Research On Textual Entailment Recognition Method Based On Joint Features

With the rapid development of the Internet,natural language processing is gradually in the forefront of science and favored by researchers.Natural language processing aims to enhance the communication between artificial intelligence and human beings,covering a number of technical tasks such as information extraction,information retrieval,machine translation,etc.it can solve the problem that human beings need a lot of energy to complete such a huge amount of data task.At the same time,the goal of natural language processing is to make the computer own human's thinking mode,understanding the text from the semantic level and completing the task more efficiently.Therefore,textual entailment recognition technology emerges as the times require.Textual entailment recognition refers to judging whether the hypothesis text can be inferred from the premise text by giving the premise text and the hypothesis text,so as to determine whether there is entailment relationship between the given texts.The traditional textual entailment recognition methods need a lot of preprocessing tools such as feature extraction,rule setting,part of speech tagging and syntactic analysis.With the development of natural language processing,researchers use neural networks to identify textual entailment.This method can simulate human brain to a certain extent and overcome the problem of sparse features through continuous vectorization.However,the comparison of semantic information is still limited,resulting in the recognition accuracy is not very ideal.Therefore,this thesis proposes a textual entailment recognition method based on joint features.Firstly,feature engineering is proposed to improve the accuracy of textual feature extraction.Feature engineering model is constructed by text vectorization,feature enhancement and feature integration to extract text feature.Secondly,an improved scheme of ESIM model based on text matching depth neural network is proposed to improve the ability of information comparison at semantic level and integrate attention mechanism into ESIM mode——traditional attention mechanism completes text soft alignment and self-attention mechanism enhances sentence representation.Finally,this paper proposes a textual entailment recognition method based on joint features,which integrates Feature Engineering with ESIM model integrated into attention mechanism,the feature fusion method based on D-S evidence theory and completes experimental comparative analysis.The experimental results show that the recognition accuracy of textual entailment recognition method based on joint features in SNLI data set and RTE-7 data set is improved,which proves the effectiveness of the method.