Recent Works for Sentiment Analysis using Machine Learning and Lexicon Based Approaches

Sentiment analysis has got attentions for its guidance significance since the influence of online public reviews dramatically increased these years. It aims to perceive people’s attitude from their online reviews about a topic by classifying the words or the document’s polarity into distinct opposite feelings. Given that the massive information and data source online are unstructured and unprocessed where we need to deal with a large amount of information containing slang and expression with distinct language, particular methods and techniques with high accuracy and precision that help us do the processing are needed. In order to assist the future investigators comprehending the influence of online reviews with heterogeneous polarity and the possible improvement of existing works, this paper shows several approaches and applications of SA to analyze and compare the different outcomes and the performance of each. It turns out that existing works conclude that lexicon and machine learning based approach are the two main categories of sentiment analysis generally. The result shows that machine learning based methods can provide an effective and satisfactory assessment of general sentiment classification in processing massive online comments, but manual coding is still dominant in high-precision sentiment analysis. It suggests that when analysis of each comment needs to be performed accurately (especially if the data set is important), a second check through manual coding is necessary.