Intelligent generation method and system of traditional Chinese medicine large language model AI question bank based on dynamic semantic partitioning, knowledge graph and multi-hop path planning

Abstract

The invention relates to a traditional Chinese medicine large language model AI question bank intelligent generation method and system based on dynamic semantic partitioning, a knowledge graph and multi-hop path planning, and the method comprises the steps: carrying out the preprocessing of traditional Chinese medicine text data, carrying out the dynamic semantic partitioning of the preprocessed data, and obtaining a plurality of semantic blocks; constructing a knowledge graph based on the preprocessed data, carrying out multi-hop path planning in the knowledge graph, and selecting path data which most meets the requirement of the question; based on the semantic blocks and the path data, training a large language model; and generating a traditional Chinese medicine question bank based on the trained large language model. Fine-grained semantics of a traditional Chinese medicine text are accurately captured through a dynamic semantic partitioning technology, efficient association of knowledge points is realized by means of a knowledge graph, and the logicality and diversity of question generation are improved by utilizing multi-hop path planning, so that the intelligence level and question quality of question bank generation are remarkably improved, and the efficiency of question bank generation is improved. Powerful technical support is provided for the fields of traditional Chinese medicine education, training, examination and the like.

Description

Technical Field

[0001] This invention relates to the field of traditional Chinese medicine education technology, and in particular to an intelligent generation method and system for a large language model AI question bank of traditional Chinese medicine based on dynamic semantic segmentation, knowledge graph and multi-hop path planning. Background Technology

[0002] Traditional question bank generation methods primarily rely on manual writing and selection of questions. This method is not only time-consuming and labor-intensive, but also struggles to ensure the comprehensiveness and consistency of the questions. Given the complexity of the Traditional Chinese Medicine (TCM) knowledge system, encompassing a vast amount of classical literature and clinical experience, manually writing questions is insufficient to cover all knowledge points, easily leading to a limited and outdated question bank that fails to meet the needs of modern TCM education.

[0003] In recent years, with the development of artificial intelligence technology, especially the application of Large Language Models (LLM), the field of medical education has begun to explore the use of these technologies to improve the efficiency and quality of question bank generation. For example, some studies have proposed automatic question generation technology based on Large Language Models. Through training with massive amounts of data, LLM can generate questions and answers that meet the requirements, significantly improving question generation efficiency. However, these methods still have shortcomings when dealing with the complex semantics and cross-knowledge point connections in the field of Traditional Chinese Medicine.

[0004] In the field of Traditional Chinese Medicine (TCM), some research and patents have attempted to leverage large language models and knowledge graphs to improve the quality of question bank generation. For example, one method and system for automatic question generation based on a large language model generates questions and answers that match user input. Furthermore, the ShenNong-TCM model, through fine-tuning based on a TCM knowledge graph, can generate TCM-related instructions and responses, but there is still room for improvement in multi-hop path planning and dynamic semantic segmentation.

[0005] While existing technologies have made some progress in generating TCM question banks, they still have limitations in handling complex semantics, knowledge associations, and the flexibility of question generation. Summary of the Invention

[0006] The purpose of this invention is to propose an intelligent question bank generation method and system based on dynamic semantic segmentation, knowledge graphs, and multi-hop path planning using a large-scale language model of Traditional Chinese Medicine (TCM). This addresses the problems existing in the prior art. Through dynamic semantic segmentation, it accurately captures the fine-grained semantics of TCM knowledge; it leverages knowledge graphs to achieve efficient association of knowledge points; and it utilizes multi-hop path planning to enhance the logic and diversity of generated questions. This method not only significantly improves the intelligence level of question bank generation but also greatly enhances the quality and educational effectiveness of the questions, providing strong technical support for TCM education.

