A Novel Morphological Rule-based Approach for Urdu Text Sentiment Analysis

Abstract

Sentiment Analysis (SA) is an exciting research area in Urdu text classification tasks. Due to its complicated morphology, sentiment analysis in Urdu poses various challenges. The lexicon-based technique is the preferred approach for Urdu sentiment analysis for its high accuracy rate. This approach requires a comprehensive manual list of positive and negative words. Sentiment classification can be effectively achieved using morphology-driven methods. For this purpose, an effective Urdu sentiment analyser employing morphological rules for performing Urdu sentiment analysis is proposed. The application of the rules depends on the input words' lexical category and prefixes. Sentiment classification rules are proposed based on the prefixes attached to negative sentiment carrier words.  A thorough evaluation with reasonable Urdu text document/dataset size is performed. The empirical assessment proved that the proposed rule-based method achieved satisfactory precision, recall, and accuracy.