Target-Based Sentiment Analysis using a BERT Embedded Model

Abstract

Opinion mining or sentiment analysis is used for understanding the opinion of the community on a specific product or a service. In this paper, we investigate the modeling power of contextualized representations from pre-trained language model BERT. There are many methods and techniques used to explore these features from unstructured comments. We developed a different approach to target extraction task in sentiment analysis using a deep learning model combining Gated Recurrent Unit (GRU) and Conditional Random Field (CRF). The task is typically modeled as a sequence labeling problem and solved using state-of-the-art labelers such as CRF. This model is trained on labeled data to extract and classify feature sets. Experiments have been conducted on two benchmark datasets and our framework achieves consistently superior results.