Loading article...
Loading article...
Stream/Area:
-
Subject:
-
Paper ID:
IJSREM64842
Volume:
10 |
6 |
Jun 2026
Pages:
1-3
Publication Frequency:
Monthly (12 issues per year)
License:International Journal of Scientific Research in Engineering and Management follows the Creative Commons Attribution 4.0 (CC BY 4.0) License, allowing free use and distribution with proper attribution.
Computer Science and Engineering
ABES Engineering College, Ghaziabad
Computer Science and Engineering
ABES Engineering College, Ghaziabad
Sentiment analysis has become one of the most important applications of Natural Language Processing because social media generates a continuous stream of short, noisy, and opinion-rich text. This paper presents SentiLab, a comparative framework for three-class Twitter sentiment classification using Positive, Neutral, and Negative labels. The study evaluates three representative model families: TF-IDF with Logistic Regression, Bidirectional LSTM with Multi-Head Self-Attention, and fine-tuned DistilBERT. The original Sentiment140 dataset was extended into a balanced three-class dataset using confidence-based relabeling so that emotionally ambiguous tweets could be modeled as Neutral rather than being forced into binary polarity. The results show that DistilBERT achieved the best overall performance with 94.7% accuracy and a macro AUC-ROC of 0.988, followed by BiLSTM+Attention at 92.4% and TF-IDF+Logistic Regression at 76.3%. In addition to model comparison, the framework was deployed as a browser-based real-time sentiment analysis system using ONNX Runtime and Transformers.js. The paper also discusses the trade-off between classification accuracy, inference latency, and deployability in practical sentiment systems.
KEYWORDS
The study shows that DistilBERT achieves the strongest overall performance for three-class Twitter sentiment analysis, BiLSTM+Attention offers a strong middle ground, and TF IDF+Logistic Regression remains a useful low-cost baseline. Future work may include multilingual sentiment analysis, aspect-based sentiment classification, quantized transformer deployment, and streaming analysis of live social media data.
[1] A. Go, R. Bhayani, and L. Huang, 'Twitter Sentiment Classification using Distant Supervision,' Stanford University,2009. [2] J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, 'BERT: Pre-training of Deep Bidirectional Transformers forLanguage Understanding,' Proc. NAACL-HLT, 2019. [3] V. Sanh et al., 'DistilBERT, a Distilled Version of BERT: Smaller, Faster, Cheaper and Lighter,' arXiv preprintarXiv:1910.01108, 2019. [4] A. Vaswani et al., 'Attention Is All You Need,' Proc. NIPS, 2017. [5] Y. Kim, 'Convolutional Neural Networks for Sentence Classification,' Proc. EMNLP, 2014. [6] S. Hochreiter and J. Schmidhuber, 'Long Short-Term Memory,' Neural Computation, vol. 9, no. 8, pp. 1735-1780, 1997. [7] D. Bahdanau, K. Cho, and Y. Bengio, 'Neural Machine Translation by Jointly Learning to Align and Translate,' arXiv preprint arXiv:1409.0473, 2015. [8] J. Pennington, R. Socher, and C. Manning, 'GloVe: Global Vectors for Word Representation,' Proc. EMNLP, 2014. [9] C. Hutto and E. Gilbert, 'VADER: A Parsimonious Rule-Based Model for Sentiment Analysis of Social Media Text,' Proc. ICWSM, 2014. [10] M. Hu and B. Liu, 'Mining and Summarizing Customer Reviews,' Proc. ACM SIGKDD, 2004.
Publication Date
6/11/2026
Peer Review Type
Double-Blind Peer Review
Review Rounds: 2
Documents
Article Views
0
Total views since publication
PDF Downloads
3
Total downloads since publication
Geographic Reach
5
Digital Preservation: This article is permanently archived on Zenodo and complies with all data preservation standards across major indexing platforms.
Plagiarism Screening
This manuscript has been checked for plagiarism using TurnitinPlagiarismChecker software.
Conflict of Interest
The authors declare no conflict of interest related to this research work.
Funding Information
No funding information available for this research.
Related research in the same subject area
Loading similar articles...