[0007] To achieve the above objectives, the present invention provides the following solution: A method for intelligent generation of a TCM big data language model AI question bank based on dynamic semantic segmentation, knowledge graph, and multi-hop path planning includes: Preprocessing of TCM text data; The preprocessed data is dynamically semantically segmented to obtain several semantic blocks; A knowledge graph is constructed based on the preprocessed data, and multi-hop path planning is performed in the knowledge graph to select the path data that best meets the requirements of the question. Based on the semantic blocks and path data, a large language model is trained. A TCM question bank is generated based on a trained large language model.

[0008] Optionally, preprocessing of TCM text data includes: The TCM text data is cleaned, segmented, and labeled.

[0009] Optionally, dynamic semantic segmentation of the preprocessed data includes: Natural language processing techniques are used to perform semantic analysis on preprocessed text data to extract key semantic information. Based on the key semantic information, the text data is divided into several semantic blocks, each containing a complete semantic unit.

[0010] Optionally, constructing a knowledge graph based on the preprocessed data includes: Extract entities from the preprocessed text related to Traditional Chinese Medicine (TCM); wherein the entities include: Chinese herbal medicines, acupoints, and symptoms; Identify the relationships between the entities; The extracted entities and relationships are integrated to construct a knowledge graph.

[0011] Optionally, multi-hop path planning in the knowledge graph includes: In a knowledge graph, based on the query requirements, a starting point and an ending point are set, and a multi-hop path search is performed to find the path from the starting node to the target node. Optimize the searched paths and select the path data that best meets the requirements of the question.

[0012] Optionally, training a large language model based on the semantic blocks and path data includes: The large language model is trained on the semantic blocks and path data, and the model parameters are adjusted until the model converges; wherein, the training process adopts distributed training technology.

[0013] Optionally, the generation of a TCM question bank based on the trained large language model includes: Design a variety of question templates; Based on the path data from the multi-hop path planning and the output of the large language model, fill in the question template and generate specific questions; The generated questions are reviewed.

[0014] A smart question bank generation system for TCM big data language models based on dynamic semantic segmentation, knowledge graphs, and multi-hop path planning, for implementing the methods described above, the system comprising: The preprocessing module is used to preprocess TCM text data; The dynamic semantic chunking module is used to perform dynamic semantic chunking on the preprocessed data to obtain several semantic blocks. The knowledge graph construction module is used to build knowledge graphs based on preprocessed data; The multi-hop path planning module is used to perform multi-hop path planning in the knowledge graph and select the path data that best meets the requirements of the question. The large model training module is used to train a large language model based on the semantic blocks and path data; The question bank generation module is used to generate a TCM question bank based on a trained large language model.

[0015] Optionally, the dynamic semantic segmentation module includes: The semantic analysis unit is used to perform semantic analysis on preprocessed text data using natural language processing technology to extract key semantic information. The dynamic segmentation unit is used to divide the text data into several semantic blocks based on the key semantic information, and each semantic block contains a complete semantic unit.

[0016] Optionally, the knowledge graph construction module includes: An entity extraction unit is used to extract entities in the field of traditional Chinese medicine from preprocessed text; wherein, the entities include: Chinese herbal medicines, acupoints, and diseases; A relationship extraction unit is used to identify the relationships between the entities; Building units are used to integrate extracted entities and relationships to construct a knowledge graph.

[0017] Optionally, the multi-hop path planning module includes: The path search unit is used in the knowledge graph to set the starting point and the ending point according to the query requirements, and to perform multi-hop path search to find the path from the starting node to the target node. The path optimization unit is used to optimize the searched paths and select the path data that best meets the requirements of the question.

[0018] Optionally, the large model training module: The large language model is trained based on the semantic blocks and path data, and the model parameters are adjusted until the model converges; the training process adopts distributed training technology.

[0019] Optionally, the question bank generation module includes: The question template design unit is used to design various question templates; The question generation unit is used to fill the question template and generate specific questions based on the path data of the multi-hop path planning and the output of the large language model; The question review unit is used to review the generated questions.

