge graph based on user queries or natural language questions.

[0040] The beneficial effects achieved by adopting the above technical solution are:

[0041] 1. Knowledge graph storage based on Neo4j graph database: Neo4j graph database is used to store standard triple data, and its node and edge connections are used to intuitively display the relationship between entities, significantly improving the efficiency of knowledge organization and retrieval speed.

[0042] 2. Multi-technology collaborative information extraction and dynamic updating: This system extracts standard information elements by combining a large language model, extracts standard technical elements based on regular expression matching, and selects a deep learning model to extract standard clause elements. Furthermore, through the APOC extension library and dynamic Cypher statements, it supports dynamic expansion and real-time updating of standard knowledge.

[0043] 3. Intelligent question-answering system integrating large language model (LLM): Combining knowledge graph with LLM (such as Kimi), using LLM's natural language generation capability to solve the problem of low flexibility in language processing of knowledge graph, while ensuring the accuracy of answers through knowledge graph, avoiding the "knowledge illusion" problem caused by relying solely on LLM.

[0044] 4. Modular system design: The system is divided into modules such as standard management and query, standard knowledge graph construction, standard entity relationship query, and standard knowledge question and answer. It supports classified storage of standard PDFs, visual display of knowledge graphs, dynamic interactive operations, and natural language question and answer, realizing the efficient management and application of tobacco agricultural standard knowledge.

[0045] Through the above-mentioned technology, the present invention realizes the accurate retrieval, dynamic update and intelligent question-answering of tobacco agricultural standard knowledge, significantly improving the efficiency and application value of standard management. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings of the embodiments of the present invention. The drawings are only used to illustrate some embodiments of the present invention, but not to limit all embodiments of the present invention thereto.

[0047] Figure 1 This is a structural block diagram of the tobacco agricultural standard knowledge graph application system according to the first embodiment of the present invention;

[0048] Figure 2 It is the standard knowledge graph construction module embedded algorithm of the first embodiment of the present invention;

[0049] Figure 3This is the implementation process of the standard knowledge question and answer module in the first embodiment of the present invention;

[0050] Figure 4 This is the system login interface of the second embodiment of the present invention;

[0051] Figure 5 This is the user registration interface of the second embodiment of the present invention;

[0052] Figure 6 This is the standard management and query interface of the second embodiment of the present invention;

[0053] Figure 7 This is the standard viewing function of the second embodiment of the present invention;

[0054] Figure 8 This is the standard knowledge graph interface of the second embodiment of the present invention;

[0055] Figure 9 This is the standard knowledge uploading and recognition function of the second embodiment of the present invention;

[0056] Figure 10 This is the standard knowledge extraction function of the second embodiment of the present invention;

[0057] Figure 11 This is the standard entity relationship query interface of the second embodiment of the present invention;

[0058] Figure 12 This is the standard knowledge question and answer interface of the second embodiment of the present invention. DETAILED DESCRIPTION

[0059] The following will be combined with the accompanying drawings of specific embodiments of the present invention to clearly and completely describe the exemplary embodiments of the present invention. Unless otherwise defined, technical or scientific terms used in the present invention should be given the common meanings understood by people with ordinary skills in the relevant field.

[0060] Example 1

[0061] like Figure 1 As shown, this embodiment discloses a tobacco agriculture standard knowledge graph application system, including a standard management and query module, a standard knowledge graph module, a standard knowledge graph construction module, a standard entity relationship query module and a standard knowledge question and answer module.

[0062] The system utilizes a client / server architecture. The front-end uses the React framework for interface interaction, Tailwind CSS for style management and responsive design, and the Lucide React unified icon system and React Hooks for component state management. The back-end is developed using the Python Flask framework, providing relevant interfaces and embedding standard feature extraction algorithms. A MySQL database is used to store user information and standard table data, MinIO for tobacco agricultural standard PDF documents, and a Neo4j graph database for standard structured information. Graph visualization is performed using cytoscape.js, and Cypher queries and result returns are constructed using py2neo.

[0063] (1) Standard management and query module

[0064] The standards management and query module implements the classification of standards, displays basic standard information, allows for viewing and downloading standard PDFs, and enables standard retrieval. Standards are categorized according to national, industry, local, and enterprise standards. The selected standard level displays the responsible department, standard number, standard name, status, and allows for viewing and downloading of relevant standards. The standard information table is stored in a MySQL database. Standard viewing and downloading operations are linked to the URL of the standard MinIO (which stores PDF documents of tobacco agricultural standards) to implement related functions. The URL is generated by MinIO, stored in MySQL, and points to the specific stored PDF file. Standard retrieval is primarily implemented using database LIKE queries.

[0065] (2) Standard knowledge graph module

[0066] The Standard Knowledge Graph module aims to visualize and dynamically update the Standard Knowledge Graph, helping users intuitively understand the content of standards and their interrelationships through a graphical interface. Specific functions include displaying the Standard Knowledge Graph and dynamically adding entity nodes and relationships.

[0067] Select the standard knowledge graph system through the relevant page to obtain the standard name, and obtain the relevant knowledge graph through the Neo4j graph database Cypher language "MATCH". Users can intuitively understand the standard content through the graphical interface.

[0068] The system now supports dynamic addition of nodes and relationships. Users can enter a node name and type through the interface, and the system automatically generates a Cypher statement to create the new node. Users can also select a node type, specify the source and target nodes, and the system automatically generates a Cypher statement to create the new relationship. This feature allows users to flexibly expand standard knowledge graphs and dynamically update and enrich standard data.

[0069] (3) Standard knowledge graph construction module

[0070] The standard knowledge graph construction module is embedded with various algorithms, including standard information element extraction algorithm, standard technical element extraction algorithm and standard clause element extraction algorithm, which assists users to quickly build a standard knowledge graph through algorithm process. Figure 2 shown.

[0071] The standard information element extraction algorithm is based on the large language model (LLM). For example, LLM (such as Kimi, GPT) is used to extract fixed fields (such as standard number, responsible department, release date, etc.) from the standard text. The standard technical element extraction algorithm is based on regular matching. For example, the regular rule of predefined technical parameters (such as total nitrogen ≤ (\d+) mg / m 3 ) to match key indicators from standard clauses. The standard clause element extraction algorithm uses a deep learning model to implement named entity recognition and relationship extraction. Named entity recognition models can use BERT_CRF, BERT_BiLSTM_CRF, or BiLSTM_Attention_CRF. Relationship extraction models can use BiLSTM, BiLSTM_Att, PCNN, or TextCNN_pos.

[0072] (4) Standard entity relationship query module

[0073] The Standard Entity Relationship Query module searches for and visualizes related triples based on user-entered head entities and relationship types. The query algorithm is written in Cypher, and its implementation logic is similar to the node and relationship addition functionality of the Standard Knowledge Graph module, using a "fill-in-the-blank" query approach.

[0074] Users enter entity names and relationship types through the front-end interface, and the system automatically generates a corresponding Cypher query statement. This Cypher query statement is sent to the Neo4j graph database via the py2neo library, and the Neo4j graph database returns triples that meet the conditions. The query results are then transmitted to the front-end for visualization, allowing users to intuitively understand the relationships between entities.

[0075] (5) Standard Question and Answer Module

[0076] The standard knowledge question answering module performs natural language processing, graph database query, knowledge sorting and other steps around the standard questions input by users, and finally provides standard-related intelligent question answering and graph display, such as Figure 3 As shown, the general steps include:

[0077] ①The user selects the question type and enters a natural language question;

[0078] ② The system uses LLM to identify the entities and intent in the question and generates the corresponding Cypher query statement;

[0079] ③ Retrieve relevant triples from the graph database, perform knowledge sorting (logical reorganization) on the triples retrieved from the graph database through LLM, and generate answers that conform to natural language habits;

[0080] ④Show the answer and related knowledge graph at the same time.

[0081] Because structured technologies are selected based on the characteristics of each piece of knowledge, entity recognition cannot rely on a single technology. Standard information technology elements have a simple structure, fixed entity types, and unified relationships between entities. LLM can efficiently identify the entity types and corresponding intent contained in questions. Standard article element entities have complex definitions, and pre-trained deep learning models can effectively identify entity information within complex semantics. Experimental verification has shown that LLM can effectively identify question intent based on the pre-defined question entities and question intent. Therefore, question intent recognition for standard article element types is completed by Prompt Engineering. In summary, the knowledge question answering module distinguishes questions, and different algorithms are embedded in different question types.

[0082] After information processing, the system queries the graph database using Cypher queries, accessing the Neo4j graph database to retrieve standard knowledge triples. To ensure the coherence of the acquired knowledge, the user's questions and acquired knowledge are fed into the LLM (Kimi), leveraging its powerful natural language processing capabilities to organize the knowledge based on the question and query answer.

