Interactive discussion platform construction method based on knowledge graph
By using a knowledge graph-based interactive seminar platform and leveraging natural language understanding and deep learning technologies, the problems of single entity relationship processing and lack of interactivity in traditional association technologies have been solved. This has enabled intelligent and efficient communication in the seminar process and generated detailed research reports.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-26
- Publication Date
- 2026-04-10
AI Technical Summary
Traditional association technologies cannot effectively handle the complex and diverse relationships between entities, lack interactivity and critical thinking, cannot simulate human thought processes, and have a limited range of display methods, failing to assist in understanding and exploring knowledge.
An interactive discussion platform based on knowledge graphs is adopted. By constructing a research task framework, natural language understanding, semantic segmentation and vectorization processing are used, combined with knowledge graph representation learning and deep learning, to enable researchers to conduct correlation analysis and generate conclusions, and to form a research report through multiple rounds of approval.
It has achieved standardization and intelligentization of the discussion process, can simulate the human thinking process, quickly acquire domain knowledge and relationships, generate effective conclusions, and support efficient discussion and communication.
Smart Images
Figure CN121835836A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of discussion and relates to a method for constructing an interactive discussion platform based on knowledge graphs. Background Technology
[0002] Traditional association techniques typically rely on tables and fields for data retrieval, resulting in a limited range of representations. They are ill-suited for handling the complex and diverse relationships between entities. Furthermore, traditional association techniques cannot support interactive, exploratory analysis based on knowledge graphs, nor can they simulate human thought processes. They lack the interactivity and analytical depth required for step-by-step verification and reasoning to arrive at conclusions.
[0003] Traditional association techniques, such as mind maps and topology maps, cannot be visualized and therefore cannot help people better understand and explore knowledge. Summary of the Invention
[0004] The technical problem solved by this invention is to overcome the shortcomings of the prior art and propose a method for constructing an interactive discussion platform based on knowledge graphs, so as to realize the standardization and intelligence of the discussion process, support the realization of efficient discussion and communication, and output effective conclusions.
[0005] The solution of the present invention is:
[0006] A method for constructing an interactive seminar platform based on knowledge graphs, comprising:
[0007] The task initiator constructs the research task framework and distributes the research task framework and research tasks in different fields to researchers in the corresponding fields.
[0008] Researchers conduct correlation analysis on the received research tasks; obtain answers, generate conclusions, and report them to the task initiator;
[0009] The task initiator reviews each conclusion once and sends the review results to the summary personnel. After the summary personnel summarize the results, they return them to the task initiator for a second review. After the second review, the research report is output.
[0010] In the aforementioned method for constructing an interactive seminar platform based on knowledge graphs, the method for constructing the research task framework is as follows:
[0011] Based on pre-set and editable research task templates, new research tasks are added to form a research task framework.
[0012] In the aforementioned method for constructing an interactive seminar platform based on knowledge graphs, the distribution of research tasks supports single-task-to-single-researcher, single-task-to-multiple-researchers, and multiple-task-to-single-researcher modes; after the research tasks are distributed, the seminar task status can be viewed and modified.
[0013] In the aforementioned method for constructing an interactive discussion platform based on knowledge graphs, the method researchers use to conduct correlation analysis on the received research tasks is as follows:
[0014] Researchers input questions into the research task framework; the research task framework uses a natural language understanding component to identify the intent of the question and extract relevant keywords and entities; and uses semantic word segmentation technology to extract feature keywords, entities, and categories;
[0015] The input question text is vectorized to obtain a corresponding set of generalized statements; the set of generalized statements is then matched with pre-defined templates for various research tasks.
[0016] When a match is successful, the vector-based knowledge graph representation learning component enters the graph knowledge query; it executes the query based on the optimal generalization statement mentioned by the variable and returns the correct answer.
[0017] When a match fails, the process proceeds to full-text semantic retrieval. In this process, a semantic segmentation component based on deep learning is used to calculate the similarity of semantic vectors between the question and the document content. The set of answers with the highest similarity is obtained. The optimal answer is selected from the set of answers and returned to the researchers.
[0018] Based on the obtained answers, the researchers generated conclusions and reported them to the task initiator.
[0019] In the aforementioned method for constructing an interactive seminar platform based on knowledge graphs, the natural language understanding component is:
[0020] It employs some basic text processing techniques, including keyword extraction and word frequency statistics, to quickly capture the main information of the text and support the rapid extraction and understanding of key information.
[0021] In the aforementioned method for constructing an interactive seminar platform based on knowledge graphs, the deep learning semantic word segmentation component is:
[0022] The characters in the corpus are embedded, and the character embedding features are input into a bidirectional LSTM. The output layer outputs the features learned by deep learning and inputs them into a CRF layer to obtain the final model.
[0023] In the aforementioned method for constructing an interactive seminar platform based on knowledge graphs, the knowledge graph representation learning component for the vectors is:
[0024] Based on the modeling principle of knowledge graph embedding representation model, vector-based knowledge representation models are divided into different types of models, including the vector-based triangle rule and the TransE model based on the norm principle.