[0020] The beneficial effects of this invention are as follows: This invention proposes a method and system for intelligent generation of TCM AI question banks based on dynamic semantic segmentation, knowledge graphs, and multi-hop path planning. The method uses dynamic semantic segmentation technology to perform fine-grained segmentation of TCM texts, ensuring semantic coherence and the completeness of knowledge points; it utilizes a knowledge graph to construct a TCM knowledge system, distinguishing core, key, and peripheral knowledge points and assigning relational weights; and it combines a multi-hop path planning algorithm to optimize the question generation path, generating high-quality questions covering various question types such as multiple choice, true / false, fill-in-the-blank, definition, and open-ended questions. The system mainly consists of a semantic segmentation module, a knowledge graph module, a path planning module, and a question bank generation module. These modules work collaboratively through electrical connections to achieve efficient and intelligent question bank generation. This invention effectively improves the intelligence level and question quality of TCM question bank generation, and is applicable to TCM education, training, and assessment, significantly improving question creation efficiency, enriching question type diversity, and ensuring appropriate question difficulty and comprehensive knowledge coverage. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in 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.

[0022] Figure 1 This is a schematic diagram of the process of intelligent generation method of AI question bank of TCM big language model based on dynamic semantic segmentation, knowledge graph and multi-hop path planning according to an embodiment of the present invention; Figure 2 This is a schematic diagram of an AI-powered question bank intelligent generation system for a large language model of traditional Chinese medicine based on dynamic semantic segmentation, knowledge graph, and multi-hop path planning, according to an embodiment of the present invention. Detailed Implementation

[0023] 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.

[0024] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0025] like Figure 1 As shown, this embodiment proposes an intelligent generation method for a TCM big language model AI question bank based on dynamic semantic segmentation, knowledge graph, and multi-hop path planning, including: Preprocessing of TCM text data; The preprocessed data is dynamically semantically segmented to obtain several semantic blocks; A knowledge graph is constructed based on the preprocessed data, and multi-hop path planning is performed in the knowledge graph to select the path data that best meets the requirements of the question. Train a large language model based on semantic blocks and path data; A TCM question bank is generated based on a trained large language model.

[0026] Furthermore, preprocessing of TCM text data includes: The TCM text data was cleaned, segmented, and labeled.

[0027] Specifically, in this embodiment, the input TCM text data is cleaned, segmented, and labeled. The specific steps are as follows: Data cleaning: Removing noisy data from text, such as punctuation marks and irrelevant characters; Word segmentation: The text is segmented using a professional TCM word segmentation tool; Annotation: Part-of-speech tagging and entity recognition are performed on the segmented text to annotate the names of Chinese medicines, acupoints, diseases, etc.

[0028] Furthermore, dynamic semantic segmentation of the preprocessed data includes: Natural language processing techniques are used to perform semantic analysis on preprocessed text data to extract key semantic information. Based on key semantic information, the text data is divided into several semantic blocks, each containing a complete semantic unit.

[0029] Specifically, in this embodiment, semantic segmentation is performed by using a dynamic semantic segmentation algorithm. Based on the semantic structure and contextual information of the text, the text is divided into multiple semantic blocks. Each semantic block contains complete semantic information and can independently express a complete meaning.

[0030] By using dynamic programming, the segmentation process is optimized to ensure the rationality and accuracy of segmentation. The sliding window algorithm ensures contextual coherence and can dynamically adjust the segment size based on semantic similarity to adapt to texts of different lengths and complexities.

[0031] Feature extraction: Multi-dimensional features are extracted for each semantic block, including part-of-speech tagging, syntactic structure analysis, semantic role tagging, etc., to mine the semantic information of the semantic block from different perspectives.

[0032] Deep learning techniques, such as convolutional neural networks (CNNs) or recurrent neural networks (RNNs), are used to further extract deeper semantic features, thereby improving the depth and accuracy of semantic understanding.

[0033] Furthermore, constructing a knowledge graph based on the preprocessed data includes: Extract entities from the preprocessed text related to Traditional Chinese Medicine (TCM); these entities include: Chinese herbal medicines, acupoints, and symptoms. Identify relationships between entities; The extracted entities and relationships are integrated to construct a knowledge graph.

