Opinion detection: Influence factors

Many online social networks (like blogs or Twitter) allow users to post and share their opinions on various topics. The detection and interpretation of these subjective comments is strategic for various organizational and business purposes, e.g., product and service benchmarking, ads placement or market intelligence. This article aims at enhancing the opinion detection process, i.e., the identification of documents that reflect an opinion, whatever their polarities - positive or negative. Our contribution consists in analyzing the factors that influence the detection of opinions. In particular, we investigate three factors: document´s time, topic, and topic category. We have conducted an experiment to detect opinions in the TREC Blog 06 dataset, using the IMDB data collection as a reference. Our experimental results report that time, topics and topic categories have an impact on the opinion detection process.