Nowadays, opinions are a ubiquitous part of the Web and sharing experiences has never been more popular. Information regarding consumer opinions is valuable for consumers and producers alike, aiding in their respective decision processes. Due to the size and heterogeneity of this type of information, computer algorithms are employed to gain the required insight. Current research, however, tends to forgo a rigorous analysis of the used features, only going so far as to analyze complete feature sets. In this paper we analyze which features are good predictors for aspect-level sentiment using Information Gain and why this is the case. We also present an extensive set of features and show that it is possible to use only a small fraction of the features at just a minor cost to accuracy.

, , , , ,
doi.org/10.1007/978-3-319-41754-7_5, hdl.handle.net/1765/96126
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Erasmus University Rotterdam

Schouten, K., Frasincar, F., & Dekker, R. (2016). An information gain-driven feature study for aspect-based sentiment analysis. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). doi:10.1007/978-3-319-41754-7_5