The Impact of Feature Extraction on Multi-Source Sentiment Analysis

Abstract

The rapid growth of internet users combined with the increasing dominance of online review sites and social media platforms, have given rise to the importance of sentiment analysis, also known as opinion mining, seeks to determine what other people believe and comment. Almost every enthusiastic or person who loves social platforms likely to articulate their ideas in the shape of comments on various social media platforms, and this is viewed as the main resource of sentiment analysis. These comments not only communicate people’s feelings, but also provide insight into their moods. Because the text on these media is unstructured, we must first preprocess it, employing six different preprocessing approaches, before extracting features from the pre-processed data. Some of the examples of feature extraction techniques are TF-IDF, word embedding, Bag of Words and word count, noun count are feature based natural language processing. Apart from the work that has already been done in text analytics, feature extraction in sentiment analysis is presently a hot topic of research. The impact of existing methodologies and approaches for feature extraction in sentiment analysis on the performance of various sentiment classification algorithms is discussed in this review study.