[0025] In the aforementioned method for constructing an interactive discussion platform based on knowledge graphs, the conclusions of the research task include viewpoints and supporting materials.
[0026] In the aforementioned method for constructing an interactive discussion platform based on knowledge graphs, the task initiator conducts an initial review of the viewpoints and supporting materials for each research task; the initial review includes approval or rejection; when approved, it is sent to the summary personnel; when rejected, it is returned to the researchers to continue the correlation analysis.
[0027] In the aforementioned method for constructing an interactive discussion platform based on knowledge graphs, the secondary approval process includes passing or rejecting the application; when passed, a research report is generated; when rejected, the application is returned to the summary personnel for further correlation analysis; the research report includes data and charts.
[0028] The beneficial effects of this invention compared to the prior art are:
[0029] (1) The knowledge graph of this invention uses a graph model to express knowledge, which can handle the complex and diverse relationships between entities well. Moreover, the knowledge graph based on graph theory and probabilistic graph model can better express and describe various relationships in the real world.
[0030] (2) This invention employs an interactive exploratory analysis based on knowledge graphs, which can simulate the human thought process and draw conclusions through step-by-step verification and reasoning. This analytical approach is closer to human thinking and can help business personnel better understand and solve business problems;
[0031] (3) The present invention builds an interactive discussion platform based on knowledge graphs, which enables researchers to quickly acquire domain-related knowledge and the relationships between knowledge, and to successfully complete the discussion tasks initiated by the task initiator. Attached Figure Description
[0032] Figure 1 This is a flowchart illustrating the construction process of the knowledge graph-based interactive discussion platform of this invention.
[0033] Figure 2 This is a schematic diagram of the TransE model of the present invention. Detailed Implementation
[0034] The present invention will be further described below with reference to the embodiments.
[0035] This invention proposes a method for constructing an interactive discussion platform based on knowledge graphs. Based on knowledge graph technology, a single-round / multi-round interactive discussion process is designed for the discussion field. Researchers can use association analysis tools, through intelligent analysis, data retrieval, document search, etc., to form conclusions and supporting evidence through human-computer interaction and push them to the task initiator's terminal. The task initiator then initiates the discussion and forms relevant conclusions.
[0036] Methods for constructing interactive discussion platforms based on knowledge graphs, such as Figure 1 As shown, it specifically includes the following:
[0037] The task initiator constructs the research task framework and distributes the research task framework and research tasks in different fields to researchers in the corresponding fields.
[0038] The method for constructing a research task framework is as follows:
[0039] Based on pre-set and editable research task templates, new research tasks are added to form a research task framework.
[0040] The distribution of research tasks supports single-task-to-single-researcher, single-task-to-multiple-researchers, and multiple-task-to-single-researcher modes; after the research task is distributed, the status of the task can be viewed and modified.
[0041] Researchers conduct correlation analysis on the research tasks they receive; obtain answers, generate conclusions, and report them to the task initiator.
[0042] The researchers used the following method to conduct correlation analysis on the received research tasks:
[0043] Researchers input questions into the research task framework; the research task framework uses a natural language understanding component to identify the intent of the question and extract relevant keywords and entities; and uses semantic word segmentation technology to extract feature keywords, entities, and categories.
[0044] The natural language understanding component is as follows:
[0045] It employs some basic text processing techniques, including keyword extraction and word frequency statistics, to quickly capture the main information of the text and support the rapid extraction and understanding of key information.
[0046] The input question text is vectorized to obtain a set of generalized statements; the set of generalized statements is then matched with pre-set templates for various research tasks.
[0047] When a match is successful, the vector-based knowledge graph representation learning component enters the graph knowledge query; it executes the query based on the optimal generalization statement mentioned by the variable and returns the correct answer.
[0048] The knowledge graph representation learning component for vectors is:
[0049] Based on the modeling principles of knowledge graph embedding representation models, vector-based knowledge representation models are divided into different types of models. Representation methods include the TransE model based on the triangle rule and norm principle of vectors, such as... Figure 2 As shown.
[0050] When a match fails, the process proceeds to full-text semantic retrieval. In this process, a semantic segmentation component based on deep learning is used to calculate the similarity of semantic vectors between the question and the document content. The set of answers with the highest similarity is obtained, and the optimal answer is selected from the set and returned to the researchers.
[0051] The semantic word segmentation component of deep learning is:
[0052] The characters in the corpus are embedded, and the character embedding features are input into a bidirectional LSTM. The output layer outputs the features learned by deep learning and inputs them into a CRF layer to obtain the final model.
[0053] Based on the obtained answers, the researchers generated conclusions and reported them to the task initiator.
[0054] The conclusions of a research task include the viewpoints and supporting materials.
[0055] The task initiator reviews each conclusion once and sends the review results to the summary personnel. After the summary personnel summarize the results, they return them to the task initiator for a second review. After the second review, the research report is output.
