Enhancing the Performance of Sentiment Analysis using Enhanced Preprocessing Technique with Agriculture Review Score Calculation

The discipline of natural language processing is now placing the highest level of research focus on sentiment analysis. Nevertheless, despite its significant capabilities, the use of this technology remains limited in the agricultural industry. The objective of this research is to analyse the online review characteristics, review length, and review sentiment score across different food products. Furthermore, the analysis examines the differences in customer sentiment ratings based on the length of the reviews. This study proposes an enhanced preprocess method, which can adjust the selection and weighting of features or attributes of the parsed tree structures dynamically, based on the specific context or target sentiment being analyzed. This technique aims to improve the accuracy and relevance of sentiment analysis by tailoring the feature selection process to better capture the sentiment expression in different contexts. Furthermore, it uses the Valence Aware Dictionary and Sentiment Reasoner (VADER) lexicon-based categorisation approach to assess consumer sentiment across many domains. The approach entails creating a specialised vocabulary for the agricultural domain using the VADER lexicon. The reviews are then classified using this dictionary. The variations in sentiment ratings across the different farm evaluations provide a thorough feedback to the companies. Additionally, it conveys a brand’s assessment by customers and provides performance feedback for expanding into other areas. By evaluating various reviews, the VADER Lexicon classifier with Treebank dynamic filtering (VADER_TDF) method achieves $\mathbf{9 8 \%}$ of accuracy, $\mathbf{9 3 \%}$ of precision, 91% of recall and 90% of F1-score.