A Neutrosophic Approach to Improving Sentiment  Classification Accuracy in Social Media Analytics

Raad A. Qasim; Sajjad abbas; Habeeb Noori Jumaah; Maher Khalaf Hussein; Huda E. Khalid

Authors

Raad A. Qasim University of Telafer /Faculty of Education / Department of Computer Science / Nineveh / Iraq /
Sajjad abbas University of Telafer / University Presidency/ Nineveh / Iraq/
Habeeb Noori Jumaah University of Telafer / University Presidency /Nineveh / Iraq/
Maher Khalaf Hussein University of Telafer / University Presidency /Nineveh / Iraq/
Huda E. Khalid University of Telafer, The Administration Assistant for the President of the Telafer University, Telafer, Iraq;

Keywords:

Neutrosophic logic, Sentiment Analysis, Social Media Analytics, Uncertainty Environment, Natural Language Processing.

Abstract

Traditional sentiment analysis methods often struggle with the inherent ambiguity
and uncertainty present in social media text, where opinions can be simultaneously positive,
negative, and neutral. This paper proposes a novel neutrosophic-based approach to sentiment
classification that addresses the limitations of binary and ternary classification systems. By
incorporating neutrosophic logic's three-valued framework (truth, indeterminacy, and
falsity), our method better captures the nuanced nature of social media sentiment expressions.
Experimental results on multiple social media datasets demonstrate significant improvements
in classification accuracy, with our neutrosophic approach achieving 15% to 20% better
performance in handling ambiguous and mixed-sentiment posts compared to conventional
methods. The proposed framework shows particular effectiveness in processing sarcastic,
ironic that are contextually dependent expressions common in social media platforms.

DOI: 10.5281/zenodo.15825643