Construction And Application Of Knowledge Graph Of Chinese Herbal Medicine Based On Deep Learning

Chinese herbal medicine has a history of more than two thousand years,which plays an important role in the Chinese inheritance.To address the unrelated and unstructured characteristics of Chinese herbal knowledge in the information age,this study adopts knowledge graph technology to solve the multi-source heterogeneity problem of Chinese herbal data.In terms of data processing,we have extracted the herbal entities in the text by data preprocessing of unstructured texts from professional books and websites,and introduced the Attention mechanism to improve the BiLSTM+CRF named entity recognition model to solve the named entity recognition problem of long sentences about herbal texts.In terms of pattern,we have designed the conceptual model of herbal knowledge graph according to its professional classification system.After knowledge fusion and representation,the storage and visualization of the knowledge graph of Chinese herbal medicines has been established by the Neo4 j graph database.We have also designed and set the query system of Chinese herbal medicines information.The main work consists of the following three areas:(1)Pre-process Chinese herbal medicine data and design an improved named entity identification model.The professional data are based on the classical textbook Traditional Chinese Medicine,supplemented by a professional website of Chinese medicine.We have used Python’s Requests and Beautiful Soup crawler module to crawl the data,and use YEDDA annotation to pre-process the source data such as corpus annotation.After combining BiLSTM+CRF named entity recognition model to Attention mechanism,the mode has been improved to a six-layer recognition model.As a result,its F1 value of model recognition evaluation index has arrived 94.42%,which is 1.74 percentage points higher than the original one.This improves the accuracy of named entity recognition mission.(2)Establish a knowledge graph of Chinese herbal medicines based on seven types of entities,and realize data visualization.The knowledge graph in traditional Chinese medicine field mainly focuses on diagnosis and treatment.But this graph adds "abstracts from ancient texts" and "specific abstracts",based on "drug","efficacy","disease","property" and "category".We have designed the knowledge graph conceptual model of Chinese herbal medicine,which includes seven types of entities and six types of relationships.We have collected 21345 entities,which are stored in the Neo4 j graph database and displayed visually.(3)Create a query system for Chinese herbal medicine based on knowledge graph.The system directly connects to the graph database and displays the information visually from knowledge graph by Cytoscape.js.This integrates information retrieval,knowledge navigation and mapping knowledge visualization together.This study strengthens the relationship between abstracts from ancient texts and entities,by improving the conceptual model of herbal knowledge graph and the named entity recognition model.These works offer evidence-based inquiry into the Chinese herbal knowledge,which promotes further research on knowledge graph of Chinese medicine.