Abstract

Sentiment analysis is the computational analysis of end users' attitudes, opinions, and feelings around a specific subject or product. The process of gathering and examining people's thoughts, feelings, attitudes, views, etc. about various subjects, goods, and services is known as sentiment analysis. Sentiment analysis categorizes the message as positive, negative, or neutral based on its polarity. Researchers have recently concentrated on sentiment analysis of social media posts using lexical and machine learning methods. Social media is a type of microblogging platform where users can leave comments in slang, which is made up of idioms, misspellings, symbols, and ironic statements. Social media data also suffer from the "curse of dimension," or the high dimensional character of the data necessitating certain feature extraction and pre-processing steps to increase classification accuracy. To provide scholars with an international survey on sentiment analysis and associated disciplines, this paper includes a comparative assessment of sentiment analysis methodologies followed by a comparison of their accuracy. This paper presents a thorough introduction to sentiment analysis based on current research, and it then examines feature extraction and machine learning techniques for sentiment analysis using various data sets. Additionally, this article investigates the use of several machine learning, deep learning, and lexicon investigation approaches. The purpose of this study is to improve the understanding of sentiment analysis so that academics and practitioners can choose appropriate techniques for sentiment categorization based on the kind of data being studied.