Product Review Classification from Twitter Data using Semisupervised Learning

Abstract

Twitter and similar microblogging platforms have gained popularity in Latin America, fostering communities discussing various topics, including product experiences. As businesses aim to understand customer preferences, sentiment analysis and topic understanding become crucial. This paper focuses on classifying short-text comments from product reviews extracted from Twitter. Two semisupervised techniques, Label Propagation and a variation of Structural Learning, are employed to improve classification performance with limited labeled data. The challenges posed by short text and noisy data are addressed, and experimental results are presented, showcasing the effectiveness of the proposed methods compared to traditional supervised learning. Future directions for further improvements are also discussed.