Positionless aspect based sentiment analysis using attention mechanism

• DOI: 10.1016/j.knosys.2021.107136
• Publication Date: 2021-05-13T15:59:42Z
• Authors: Rohan Kumar Yadav; Lei Jiao; Morten Goodwin; Ole-Christoffer G...
• Keywords: Word embedding; Sentiment Analysis; Polarity (international relations)

Aspect-based sentiment analysis (ABSA) aims at identifying fine-grained polarity of opinion associated with a given aspect word. Several existing articles demonstrated promising ABSA accuracy using positional embedding to show the relationship between an word and its context. In most cases, depends on distance remaining words in context, known as position index sequence. However, these techniques usually employ both complex preprocessing approaches additional trainable architectures obtain state-of-the-art performance. this paper, we simplify by including lexicon replacement masking that carry information word's eliminate embedding. We then adopt novel concise architecture two Bidirectional GRU along attention layer classify based context words. Experiment results simplified significantly improve publicly available datasets, obtaining 81.37%, 75.39%, 80.88%, 89.30% restaurant 14, laptop 15, 16 respectively.