Combining rule-based and bag-of-words for phase-level sentiment analysis of blog comments

Abstract

Blogs are one of the platforms that express personal opinions, which are intended to create awareness and used as an instrument to establish trust among customers products and services or about a specific topic. A new classifying model was experimented to improve the sentiment classification of the blog comments. This technique combined Rule-Based (RB) and Bag-of-Word (BoW) model to solve major weaknesses of this Bow model in the conduct of Sentiment Analysis (SA) evaluations. The proposed technique was experimented to esti-mate the Philippine Internet customers’ satisfaction related to the quality of the services provided by the ISPs in the Philippines. In addition, automatic word seeding, building of sentiment dictionary utilizing an online dictionary, n-gram, tokenization, stemming and other SA tech-niques were applied to extract useful information from the blog comment dataset. The results of the research showed that the configurations involving BoW-RB + SVM + bi-gram + Porter stemmer achieved a high classification accuracy of 88%.Capturing the contextual meaning of the neighboring words of a given sentimental word provides a significant help to increase the classifi-cation performance of the proposed classifier method.