Enhanced Approach of Using Custom Heuristic Rules with Stanford and Naive Bayes Classification Technique
Tanvi Kadam, Parminder Kaur · 2018 International Conference On Advances in Communication and Computing Technology (ICACCT) · 2018
Expressive nature of user provides opinions, feedbacks, ideas or suggestions for some entities on the internet and hence there is a constant growth in the amount of knowledge provided for those entities. To get to know about the entities on the internet site, user will go through reviews of the other users provided for entities. The systems these days cannot only depend on user's star rating, so the opinions have to be analyzed using comments and natural language processing. Opinion mining works for classifying the reviews based on polarity. The basic problem of opinion mining is identifying the required aspect in order to find the exact opinion. In this paper, reviews are collected for smart phone and hotel entities. Heuristic rules for noun and adjective phrases from the reviews are created using regular expression on parts-of-speech (POS) tags. Enhanced heuristic rules are implemented with Stanford natural language processing (SNLP) classification and are compared with Naive Bayes (NB) classification. The results proved the high accuracy of enhanced Heuristic rules approaches using SNLP classifier over NB for the similar entities.