[0034] Specifically, in this embodiment, knowledge extraction involves automatically extracting entities and relationships from TCM texts. Entities include Chinese medicine, symptoms, and treatment methods, while relationships include etiology, efficacy, and treatment methods. Entities and their relationships are extracted from the text using a Large Language Model (LLM), and the entities are classified and scored to distinguish between core knowledge points, key knowledge points, and peripheral knowledge points.

[0035] By employing Named Entity Recognition (NER) and Relation Extraction (RE) techniques, the accuracy of knowledge extraction is improved, ensuring that the extracted knowledge points are accurate.

[0036] Knowledge Fusion: The extracted knowledge is integrated with the existing TCM knowledge base to eliminate redundancy and conflicts, ensuring the unity and consistency of knowledge. Entity alignment and relation alignment techniques are used to further optimize the knowledge fusion process and improve its efficiency and quality.

[0037] Knowledge Graph Construction: Based on the fused knowledge, a knowledge graph for the field of Traditional Chinese Medicine is constructed, clarifying the relationships between entities. Knowledge points are treated as nodes, and the relationships between them are represented as edges, with each edge assigned a weight to indicate the strength of the association. Graph database technology, such as Neo4j, is used to store and manage the knowledge graph, improving query and retrieval efficiency and providing efficient knowledge retrieval support for subsequent path planning and question generation.

[0038] Furthermore, multi-hop path planning in knowledge graphs includes: In a knowledge graph, based on the query requirements, a starting point and an ending point are set, and a multi-hop path search is performed to find the path from the starting node to the target node. Optimize the searched paths and select the path data that best meets the requirements of the question.

[0039] Specifically, in this embodiment, path search: In the knowledge graph, multi-hop path search is performed according to the query requirements to find the optimal path from the starting node to the target node. Single-hop paths are used to generate simple question types, while multi-hop paths are used to generate complex question types. Through the multi-hop path planning algorithm, comprehensive questions covering multiple knowledge points can be generated.

[0040] Graph search algorithms, such as breadth-first search (BFS) or depth-first search (DFS), are used to ensure the comprehensiveness and diversity of paths.

[0041] Path optimization: The searched paths are optimized to ensure their rationality and effectiveness. A path scoring mechanism is used to evaluate the relevance and importance of each path, selecting the optimal path. The path weights are dynamically adjusted using a path weight calculation formula, combined with the node's base weight and temporary counts, to ensure comprehensive question coverage and diversity.

[0042] Path Output: Outputs the optimized path for use by subsequent modules, presenting the path information in a structured manner for easier processing and analysis.

[0043] Furthermore, training a large language model based on semantic blocks and path data includes: The large language model is trained using semantic blocks and path data, and the model parameters are adjusted until the model converges; the training process employs distributed training technology.

[0044] Specifically, in this embodiment, model selection involves choosing a suitable large-scale language model for the field of Traditional Chinese Medicine (TCM), such as Transformer or BERT. The model is then customized based on the characteristics of TCM texts to enhance its adaptability. Furthermore, the model architecture is optimized by combining TCM expertise with the characteristics of the corpus, thereby improving the model's language understanding and generation capabilities within the TCM domain.

[0045] Model Training: The large language model is trained using the prepared data, and model parameters are adjusted until the model converges. Through training with a large amount of TCM text data, the model acquires accurate language expression and logical reasoning abilities. Distributed training technology is employed to improve training efficiency and model performance, ensuring that the model can complete training within a reasonable time and possess efficient reasoning capabilities.

[0046] Furthermore, the generation of a TCM question bank based on the trained large language model includes: Design a variety of question templates; Based on the path data from the multi-hop path planning and the output of the large language model, fill in the question template and generate specific questions. The generated questions are reviewed.

[0047] Specifically, in this embodiment, the generation rules are as follows: rules for generating the question bank are formulated, including question types (multiple choice, fill-in-the-blank, short answer, etc.), difficulty (beginner, intermediate, advanced), distribution of knowledge points, etc. The generation rules are dynamically adjusted according to the needs of TCM education to ensure the diversity and practicality of the question bank.

