Abstract | ||
---|---|---|
In this paper, we provide a theoretical framework for feature selection in tree kernel spaces based on gradient-vector components of kernel-based machines. We show that a huge number of features can be discarded without a significant decrease in accuracy. Our selection algorithm is as accurate as and much more efficient than those proposed in previous work. Comparative experiments on three interesting and very diverse classification tasks, i.e. