Language large model-based literature knowledge graph classification system

Abstract

The invention discloses a literature knowledge graph classification system based on a language large model, and relates to the technical field. The method comprises a literature interpretation model and a literature classification model. The literature interpretation model interprets the sample literature to obtain constituent elements of the literature, and knowledge graph elements capable of forming knowledge graph classification are screened from the constituent elements; the literature classification model searches from the literature library according to the constituent elements, all the related literatures are input into the literature interpretation model for interpretation to obtain the constituent elements of each related literature, the constituent elements of the sample literatures serve as classification framework elements, and a literature knowledge graph classification framework diagram of the sample literatures and the related literatures is formed. According to the method, the literature composition elements are interpreted through the literature interpretation model, the literature knowledge graph classification framework diagram of the sample literature and the related literature is formed through the literature classification model, under a large model, the literature knowledge graph is more complete, better in hierarchy and good in classification effect, and the structure layer of the literature knowledge graph can be known more easily.

Description

Technical Field

[0001] This invention belongs to the field of document knowledge graph classification technology, and in particular relates to a document knowledge graph classification system based on a large language model. Background Technology

[0002] A document knowledge graph is a knowledge graph constructed based on the structure of documents. It presents a knowledge system in the form of a structured graph by mining entities, relationships, semantic connections, research directions, and research methods within the documents. It integrates core elements (such as authors, institutions, keywords, research topics, etc.) and research methods from the documents, and reveals hidden knowledge connections through graph networks. It is widely used in academic research, intelligence analysis, knowledge management, and other fields. Existing document classification mainly follows the categories of fields and topics, but there are few methods that apply knowledge graphs for classification, and even fewer that classify the entire document database under a large model. This results in incomplete classifications and low-quality classification graphs. Furthermore, it is difficult to effectively classify documents by related fields, leading to even fewer methods for classifying related domain documents using knowledge graphs.

[0003] To address these issues, we propose a document knowledge graph classification system based on a large language model. Summary of the Invention

[0004] The purpose of this invention is to provide a document knowledge graph classification system based on a large language model. It interprets the constituent elements of documents through a document interpretation model and forms a document classification model, thus solving the problems of low classification quality of existing document knowledge graphs and failure to utilize large models and related fields to form a more comprehensive document knowledge graph.

[0005] To solve the above-mentioned technical problems, the present invention is achieved through the following technical solution:

[0006] This invention is a document knowledge graph classification system based on a large language model, including a document interpretation model and a document classification model;

[0007] The document interpretation model interprets the sample documents, interprets the research fields and research directions of the documents, obtains the constituent elements of the documents, filters the constituent elements, and selects knowledge graph elements that can constitute knowledge graph classification from the constituent elements.

[0008] The document classification model searches the document database based on the constituent elements of the sample document, searches for and downloads files related to the sample document, and inputs all related documents into the document interpretation model for interpretation to obtain the constituent elements of each related document. The constituent elements of the sample document are used as classification framework elements. The document classification model establishes a knowledge graph framework based on the classification framework elements, and constructs the constituent elements of each related document into the knowledge graph framework, forming a document knowledge graph classification framework diagram of the sample document and related documents.

[0009] The present invention is further configured such that the constituent elements of the sample documents include the paper field, paper topic, research direction, research method, research results and author information. The paper topics of related papers are listed, all papers with the same research direction and research method are listed, the similarities and differences are identified, a related knowledge graph is formed and classified, the differences and similarities of the research results of the same research direction and research method are classified, and the differences of the research results of different research directions and research methods are further classified.

[0010] The present invention is further configured such that the literature classification model classifies the topics of papers within the same paper field, divides them into multiple topic categories according to major categories, and then divides each topic category into multiple research directions that match that topic, and the research methods of each research direction must meet the requirements of its corresponding research topic.

[0011] The present invention is further configured such that the research methods under the research direction of the sample literature and the research methods of related literature will be classified on the literature knowledge graph classification framework diagram. Each research method will be marked with the literature from which it comes, and clicking on the corresponding literature can view the specific content of the research method in the literature.