[0048] By combining path complexity and feedback from large language models, the difficulty of the questions is comprehensively evaluated to ensure that the difficulty of the generated questions is suitable for different levels of teaching needs.

[0049] Model Invocation: The pre-trained large language model is invoked to generate questions according to the generation rules. The model's generation capabilities are utilized to ensure the fluency and logical coherence of the questions. The question types and quantities are dynamically allocated based on the importance of the path. The higher the importance of the path, the more types and quantities of questions are generated, ensuring comprehensive and diverse question coverage.

[0050] Question Review: Generated questions undergo manual review to ensure their accuracy and reasonableness. A team of human reviewers reviews each question individually to guarantee quality. Automated review technologies, such as Natural Language Processing (NLP), are employed to assist human review, improving efficiency. An automated review system performs initial screening of questions, reducing the workload of manual review.

[0051] Question Bank Output: The system compiles approved questions into a question bank and outputs the final results. Multiple output formats are provided, such as JSON and XML, for easy storage and use of the question bank. Users can customize question difficulty, question type ratios, and knowledge point coverage to meet the needs of different users.

[0052] The principle of the method in this embodiment is explained below: Dynamic semantic segmentation principle: Based on the semantic structure and context information of the text, a dynamic programming algorithm is used to segment the text to ensure the rationality and accuracy of the segmentation. The sliding window algorithm ensures the coherence of the context and can dynamically adjust the segment size according to semantic similarity.

[0053] Knowledge graph construction principle: Through knowledge extraction, fusion and graph construction techniques, a structured TCM knowledge system is formed. Entities and relationships are extracted from text using a large language model, and a weighted graph is constructed to clarify the relationships between entities.

[0054] Multi-hop path planning principle: Using graph search algorithms, the optimal multi-hop path is found in the knowledge graph to improve knowledge relevance. By using the path weight calculation formula, combined with the basic weight and temporary count of the node, the path weight is dynamically adjusted to ensure the coverage and diversity of questions.

[0055] The training principle of the large language model is to train the model with a large amount of TCM text data, so that the model has the ability to understand and generate language in the field of TCM. Combined with the professional knowledge and corpus characteristics of TCM, the model architecture is optimized to improve the adaptability of the model.

[0056] Question bank generation principle: Based on a trained large language model and generation rules, TCM questions that meet the requirements are automatically generated. The difficulty of the questions is comprehensively evaluated by combining path complexity and feedback from the large language model to ensure that the difficulty of the generated questions is adapted to the teaching needs of different levels.

[0057] like Figure 2 As shown, this embodiment also proposes an AI-powered question bank intelligent generation system for TCM big language models based on dynamic semantic segmentation, knowledge graphs, and multi-hop path planning, including: Data preprocessing module 101 is used to preprocess TCM text data; The dynamic semantic segmentation module 102 is used to perform dynamic semantic segmentation on the preprocessed data to obtain several semantic blocks. Knowledge graph construction module 103 is used to construct knowledge graphs based on preprocessed data; The multi-hop path planning module 104 is used to perform multi-hop path planning in the knowledge graph and select the path data that best meets the requirements of the question. Large model training module 105 is used to train large language models based on semantic blocks and path data; The question bank generation module 106 is used to generate a TCM question bank based on a trained large language model.

[0058] Furthermore, the data preprocessing module 101 is responsible for cleaning, segmenting, and labeling the input TCM text data. The specific steps are as follows: Data cleaning: Removing noisy data from text, such as punctuation marks and irrelevant characters.

[0059] Word segmentation: The text is segmented using a professional TCM word segmentation tool.

[0060] Annotation: Part-of-speech tagging and entity recognition are performed on the segmented text to annotate the names of Chinese medicines, acupoints, diseases, etc.

