Research On Aspect-Based Sentiment Analysis

Aspect-based sentiment analysis(ABSA)is a fine-grained task in sentiment analysis whose goal is to identify the sentiment polarity of a particular target in its context.ABSA contains two basic subtasks: Aspect Category Sentiment Analysis(ACSA)and Aspect Term Sentiment Analysis(ATSA).Different context words have different effects on determining the sentiment polarity of sentences facing a specific aspect category/ term,which increases the difficulty of analyzing aspect-based sentiment.Existing methods use deep learning models to model context,aspects,and the relationships between them,but some useful clues(eg,context,vocabulary,and syntax)are still not fully considered and utilized.In the same time,research on ACSA is limited because most of them use the general model with ATSA as the core research.Therefore,this paper proposes a new method to solve the above problems.First,this article uses aspect category labels as a guide,and proposes a new method for ACSA.By constructing an adversarial training model,information that is not directly related to and directly related to the aspect category is captured to obtain the shared and private representation of the aspect category.In the end,the two parts are merged in a splicing manner,which greatly excavates and retains the category-related context.Secondly,this article uses context,vocabulary and syntactic prompt features as guidance to improve the classification accuracy of sentiment analysis based on aspect terms.A subnetwork is introduced to represent the term in the sentence and dictionary embedding is used to incorporate other vocabulary cues to fully consider the context information.At the same time,this paper designs a new attention module to explain in detail the syntactic dependence cues between words in attention reasoning.