IEEE Project Abstract

Tourist reviews are information sources fortravelers to know about tourist places. Unfortunately, somereviews are irrelevant and become noisy data. Aspect-basedsentiment classification methods have shown promise insuppressing the noise. However, not much research has beendone on automatic aspect identification, and identification ofimplicit, infrequent and co-referential aspects, resulting inmisclassifications. This paper presents a framework of aspectbased sentiment classification that will not only identify theaspects very efficiently but can perform classification taskwith high accuracy. The framework has been implemented asa mobile app that helps tourists find the best restaurant orhotel in a city, and performance has been evaluated byconducting experiments on real-world datasets with excellentresults (85% identification and 90% classification).