[0012] The present invention is further configured such that the author information includes the sample document author, document publication information, publishing institution, and publishing journal.

[0013] The present invention is further configured such that, in establishing the literature graph classification database, the literature classification model classifies the sample literature to form a literature knowledge graph classification framework diagram of the sample literature and related literature, and then saves it to the literature graph classification database;

[0014] When interpreting and classifying new sample documents using document interpretation and classification models, a document knowledge graph classification framework diagram of the new sample documents and related documents is obtained. Then, the document knowledge graph classification framework diagrams of sample documents with similar themes to the new sample documents are searched from the document graph classification database. Knowledge graphs that are different from the document knowledge graph classification framework diagrams of the new sample documents are extracted from the document knowledge graph classification framework diagrams of the sample documents searched from the document graph classification database, and the extracted knowledge graphs are constructed on the document knowledge graph classification framework diagram of the new sample documents.

[0015] The present invention is further configured such that the document classification model includes a cross-domain search module. When the cross-domain search module is activated, documents that are associated with the sample documents and span multiple domains are searched. These documents that span multiple domains are then classified into a knowledge graph using the document classification model to obtain a cross-domain document knowledge graph classification. The cross-domain document knowledge graph classification is also structured onto the document knowledge graph classification framework diagram.

[0016] The present invention is further configured such that when searching for documents that are associated with sample documents and span multiple fields, the specific associated cross-field can be selected.

[0017] The present invention is further configured such that the document classification model includes a time model, which timestamps the relevant documents associated with the sample document to form a timeline of all documents on the entire document knowledge graph classification framework, thereby forming a knowledge graph of the development history of the relevant documents associated with the sample document.

[0018] The present invention has the following beneficial effects:

[0019] 1. This invention uses a document interpretation model to interpret the constituent elements of sample documents. The document classification model searches for relevant documents from the document database based on the constituent elements, then interprets the relevant documents to determine their constituent elements. Then, the constituent elements in each relevant document are structured into a knowledge graph framework to form a document knowledge graph classification framework diagram of sample documents and relevant documents. Under the large model, the document knowledge graph is more complete, has better hierarchy, better classification effect, and is more conducive to understanding the document knowledge graph structure layer.

[0020] 2. This invention also searches for cross-domain literature related to the sample literature, interprets it, and forms a cross-domain literature knowledge graph classification to supplement and solve the problem that it is difficult to construct and classify literature knowledge graphs across multiple domains, and therefore cannot be effectively constructed.

[0021] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0022] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a schematic diagram of a document knowledge graph classification system based on a large language model.

[0024] Figure 2 This is a schematic diagram of the constituent elements of the sample documents.

[0025] Figure 3 This is a schematic diagram illustrating the constituent elements of sample documents and related documents. Detailed Implementation

[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] Please see Figure 1-3 This invention is a document knowledge graph classification system based on a large language model, including a document interpretation model and a document classification model;

[0028] The document interpretation model interprets the sample documents, interprets the research fields and research directions of the documents, obtains the constituent elements of the documents, filters the constituent elements, and selects knowledge graph elements that can constitute knowledge graph classification from the constituent elements.

[0029] The document classification model searches the document database based on the constituent elements of the sample document, searches for and downloads files related to the sample document, and inputs all related documents into the document interpretation model for interpretation to obtain the constituent elements of each related document. The constituent elements of the sample document are used as classification framework elements. The document classification model establishes a knowledge graph framework based on the classification framework elements, and constructs the constituent elements of each related document into the knowledge graph framework, forming a document knowledge graph classification framework diagram of the sample document and related documents.

[0030] The construction and classification of literature knowledge graphs should be done by domain. A sample document has few constituent elements, resulting in a simple knowledge graph. These constituent elements are extracted, and a literature classification model is used to search for related documents in the paper database. These related documents are then interpreted to form their constituent elements, which are then used in conjunction with the sample document's constituent elements to construct a more comprehensive and richer literature knowledge graph classification framework. Within a domain, multiple themes (using sample documents to construct the literature knowledge graph) can be explored. This allows for the integration of literature knowledge graphs from multiple themes within a domain to form a comprehensive and richer knowledge graph classification for that domain.