[0056] The task initiator reviews the viewpoints and supporting materials for each research task once; the review includes approval or rejection; when approved, it is sent to the summary personnel; when rejected, it is returned to the researchers to continue the correlation analysis.
[0057] The secondary approval process includes approval or rejection; when approved, a research report is generated; when rejected, the report is returned to the summary personnel for further correlation analysis; the research report includes data and charts.
[0058] The knowledge graph of this invention uses a graph model to represent knowledge, which can handle the complex and diverse relationships between entities. Moreover, the knowledge graph based on graph theory and probabilistic graphical models can better express and describe various relationships in the real world.
[0059] This invention employs an interactive, exploratory analysis based on knowledge graphs, which can simulate the human thought process, drawing conclusions through step-by-step verification and reasoning. This analytical approach is closer to human thinking and can help business personnel better understand and solve business problems.
[0060] This invention builds an interactive discussion platform based on knowledge graphs, enabling researchers to quickly acquire domain-related knowledge and the relationships between knowledge points, and to successfully complete discussion tasks initiated by task initiators.
[0061] Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make possible changes and modifications to the technical solutions of the present invention by utilizing the methods and techniques disclosed above without departing from the spirit and scope of the present invention. Therefore, any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solutions of the present invention shall fall within the protection scope of the technical solutions of the present invention.
Claims
1. A method for constructing an interactive discussion platform based on knowledge graphs, characterized in that: include: The task initiator constructs the research task framework; The research task framework and research tasks in different fields will be distributed to researchers in the corresponding fields. Researchers conduct correlation analysis on the received research tasks; obtain answers, generate conclusions, and report them to the task initiator; The task initiator reviews each conclusion once and sends the review results to the summary personnel. After the summary personnel summarize the results, they return them to the task initiator for a second review. After the second review, the research report is output.
2. The method for constructing an interactive discussion platform based on knowledge graphs according to claim 1, characterized in that: The method for constructing a research task framework is as follows: Based on pre-set and editable research task templates, new research tasks are added to form a research task framework.
3. The method for constructing an interactive discussion platform based on knowledge graphs according to claim 1, characterized in that: The distribution of research tasks supports single-task-to-single-researcher, single-task-to-multiple-researchers, and multiple-task-to-single-researcher modes; after the research task is distributed, the status of the task can be viewed and modified.
4. The method for constructing an interactive discussion platform based on knowledge graphs according to claim 3, characterized in that: The researchers used the following method to conduct correlation analysis on the received research tasks: Researchers input questions into the research task framework; the research task framework uses a natural language understanding component to identify the intent of the question and extract relevant keywords and entities; and uses semantic word segmentation technology to extract feature keywords, entities, and categories; The input question text is vectorized to obtain a corresponding set of generalized statements; the set of generalized statements is then matched with pre-defined templates for various research tasks. When a match is successful, the vector-based knowledge graph representation learning component enters the graph knowledge query; it executes the query based on the optimal generalization statement mentioned by the variable and returns the correct answer. When a match fails, the process proceeds to full-text semantic retrieval. In this process, a semantic segmentation component based on deep learning is used to calculate the similarity of semantic vectors between the question and the document content. The set of answers with the highest similarity is obtained. The optimal answer is selected from the set of answers and returned to the researchers. Based on the obtained answers, the researchers generated conclusions and reported them to the task initiator.
5. The method for constructing an interactive discussion platform based on knowledge graphs according to claim 4, characterized in that: The natural language understanding component is: It employs some basic text processing techniques, including keyword extraction and word frequency statistics, to quickly capture the main information of the text and support the rapid extraction and understanding of key information.
6. The method for constructing an interactive discussion platform based on knowledge graphs according to claim 4, characterized in that: The semantic word segmentation component of the deep learning is: The characters in the corpus are embedded, and the character embedding features are input into a bidirectional LSTM. The output layer outputs the features learned by deep learning and inputs them into a CRF layer to obtain the final model.
7. The method for constructing an interactive discussion platform based on knowledge graphs according to claim 4, characterized in that: The knowledge graph representation learning component for the vector is: Based on the modeling principle of knowledge graph embedding representation model, vector-based knowledge representation models are divided into different types of models, including the TransE model based on the triangle rule and norm principle of vectors.
8. The method for constructing an interactive discussion platform based on knowledge graphs according to claim 4, characterized in that: The conclusions of the research task include viewpoints and supporting materials.
9. The method for constructing an interactive discussion platform based on knowledge graphs according to claim 8, characterized in that: The task initiator conducts a one-time review of the viewpoints and supporting materials for each research task; the one-time review includes approval or rejection; when approved, it is sent to the summary personnel; when rejected, it is returned to the researchers to continue the correlation analysis.
10. The method for constructing an interactive discussion platform based on knowledge graphs according to claim 8, characterized in that: The secondary approval process includes approval or rejection; when approved, a research report is generated; when rejected, the report is returned to the summary personnel for further correlation analysis; the research report includes data and charts.