Deep Learning Techniques for Aspect Based Sentiment Analysis
Sijin Chen, Gesangzeren Fnu · 2022
Sentiment analysis is an important tool, which is aiming to use the natural language process to detect and extract the information inside the text in order to understand the emotional states of human beings. Sentiment analysis has been used widely in the world to extract information from a large number of user-generated content. However, with the development of the user needs, sentiment analysis requires improvement with the granularity at the aspect level. It needs to be performed at a more fine-grained level. Aspect-based sentiment analysis (ABSA) is a technique that classifies data by aspect and identifies the sentiment attributed to each aspect. ABSA has been proposed to meet the needs at the aspect level. Aspect extraction and sentiment classification of product reviews and sentiment classification of target-dependent tweets are two main distinct targets of the improvement. Deep learning approaches have the advantage that its ability to capture the syntactic and semantic features within the text, and it does not require the high-level feature engineering. Deep learning has become the prospect of the industry. More and more researchers have tried to apply the deep learning approaches in the ABSA tasks in recent years. Nevertheless, there is a lack of summary of this kind of application, so we review and summarize some of the work for recent years.