[0031] The constituent elements of the sample documents include the paper field, paper topic, research direction, research method, research results, and author information. The paper topics of related papers are listed, and all papers with the same research direction and research method are listed. The similarities and differences are identified, and a related knowledge graph is formed and classified. The differences and similarities of the research results with the same research direction and research method are classified, and the differences of the research results with different research directions and research methods are further classified.

[0032] There are various ways to classify the constituent elements of literature; this study primarily uses research direction and methodology to build a knowledge graph. This effectively categorizes sample literature and related literature's research directions and methods, presenting them in a reasonable and effective manner as a knowledge graph. This allows for more effective interpretation of the knowledge graph content.

[0033] The literature classification model categorizes the topics of papers within the same research field, dividing them into multiple topic categories according to major categories. Each topic category is then further divided into multiple research directions that match that topic. The research methods of each research direction must conform to the requirements of its corresponding research topic.

[0034] Expanding the construction of literature knowledge graphs across domains allows for the creation of knowledge graph classifications for most of the paper topics within a domain. This enables the classification of knowledge graphs for most topics within a domain, which can then be merged to form a domain-specific knowledge graph classification.

[0035] The research methods under the research directions of the sample literature and the research methods of related literature will be classified on the literature knowledge graph classification framework. Each research method will be marked with the literature from which it comes. Clicking on the corresponding literature will allow you to view the specific content of the research method in the literature.

[0036] For knowledge graph classification, only the classification results can be seen, and the research-related content and results cannot be understood. Hyperlinks are set up (the relevant research methods are extracted after downloading the relevant literature), and you can view the content by clicking on them.

[0037] The author information includes the authors of the sample documents, publication information, publishing institution, and journal. This author information facilitates understanding the source and publication date of important papers in the knowledge graph classification, enabling the creation of a document knowledge graph to identify the sources of documents and determine which institutions publish more valuable papers.

[0038] The document classification database is established by classifying sample documents using the document classification model to form a document knowledge graph classification framework diagram of sample documents and related documents, which is then saved to the document classification database.

[0039] When interpreting and classifying new sample documents using document interpretation and classification models, a document knowledge graph classification framework diagram of the new sample documents and related documents is obtained. Then, the document knowledge graph classification framework diagrams of sample documents with similar themes to the new sample documents are searched from the document graph classification database. Knowledge graphs that are different from the document knowledge graph classification framework diagrams of the new sample documents are extracted from the document knowledge graph classification framework diagrams of the sample documents searched from the document graph classification database, and the extracted knowledge graphs are constructed on the document knowledge graph classification framework diagram of the new sample documents.

[0040] Establishing a database allows for the storage of existing literature knowledge graphs. When creating knowledge graphs for similar literature in the future, the database can be directly accessed for comparison and analysis, or used as a reference to discover new updates to the knowledge graph on related topics. This facilitates understanding literature development and allows for comparison of research methods related to the same research topic, revealing the evolution and improvement of research methods.

[0041] The document classification model includes a cross-domain search module. When the cross-domain search module is activated, documents that are associated with the sample documents and span multiple domains are searched. These documents that span multiple domains are then classified into a knowledge graph using the document classification model to obtain a cross-domain document knowledge graph classification. This cross-domain document knowledge graph classification is also structured onto the document knowledge graph classification framework diagram.

[0042] Interdisciplinary research is increasing and becoming more widespread, but related knowledge graph research is scarce, failing to provide direction for such research. Therefore, establishing interdisciplinary knowledge graphs can assist in conducting or researching interdisciplinary fields.

[0043] When searching for literature that is related to sample documents and spans multiple fields, users can select specific cross-field relationships. Users can choose cross-field relationships to more accurately find topics in several related fields to build knowledge graphs, and can select specific related fields for knowledge graph classification.