[0083] Corresponding to the above method, this embodiment also discloses a visualization method of a tobacco agricultural standard knowledge graph application system, comprising:

[0084] Information elements, technical elements and clause elements are extracted from the tobacco agricultural standard text to form structured triples.

[0085] Import triple data into the Neo4j graph database to build a knowledge graph.

[0086] The knowledge graph is dynamically displayed through visualization tools, supporting the dynamic addition of nodes and relationships.

[0087] Retrieve and return relevant knowledge from the knowledge graph based on user queries or natural language questions.

[0088] Example 2

[0089] System user information is stored in the MySQL database to ensure efficient management and secure storage of user data. The user data table design includes basic account information, such as email address, mobile phone number, password (encrypted storage), registration time, last login time and other fields to ensure that the identity can be effectively verified and the user's activity history can be tracked during the user registration and login process. The system user login interface is as follows: Figure 4 As shown, users can choose to log in via email or mobile phone number. In the email login section, users enter their registered email address, password, and verification code for verification; in the mobile phone number login section, users enter their mobile phone number and verification code for login.

[0090] The system provides a registration function. Users need to fill in their email address and mobile phone number, set a password and confirm the password, and enter a verification code to ensure the security of the registration operation. After completing these operations, the user can click the Register and Login button, and the system will verify the registration information and successfully create the user account. Figure 5 shown.

[0091] (1) Standard management and query module

[0092] The standard management and query module provides two main functions: standard classification and standard information display, helping users to effectively manage and query tobacco agricultural standards, such as Figure 6 The system's search function allows users to quickly search for specific standards by entering a standard number or name, improving standard retrieval efficiency. Furthermore, the combination of standard classification and the query interface enables users to quickly find the required standard information and perform operations, ensuring ease of use when browsing and downloading standard documents.

[0093] By clicking on the three operations under the operation bar, you can download, view and structure the corresponding standards. For example, click on the "Standard View" corresponding to "Guidelines for the Rational Use of Pesticides (I)" to view its PDF document and obtain the standard content. Figure 7 The standard download function can download the PDF locally, and the standard graph function can be clicked to switch to the standard knowledge graph module.

[0094] (2) Standard knowledge graph module

[0095] The standard knowledge graph module displays the corresponding standard structured graph, such as Figure 8 By clicking Figure 6 The standard map under "Enterprise Standards - Henan Province Flue-cured Tobacco Quality Standards" allows you to view its structured knowledge. To adapt to dynamic changes in standards, add standard entities by entering the node name and node type on the right side of the interface. Add relationships between entities by selecting the source and target nodes, entering the relationship type, and clicking Add Relationship.

[0096] (3) Standard knowledge graph construction module

[0097] The standard knowledge graph construction module uses experimental algorithms to assist users in completing the construction of the tobacco agriculture standard knowledge graph. The task bar on the left side of this module uses "unfinished" and "completed" to prompt users to complete the task status; "in progress" indicates the current task; and "to be completed" indicates the next task to be completed by the user. Figure 9 As shown, after uploading the "Technical Regulations for Mature Harvesting of Flue-cured Tobacco" standard, the identified content is displayed in real time.

[0098] Standard knowledge extraction and storage To ensure the correctness of the extracted triples, the extracted data viewing and downloading functions are provided. If the user needs to adjust the extracted data, the data can be downloaded, manually reviewed and corrected, and then uploaded and confirmed. Figure 10 shown.

[0099] (4) Standard entity relationship query module

[0100] The main function of the standard entity relationship query module is to provide graph query, helping users to quickly obtain knowledge triples based on the input head entity and relationship type. This module implements the query function through three input boxes, where users only need to enter one of the fields to trigger the relevant query operation. Figure 11 As shown: Users can select or enter a standard header entity (such as "Technical Regulations for Tobacco Wet Seedling Cultivation") and select the corresponding relationship type (such as "Reference Standard"). The system will then generate a Cypher query statement based on the input information and extract qualified knowledge triples from the graph database.

[0101] (5) Standard knowledge question and answer module

[0102] The main function of the standard knowledge question and answer module is to provide intelligent question and answer services for standard-related questions and to display the relevant knowledge graph in a visual way. In this module, users first select the question type (such as standard information, clause elements, etc.), and then fill in the specific question in the input box. After clicking the query, the system will use the preset algorithm and model, combined with the knowledge graph in the graph database, to generate the corresponding answer and display the answer source graph, such as Figure 12 shown.

