Research On Optimization Of Relation Extraction Model For TCM Texts

Chinese medicine culture is the essence of intelligence of the Chinese people,covering a lot of Chinese medicine knowledge.Due to the large scale of ancient Chinese medicine classics,to extract TCM knowledge from TCM texts that have important reference value for modern medicine,it is necessary to extract information from TCM texts.The goal of relationship extraction is to find and extract the relationship between entities from the text.With a large number of mature applications of deep learning models in the field of natural language processing,many deep learning models have achieved very good results in all kinds of research of natural language processing.However,due to the complex structure and the large scale of the model,its application in the special and complex language structure of TCM text is limited.To further improve the performance of the relationship extraction model in the task of extracting TCM text relations,at the same time,reduce the amount of model calculation as much as possible and improve the stability of the model.The main research work is as follows:First,conduct extensive research on the current models that are widely used in the field of natural processing and show good performance.Through summary and classification,they can be roughly split into three types: traditional deep learning models and traditional deep learning models overlay attention modules and new deep learning models based on attention mechanisms.Through the analysis and research of these models,it can be concluded that the role of attention mechanisms is very important for the extraction of deep semantic relations.The performance of the deep learning model represented by Transformer in relation extraction tasks under different backgrounds is greatly improved compared with the traditional deep learning model.The research results on the optimization of relation extraction model show that the scale and computational complexity of the new deep learning model based on the attention mechanism are the key issues that limit its application.Secondly,in view of the structural characteristics of the Transformer model in the task of extracting TCM text relations,we propose the multi-head attention mechanism in the Transformer model.Two methods are used to analyze the importance of all attention heads in the model.After the analysis results are obtained,use the method based on regularization suppresses unimportant attention heads in the model,thereby reducing the number of model calculations and improving model accuracy.By inputting the corpus of the same scale into the unoptimized and optimized Transformer model to compare the effects of TCM relationship extraction,experiments prove that the optimization method is effective for TCM text relationship extraction tasks.Figure 20;Table 8;Reference 47...