Semi-supervised sentiment clustering on natural language texts

Abstract In this paper, we propose a semi-supervised method to cluster unstructured textual data called sentiment clustering on natural language texts. The aim is identify clusters homogeneous with respect the overall of texts analyzed. combines different techniques and methodologies: Sentiment Analysis, Threshold-based Naïve Bayes classifier, Network-based Semi-supervised Clustering. It involves steps. first step, text transformed into structured text, it categorized positive or negative classes using analysis algorithm. second classifier applied define specific value for topics. last Clustering partition instances disjoint groups. proposed algorithm tested collection reviews written by customers Booking.com . results have highlighted capacity that are distinct, non-overlapped, sentiment. Results also easily interpretable thanks network representation helps understand relationship between them.