Systematic Review of Deep Learning Models in Sentiment Analysis ( 2020-2025)
TL;DR
The study concluded that advances in deep learning models have significantly improved the accuracy of sentiment analysis, and recommends developing models capable of more effectively handling linguistic complexities, such as sarcasm, irony, regional dialects, and multiple dialects.
Abstract
Sentiment analysis is one of the most important branches of natural language processing. Sentiment analysis has become a key research area in natural language processing, driven by rapid advancements in deep learning architectures. Where It is used to extract opinions and emotional sentiments from texts about a specific topic or new product, as well as from political and financial news headlines. In recent years, recurrent neural network models, convolutional neural networks, and, notably, transformer-based architectures have significantly improved performance across a wide range of natural language processing tasks. This paper presents a systematic review of deep learning approaches used in sentiment analysis published between 2020 and 2025, following established systematic review guidelines, including a specific search strategy . After applying inclusion and exclusion criteria, a number of final studies were selected for analysis and comparison in terms of methodologies, accuracy, and data used.The study concluded that advances in deep learning models have significantly improved the accuracy of sentiment analysis. The review also recommends developing models capable of more effectively handling linguistic complexities, such as sarcasm, irony, regional dialects, and multiple dialects.