Sentence Level Feature Selection for Production Prescriptive Opinion Mining

Abstract

The rise of online reviews and social media has fueled the need for advanced opinion mining techniques that can extract and analyze subjective information from text data. This research introduces a novel framework for Sentence Level Feature Selection for Production Prescriptive Opinion Mining (SLFSPPOM), aimed at enhancing sentiment analysis accuracy in production and service domains. The framework includes three key modules to extract actionable insights from online reviews and social media data. The Context Graspable Aspect-based Sentiment Analysis centered Sentence-level Opinion Extractor (CGASA-SOE) performs fine-grained, context-aware aspect-based sentiment analysis, capturing specific sentiments toward product or service features. The Part-of-Speech Tagging Feature Representation Module (POST-FRM) refines feature extraction using linguistic patterns to improve sentiment and aspect recognition. The Fuzzy Rule-based Prescription Generator (FRPG) converts extracted data into tailored production-centric recommendations. By combining these modules, SLFSPPOM provides a comprehensive solution that translates detailed opinions into actionable insights for decision-makers. This framework offers industries a robust tool for gaining nuanced insights and optimizing customer satisfaction and operational strategies. One of the well-known benchmark datasets from Amazon reviews is used to evaluate the performance in terms of Accuracy, Precision, Sensitivity, Specificity, and F-Score of the proposed work. A vivid improvement in performance is observed through the experiments carried out.