A Systematic Analysis of Deep Learning Based Twitter Sentiment Analysis: Emerging Trends and Challenges

Sentiment analysis is a powerful area of research in the realm of natural-language-processing(NLP), with several practical applications. Businesses and governments can gain valuable insights from the immense amounts of user-generated data available through diverse social media platforms. However, it is now challenging to efficiently mine pertinent information from the available data. Readers, product vendors, and legislators all pay close attention to the opinions offered by internet users. Sentiment-analysis(SA) has received a lot of interest as a means of analyzing and organizing the unstructured data that comes through social media. Sentiment analysis is a method of text organization used to categorize the author's attitude or feelings as positive, negative, neutral, or otherwise. Deep learning models are useful because of their automated learning potential, this problem has been solved by combining sentiment analysis with deep-learning(DL). In this study, we undertake a thorough literature evaluation of DL strategies to document-based sentiment analysis, with the goal of extracting various features from social media content, namely from Twitter data. In addition, it offers a concise overview, evaluation, and improvement of deep learning's latest breakthroughs in the areas of sentiment analysis approaches and document applications. This paper investigates the difficulties that researchers have run into and other issues that may arise when attempting to analyze social media sentiment. It explains why and how SA is performed, what problems it solves, and what kinds of problems it solves. It also identifies significant difficulties with the datasets, text languages, investigation methods, and assessment measures, and compares the results of several investigations. This study aids the state of the art in SA and can guide practitioners toward the most useful approach.