[0061] Furthermore, the dynamic semantic segmentation module 102 includes: The semantic analysis unit is used to perform semantic analysis on preprocessed text data using natural language processing technology to extract key semantic information. Dynamic chunking units are used to divide text data into several semantic blocks based on key semantic information. Each semantic block contains a complete semantic unit.

[0062] Furthermore, the knowledge graph construction module 103 includes: The entity extraction unit is used to extract entities in the field of traditional Chinese medicine from preprocessed text; the entities include: Chinese herbal medicines, acupoints, and diseases. The relation extraction unit is used to identify the relationships between entities; Building units are used to integrate extracted entities and relationships to construct a knowledge graph.

[0063] Furthermore, the multi-hop path planning module 104 includes: The path search unit is used in the knowledge graph to set the starting point and the ending point according to the query requirements, and to perform multi-hop path search to find the path from the starting node to the target node. The path optimization unit is used to optimize the searched paths and select the path data that best meets the requirements of the question.

[0064] Furthermore, large model training module 105: The large language model is trained based on semantic blocks and path data, and the model parameters are adjusted until the model converges; the training process adopts distributed training technology.

[0065] Furthermore, the question bank generation module 106 includes: The question template design unit is used to design various question templates; The question generation unit is used to fill the question template and generate specific questions based on the path data of the multi-hop path planning and the output of the large language model; The question review unit is used to review the generated questions.

[0066] Description of modules and functions in this embodiment: Dynamic semantic segmentation module: responsible for semantic segmentation of text, improving the depth of semantic understanding, and providing high-quality semantic units for subsequent knowledge graph construction and question generation.

[0067] Knowledge Graph Construction Module: Constructs a knowledge graph for Traditional Chinese Medicine (TCM) to provide a foundation for knowledge association. Through technologies such as entity recognition, relation extraction, and graph construction, it transforms knowledge points in TCM texts into a visualized knowledge network.

[0068] Multi-hop path planning module: Performs path planning in the knowledge graph to enhance knowledge relevance. It finds the optimal path from the starting node to the target node through a graph search algorithm, providing path support for generating questions of different difficulty levels.

[0069] Large Language Model Training Module: Trains a large language model for Traditional Chinese Medicine (TCM), providing the core capability for question generation. Through training with a large amount of TCM text data, the model acquires accurate language expression and logical reasoning abilities.

[0070] Question bank generation module: Based on the model, a TCM question bank is generated to achieve the final goal. It automatically generates high-quality questions covering various question types according to the generation rules, and ensures the accuracy and rationality of the questions through a combination of manual and automated review.

[0071] This invention relates to an intelligent generation method and system for a TCM (Traditional Chinese Medicine) large-scale language model AI question bank based on dynamic semantic segmentation, knowledge graphs, and multi-hop path planning. It aims to address problems in existing TCM question bank generation processes, such as insufficient semantic understanding, weak knowledge correlation, and unreasonable path planning. This invention accurately captures the fine-grained semantics of TCM texts through dynamic semantic segmentation technology, achieves efficient association of knowledge points using knowledge graphs, and enhances the logic and diversity of question generation through multi-hop path planning. This significantly improves the intelligence level and question quality of question bank generation, providing strong technical support for TCM education, training, and assessment.

[0072] Through the above technical solution, the present invention can generate TCM question banks efficiently and intelligently, significantly improving the quality and practicality of the question banks, meeting the needs of TCM education and research, and providing an innovative intelligent solution for the field of TCM education.

[0073] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for intelligent generation of a TCM big language model AI question bank based on dynamic semantic segmentation, knowledge graph, and multi-hop path planning, characterized in that... include: Preprocessing of TCM text data; The preprocessed data is dynamically semantically segmented to obtain several semantic blocks; A knowledge graph is constructed based on the preprocessed data, and multi-hop path planning is performed in the knowledge graph to select the path data that best meets the requirements of the question. Based on the semantic blocks and path data, a large language model is trained. A TCM question bank is generated based on a trained large language model.