[0044] The document classification model includes a time model, which timestamps related documents associated with the sample documents to form a timeline of all documents on the entire document knowledge graph classification framework, thus forming a knowledge graph of the development history of related documents associated with the sample documents.

[0045] You can clearly see the development timeline of knowledge graphs in relevant literature, and accurately understand the development timeline of knowledge graphs.

[0046] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0047] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A document knowledge graph classification system based on a large language model, characterized in that: This includes document interpretation models and document classification models; The document interpretation model interprets the sample documents, interprets the research fields and research directions of the documents, obtains the constituent elements of the documents, filters the constituent elements, and selects knowledge graph elements that can constitute knowledge graph classification from the constituent elements. The document classification model searches the document database based on the constituent elements of the sample document, searches for and downloads files related to the sample document, and inputs all related documents into the document interpretation model for interpretation to obtain the constituent elements of each related document. The constituent elements of the sample document are used as classification framework elements. The document classification model establishes a knowledge graph framework based on the classification framework elements, and constructs the constituent elements of each related document into the knowledge graph framework, forming a document knowledge graph classification framework diagram of the sample document and related documents.

2. The document knowledge graph classification system based on a large language model according to claim 1, characterized in that, The constituent elements of the sample documents include the paper field, paper topic, research direction, research method, research results, and author information. The paper topics of related papers are listed, and all papers with the same research direction and research method are listed. The similarities and differences are identified, and a related knowledge graph is formed and classified. The differences and similarities of the research results with the same research direction and research method are classified, and the differences of the research results with different research directions and research methods are further classified.

3. The document knowledge graph classification system based on a large language model according to claim 2, characterized in that, The literature classification model categorizes the topics of papers within the same research field, dividing them into multiple topic categories according to major categories. Each topic category is then further divided into multiple research directions that match that topic. The research methods of each research direction must conform to the requirements of its corresponding research topic.

4. A document knowledge graph classification system based on a large language model according to claim 2, characterized in that, The research methods under the research directions of the sample literature and the research methods of related literature will be classified on the literature knowledge graph classification framework. Each research method will be marked with the literature from which it comes. Clicking on the corresponding literature will allow you to view the specific content of the research method in the literature.

5. A document knowledge graph classification system based on a large language model according to claim 2, characterized in that, The author information includes the authors of the sample documents, publication information, publishing institution, and journal.

6. The document knowledge graph classification system based on a large language model according to claim 1, characterized in that, The document classification database is established by classifying sample documents using the document classification model to form a document knowledge graph classification framework diagram of sample documents and related documents, which is then saved to the document classification database. When interpreting and classifying new sample documents using document interpretation and classification models, a document knowledge graph classification framework diagram of the new sample documents and related documents is obtained. Then, the document knowledge graph classification framework diagrams of sample documents with similar themes to the new sample documents are searched from the document graph classification database. Knowledge graphs that are different from the document knowledge graph classification framework diagrams of the new sample documents are extracted from the document knowledge graph classification framework diagrams of the sample documents searched from the document graph classification database, and the extracted knowledge graphs are constructed on the document knowledge graph classification framework diagram of the new sample documents.

7. A document knowledge graph classification system based on a large language model according to claim 1, characterized in that, The document classification model includes a cross-domain search module. When the cross-domain search module is activated, documents that are associated with the sample documents and span multiple domains are searched. These documents that span multiple domains are then classified into a knowledge graph using the document classification model to obtain a cross-domain document knowledge graph classification. This cross-domain document knowledge graph classification is also structured onto the document knowledge graph classification framework diagram.

8. A document knowledge graph classification system based on a large language model according to claim 7, characterized in that, When searching for literature that is related to the sample documents and spans multiple fields, you can select the specific cross-field association.

9. A document knowledge graph classification system based on a large language model according to claim 1, characterized in that, The document classification model includes a time model, which timestamps related documents associated with the sample documents to form a timeline of all documents on the entire document knowledge graph classification framework, thus forming a knowledge graph of the development history of related documents associated with the sample documents.

Images