Adaptation of a BERT Model to the India Restaurant data using Rule-based Approach for Aspect-based Sentiment Analysis

Aspect-Based Sentiment Analysis (ASBA) concept has been worked extensively as a popular technique in consumer research. The commercial applications though limited are being tested. The research mainly focuses on using the existing ASBA model to be adopted, tested and validated for use in the context of consumer reviews for restaurants in Indian geography. We used the rule-based approach instead of the supervised or unsupervised based machine learning approach to adapt the ABSA model with a hypothesis that the rule-based approach would yield F Square estimates equivalent to supervise or unsupervised learning approach.Our research indicated the rule-based approach was equivalent to the supervised approach of training and presented better F square estimates than the unsupervised approach. However, it required more manpower and effort to build the rule-based adaptation while the adopted model will demand constant revisions and maintenance based on the change in the consumer vocabulary.