Research On Text Sentiment Triplet Extraction Algorithm

With the rapid development of Internet technology,people actively share their views and attitudes on social events,public figures and consumer services online.Through analysis and mining of text comments from various fields,sentiment analysis helps to grasp the trend of public opinion,understand the media and public’s views on hot events,help users understand the reputation of products,and help businesses improve their products and services.Therefore,sentiment analysis has become a current research hotspot and has great value.In order to analyze user comments in a fine-grained manner,the sentiment analysis task gradually evolved from simply judging the sentiment polarity of the text to extracting sentiment triplets of the text.Text Sentiment Triplet Extraction(TSTE)aims to extract triplets from sentences,where each triplet includes an entity,its associated sentiment,and the opinion span explaining the reason for the sentiment.Most existing research addresses this problem in a multi-stage pipeline manner,which neglects the mutual information between such three elements and has the problem of error propagation.Aiming at the deficiencies of existing work,a Semantic and Syntactic Enhanced Text Sentiment Triplet Extraction Model(S3E2)is proposed,which makes full use of the syntactic and semantic relations between triplets,and extract them together.Specifically,for the task of TSTE,S3E2 designs a Graph-Sequence dual representation and modeling paradigm: using graphs to represent the semantic and syntactic relationships between word pairs in sentences,and encoding them through Graph Neural Networks(GNNs).At the same time,the original sentence is modeled through the Bidirectional Long and Short-Term Memory networks(Bi LSTM)to preserve the order information.Under this setting,the model further applies a more effective inference strategy to extract triplets.Extensive experiments on four benchmark datasets show that the performance of S3E2 is significantly better than existing methods,which proves the superiority and flexibility of S3E2 in an end-to-end manner.