Mr. Vaibhav Gajananrao Ingole
Download PDFSentiment analysis, or opinion mining, focuses on detecting, extracting, and categorizing subjective information within textual data. This includes opinions, attitudes, and emotions. One of the main drivers of the surge in sentiment analysis over the last two decades has been the emergence of social media, review websites, blogs, and online discussion forums. Each of these contains an abundance of data composed of interpersonal opinions and have created a significant need for automated systems to analyze sentiments. Initially, sentiment analysis focused on lexicon-based and classical machine learning methods. Recently, there has been a shift toward representations and architectures based on deep learning. It examines core sentiment analysis tasks at the document, sentence, and aspect level, as well as analysis of sentiments within social media. It reviews lexicon-based methods, along with sentiment analysis using Naïve Bayes, Support Vector Machines, and Maximum Entropy, as well as recursive and convolutional neural networks and distributed document representations. It analyses challenges including negation, sarcasm, ambiguity, domain adaptation, multilingual text, code-mixed languages, noisy social media data, and a lack of annotated data. It also identifies areas for future research focusing on code-mixed and multilingual sentiment analysis, where current methods are largely inadequate.
