Aspect Based Sentiment Analysis Using NeuroNER and Bidirectional Recurrent Neural Network

Abstract

Nowadays, understanding sentiments of what customers say, think and review plays an important part in the success of every business. In consequence, Sentiment Analysis (SA) has been becoming a vital part in both academic and commercial standpoint in recent years. However, most of the current sentiment analysis approaches only focus on detecting the overall polarity of the whole sentence or paragraph. That is the reason why this work focuses on another approach to this task, which is Aspect Based Sentiment Analysis (ABSA). The proposed ABSA system in this paper has two main phases: aspect term extraction and aspect sentiment prediction. For the first phase, as to deal with the named-entity recognition (NER) task, it is performed by reusing the NeuroNER [1] program without any modifications because it is currently one of the best NER tool available. For the sentiment prediction task, a bidirectional gated recurrent unit (BiGRU) Recurrent Neural Network (RNN) model which processes 4 features as input: word embeddings, SenticNet [2], Part of Speech and Distance is implemented. However, this network architecture performance on SemEval 2016 [3] dataset showed some drawbacks and limitations that influenced the polarity prediction result. For this reason, this work proposes some adjustments to the mentioned model to solve the current problems and improve the accuracy of the second task.