Enhanced Lexicon based Hybrid Method for Slang and Punctuation Scoring for Aspect Based Sentiment Analysis

In the contemporary landscape of sentiment analysis from customer reviews on product and service evaluations, the focus has shifted markedly towards the nuanced evaluation of their finer sentiment. This underscored the need to deciphering and understanding customer sentiments with aspects towards Aspect-Based Sentiment Analysis (ABSA). In earlier literature informal words like slang and informal sentiment carrier like punctuation were mostly trimmed or left out through data preprocessing. With the time both researcher and business are demanding to include more granular details like slang and punctuation for enhancing the accurate sentiment for Aspect sentiment classification. Evaluation of slang and punctuation, in particular, poses a substantial hurdle as sentiment of a sentence is influenced by the interplay between word semantics and punctuation, and it needs to use lexicon thus demanding innovative solutions. In addition, some literature do not handle the negations and intensifiers in a sentence; instead, it gives only overall positive and negative sentiment polarity. This paper introduces an enhanced lexicon-based hybrid method for the classification using slang and punctuations score for ABSA. Experiments conducted on benchmark SemEval datasets serve as a robust validation of the efficacy of the proposed model. Hybrid method with metrics experienced substantial improvements, with the yielding impressive results for classification with a precision rate of 93.77%, a recall rate of 92.56%, and an overall accuracy rate of 94.82%. These outcomes further solidify the efficacy of the proposed hybrid method with two Amazon datasets.