Optimizing Sentiment Analysis: A Cognitive Approach with Negation Handling via Mathematical Modelling

Negation handling is a crucial aspect of sentiment analysis, as it presents challenges to accurate sentiment classification by altering polarity and reducing reliability. Traditional lexicon-based approaches often lack adequate techniques for modeling negation and fail to identify the appropriate negation window. Moreover, building machine learning models for negation handling in conversational text data proves difficult due to the intricate syntactic structure of negation. To address these issues, we propose a novel unsupervised cognitive sentiment classification approach. Our research introduces the multi-criteria decision-making (MCDM)–based “Negation Handling of the Text Using the VIKOR Optimization Technique” (NEGVOT) model, which effectively handles negation in sentiment analysis. By employing the decision science method, the NEGVOT model provides a solution for correctly labeling text sentiment in both negation-free and negation-containing texts. Our approach utilizes a lexicon database to obtain context scores of textual comments and integrates emotional scores to achieve accurate sentiment classification. Through experiments conducted on three benchmarked datasets, we demonstrate that the NEGVOT model performs comparably to state-of-the-art models. The NEGVOT model achieves the accuracy of 83%, 85%, and 82% over three datasets. It significantly enhances the accuracy of sentiment orientation tagging by effectively handling sentences with negation. We employ statistical analysis to support the relevance of our findings. The NEGVOT paradigm ensures logical and consistent results while exhibiting a strong generalization capacity, enabling sentiment classification for texts containing negations. This study makes notable contributions to the advancement of unsupervised techniques and provides a robust framework for handling negation in sentiment classification tasks.