Supervised Semantic Analysis of Product Reviews Using Weighted k-NN Classifier

On line shopping, today, has become the call of the day. People are showing more inclination towards on line shops, due to large variety of options at fingertips, ease of access, access to global products. Further the buyer also benefits from information regarding user review of products, comparison of similar products etc. Therefore the importance of product review is also escalating exponentially. Most of the existing sentiment analysis systems require large training datasets and complex tools for implementation. This paper presents a Two-Parse algorithm with a training dataset of approximately 7,000 keywords, for automatic product review analysis. The proposed algorithm is more efficient as compared to some of the popular review analysis systems with enormous datasets. This algorithm is a solution to a very common problem of high polarity of datasets. This paper proposes a Weighted k-Nearest Neighbor (Weighted k-NN) Classifier which achieves a better efficiency than the classical k-Nearest Neighbor Classifier. The proposed Classifier is capable of successfully classifying weakly and mildly polar reviews along with the highly polar ones. The Classifier provides an option of modifying the parameters according to need of the system and thus overcomes the problem of static parameters in classical machine learning algorithms.