[0103] The present invention stores the acquired triple data in a Neo4j database and visualizes the data. Based on the constructed knowledge graph, trained named entity model, designed regular template, and LLM model algorithm, the tobacco agricultural standards knowledge graph application system architecture is designed to fully utilize the constructed knowledge graph and assist users in efficiently querying standard knowledge.

[0104] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the above-described embodiments, those skilled in the art should understand that any person skilled in the art can modify the technical solutions described in the above-described embodiments within the technical scope disclosed by the present invention, or replace some of the technical features therein with equivalents. Such modifications or replacements do not deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be subject to the scope of protection of the claims.

Claims

1. A tobacco agricultural standard knowledge graph application system, characterized in that: Includes the following modules: The standards management and query module is used to classify, store and manage tobacco agricultural standards according to national standards, industry standards, local standards and enterprise standards, and supports viewing and downloading standard PDF documents; Standard knowledge graph module, used to dynamically display entity nodes and their relationships through visualization tools, and supports dynamic addition of nodes and relationships; A standard knowledge graph construction module is used to automatically extract "entity-relationship-entity" triples from standard texts, store the structured triple data of tobacco agricultural standards in a Neo4j graph database, and generate a knowledge graph; Standard entity-relationship query module, which generates Cypher query statements based on the head entity and relationship type input by the user, and retrieves and visualizes related triples from the graph database; The standard knowledge question-and-answer module is used to receive natural language questions input by users, generate professional answers by combining the large language model and graph database query results, and display relevant knowledge graphs.

2. The tobacco agricultural standard knowledge graph application system according to claim 1, characterized in that: The standard management and query module also includes: The standard information table is stored in the MySQL database and contains the standard number, standard name, responsible department, status, and the URL of the PDF document stored in MinIO; The standard search function is implemented based on database LIKE query, allowing users to quickly locate target standards by standard number or standard name.

3. The tobacco agricultural standard knowledge graph application system according to claim 1, characterized in that: The standard knowledge graph module supports dynamic addition of nodes and relationships in the following ways: The user enters the node name and type through the interface, and the system automatically generates a Cypher statement to create a new node; The user selects the source node, target node, and relationship type, and the system automatically generates a Cypher statement to create a new relationship; Neo4j's APOC extension library is used to support the creation and modification of dynamic nodes and relationships.

4. The tobacco agricultural standard knowledge graph application system according to claim 1, characterized in that: The standard knowledge graph construction module embeds a standard information element extraction algorithm, a standard technical element extraction algorithm and a standard article element extraction algorithm. The standard information element extraction algorithm is implemented based on a large language model, the standard technical element extraction algorithm is implemented based on regular matching, and the standard article element extraction algorithm implements named entity recognition and relationship extraction based on a deep learning model.

5. The tobacco agricultural standard knowledge graph application system according to claim 1, characterized in that: The standard entity relationship query module implements the query function through the following steps: The user enters the header entity and relationship type, and the system converts it into a Cypher query statement; The Cypher query statement is sent to the Neo4j graph database through the py2neo library for execution, and the Neo4j graph database returns triple data; Use cytoscape.js to display the query results in the form of a visual map.

6. The tobacco agricultural standard knowledge graph application system according to claim 1, characterized in that: The workflow of the standard knowledge question answering module includes: The user selects the question type and enters a natural language question; The system uses LLM to identify entities and intents in the question and generates corresponding Cypher query statements; Retrieve relevant triples from the graph database, then use LLM to sort out the knowledge of the triples retrieved from the graph database and generate structured answers; The answer and related knowledge graph are displayed at the same time.

7. The tobacco agricultural standards knowledge graph application system according to claim 6, characterized in that: The standard knowledge question answering module can be used to identify entities and intents in questions, which can be implemented in two technical ways: For standard information and standard technical elements, the entities and intentions in user questions are directly identified through LLM; For standard clause elements, deep learning models are used to identify entities in the clauses, and PromptEngineering is used to guide LLM to parse the intent.

8. A visualization method for the tobacco agricultural standards knowledge graph application system according to any one of claims 1 to 7, comprising: Extract information elements, technical elements and clause elements from the tobacco agricultural standard text to form structured triples; Import triple data into Neo4j graph database to build knowledge graph; Dynamically display the knowledge graph through visualization tools, and support dynamic addition of nodes and relationships; Retrieve and return relevant knowledge from the knowledge graph based on user queries or natural language questions.

9. A computer-readable storage medium, characterized in that A computer program is stored, and when the computer program is executed by a processor, the method according to claim 8 is implemented.

Images