Sentiment Analysis of Machine Learning Algorithms: A Transformer-Based Approach

Saad Iqbalbhai Mandli

Vol. 11, Issue 1, Jan-Dec 2025

Page Number: 76 - 81

Abstract:

Sentiment analysis is one of the most widely used applications of Natural Language Processing because it helps convert unstructured opinions into measurable insights. Traditional machine learning algorithms such as Naive Bayes, Logistic Regression, Support Vector Machine, Random Forest and Gradient Boosting have been used extensively for classifying reviews, tweets, feedback and news articles into positive, negative or neutral categories. However, these methods depend heavily on manual feature engineering, bag-of-words, TF-IDF vectors and domain-specific pre-processing. Deep learning methods improved performance by learning distributed representations, but recurrent and convolutional architectures still struggle with long-range context, sarcasm, negation and aspect-level sentiment. Transformer-based models, especially BERT and its variants, address many of these limitations through self-attention, contextual embeddings and transfer learning. This paper studies sentiment analysis from the perspective of machine learning algorithms and proposes a transformer-based approach for improving classification accuracy, contextual understanding and adaptability across domains. A comparative analysis is presented between traditional machine learning, deep learning and transformer-based models. The paper concludes that transformer-based models provide superior contextual representation and generalization, although they require higher computational resources and careful fine-tuning.

References

Back Download