Tripathi, Yashaswi2021-08-042021-08-042020-0743p.http://hdl.handle.net/10263/7185Dissertation under the supervision Utpal Garain,Professor & Debapriyo Majumdar, Assistant Professor, CVPRSentiment analysis plays an important role in e-commerce, as it allows the industries to better understand the customer experience and its brand value. Aspect Based Sentiment Analysis (ABSA) is a ne-grained version of sentiment analysis. ABSA not only focuses on analysing opinions in a given review but also looks into the several aspects and their sentiments thus giving a much clearer understanding. Aspect extraction is a crucial part of this ABSA task on which much attention has not been paid until recent years. Limited number of training data has made the task further challenging. This project addresses the problem of extraction of aspects from review comments and thereby attempts to improve the state of the art results in ABSA. For language modeling, BERT is used and it's netuned on a novel Neurosyntactic model architecture. POS and dependency tags are used along review comments for extraction of aspect terms. Experiments conducted on SemEval dataset show that the proposed architecture achieves the state of the art results on the dataset.enABSAAspect DetectionNeurosyntactic architecturePOS tagsDEP tagsBERT. 5Other