A Comprehensive Evaluation of Machine Learning and Transformer-Based Techniques for Sentiment Analysis

Haoran Xi

doi:10.54097/008hs226

Authors

Haoran Xi

DOI:

https://doi.org/10.54097/008hs226

Keywords:

sentiment analysis, deep learning, pretrained language models, traditional machine learning, natural language processing.

Abstract

Emotions, as an essential part of text, play a significant role in opinion mining, enabling the judgment of users’ attitudes and the monitoring of public opinion. This study compares statistical learning methods, deep neural networks, and pre-trained models for sentiment polarity classification tasks using the Internet Movie Database (IMDB) movie review dataset, the ChnSentiCorp Chinese sentiment dataset, and financial data. Empirical evidence shows that traditional approaches, such as support vector machines—which go beyond classic algorithms—still demonstrate strong performance in classifying data. However, deep learning models based on bidirectional long short-term memory networks achieve considerable improvements in accuracy due to their ability to capture contextual information. More impressively, the Internet Movie Database (BERT) model achieves 100% on all evaluation metrics after knowledge distillation, further confirming the advantages of pre-training techniques for these datasets. These results provide insights into how different models evolve and highlight the need to balance model accuracy, computational cost, and data requirements in practical applications.

References

[1] L. Zhang, S. Wang, & B. Liu. Deep learning for sentiment analysis: A survey. Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery, 8(4), e1253. (2018).

[2] J. Devlin, M. W. Chang, K. Lee, & K. Toutanova. BERT: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics (NAACL). (2019).

[3] C. Sun, L. Huang, & X. Qiu. Utilizing BERT for aspect-based sentiment analysis via constructing auxiliary sentence. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics (NAACL). (2019).

[4] H. Xu, & K. Mao. A survey of aspect-level sentiment analysis based on deep learning. IEEE Access, 9, 71857–71870. (2021).

[5] P. Liu, W. Li, & Y. Zou. Empowering BERT with knowledge graph embedding for document classification. In Proceedings of the 28th International Joint Conference on Artificial Intelligence (IJCAI). (2019).

[6] Q. Zhong, et al. Knowledge-enriched transformer for emotion detection in conversations. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP). (2020).

[7] Y. Zhang, & Q. Yang. A comprehensive review on cross-domain sentiment analysis. ACM Computing Surveys. (2022).

[8] Z. Lin, et al. A unified model for sentiment analysis with multi-task learning. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics (ACL). (2020).

[9] P. Zhou, et al. Attention-based bidirectional long short-term memory networks for relation classification. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics (ACL). (2019).

[10] X. Huang, et al. Prompt-based fine-tuning for sentiment classification in few-shot settings. In Proceedings of the 29th International Conference on Computational Linguistics (COLING). (2023).