Abstract | ||
---|---|---|
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. |
Year | DOI | Venue |
---|---|---|
2016 | 10.1007/978-3-319-41754-7_5 | Lecture Notes in Computer Science |
Keywords | Field | DocType |
Sentiment analysis,Aspect-level sentiment analysis,Data mining,Feature analysis,Feature selection,Information gain | Data mining,Feature selection,Sentiment analysis,Computer science,Information gain,Decision process,Pattern recognition (psychology) | Conference |
Volume | ISSN | Citations |
9612 | 0302-9743 | 0 |
PageRank | References | Authors |
0.34 | 8 | 3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Kim Schouten | 1 | 161 | 15.79 |
Flavius Frasincar | 2 | 1367 | 117.14 |
Rommert Dekker | 3 | 763 | 101.60 |