A computationally efficient BiLSTM based approach for the binary sentiment classification

Abstract

Sentiment analysis is a crucial task in natural language processing (NLP), aiming at the extraction of opinions expressed in the texts. This paper aims at a computationally efficient approach for the classification of positive and negative sentiments by using publicly available datasets about movie reviews, namely, Movie Review (MR) and Stanford Sentiment Tree (SST2) datasets. We exploited merely one bidirectional long short-term memory (BiLSTM) layer with a global maximum pooling layer and got F1 scores of 85.78 and 80.21 for SST2 and MR datasets, respectively. We concluded that our results are competitive with recently published methods with complex architectures. Also, our approach requires minimal computational cost and thus may help in real-time applications in general opinion categorization.