Gradual Weight Updating for Sentiment Mining

Abstract

Nowadays, many people prefer the use of social media for communicating and exchanging opinions with each other over face to face communication. This has lead to a generation of a tremendous amount of textual opinioned data. Understanding this opinioned data is useful from all perspectives. But the major challenge exists here is how to extract the exact sentiment hidden behind this huge data. To solve this problem, keyword spotting or dictionary-based approaches are followed. In this paper, we present a Gradual Weight Updating for sentiment mining. It not only considers the polarity of each word similar to the unigram methodology but, it also focuses on the entire cluster of words that contains the unigram. The different steps it follows for sentiment extraction of the word are polarity fetching, cluster marking, weight tagging, valence shifter, adversative conjunction handling, and final score generation. The paper contributions in the area of domain independent opinioned word extraction and accurate polarity mining with the help of context marking approach. We used the various opinionated datasets to compare and illustrate the performance of our proposed system.