2. The method for intelligent generation of a TCM big language model AI question bank based on dynamic semantic segmentation, knowledge graph, and multi-hop path planning as described in claim 1, is characterized in that... Preprocessing of TCM text data includes: The TCM text data is cleaned, segmented, and labeled.

3. The method for intelligent generation of TCM big language model AI question bank based on dynamic semantic segmentation, knowledge graph and multi-hop path planning as described in claim 1, is characterized in that, Dynamic semantic chunking of preprocessed data includes: Natural language processing techniques are used to perform semantic analysis on preprocessed text data to extract key semantic information. Based on the key semantic information, the text data is divided into several semantic blocks, each containing a complete semantic unit.

4. The method for intelligent generation of a TCM big language model AI question bank based on dynamic semantic segmentation, knowledge graph, and multi-hop path planning as described in claim 1, is characterized in that... Building knowledge graphs based on preprocessed data includes: Extract entities from the preprocessed text related to Traditional Chinese Medicine (TCM); wherein the entities include: Chinese herbal medicines, acupoints, and symptoms; Identify the relationships between the entities; The extracted entities and relationships are integrated to construct a knowledge graph.

5. The method for intelligent generation of a TCM big language model AI question bank based on dynamic semantic segmentation, knowledge graph, and multi-hop path planning as described in claim 1, is characterized in that... Multi-hop path planning in the knowledge graph includes: In a knowledge graph, based on the query requirements, a starting point and an ending point are set, and a multi-hop path search is performed to find the path from the starting node to the target node. Optimize the searched paths and select the path data that best meets the requirements of the question.

6. The method for intelligent generation of a TCM big language model AI question bank based on dynamic semantic segmentation, knowledge graph, and multi-hop path planning as described in claim 1, is characterized in that... Training a large language model based on the aforementioned semantic blocks and path data includes: The large language model is trained based on the semantic blocks and path data, and the model parameters are adjusted until the model converges; the training process adopts distributed training technology.

7. The method for intelligent generation of a TCM big language model AI question bank based on dynamic semantic segmentation, knowledge graph, and multi-hop path planning as described in claim 1, is characterized in that... The TCM question bank generated based on the trained large language model includes: Design a variety of question templates; Based on the path data from the multi-hop path planning and the output of the large language model, fill in the question template and generate specific questions; The generated questions are reviewed.

8. A smart question bank generation system for a large language model of Traditional Chinese Medicine based on dynamic semantic segmentation, knowledge graph and multi-hop path planning, characterized in that, The system for implementing the method as described in any one of claims 1-7 comprises: The preprocessing module is used to preprocess TCM text data; The dynamic semantic chunking module is used to perform dynamic semantic chunking on the preprocessed data to obtain several semantic blocks. The knowledge graph construction module is used to build knowledge graphs based on preprocessed data; The multi-hop path planning module is used to perform multi-hop path planning in the knowledge graph and select the path data that best meets the requirements of the question. The large model training module is used to train a large language model based on the semantic blocks and path data; The question bank generation module is used to generate a TCM question bank based on a trained large language model.

9. The intelligent generation system for a TCM big language model AI question bank based on dynamic semantic segmentation, knowledge graph, and multi-hop path planning as described in claim 8, is characterized in that, The dynamic semantic segmentation module includes: The semantic analysis unit is used to perform semantic analysis on preprocessed text data using natural language processing technology to extract key semantic information. The dynamic segmentation unit is used to divide the text data into several semantic blocks based on the key semantic information, and each semantic block contains a complete semantic unit.

10. The AI-powered question bank intelligent generation system for a large language model of Traditional Chinese Medicine based on dynamic semantic segmentation, knowledge graph, and multi-hop path planning as described in claim 8, is characterized in that... The knowledge graph construction module includes: An entity extraction unit is used to extract entities in the field of traditional Chinese medicine from preprocessed text; wherein, the entities include: Chinese herbal medicines, acupoints, and diseases; A relationship extraction unit is used to identify the relationships between the entities; Building units are used to integrate extracted entities and relationships to construct a knowledge graph.

Images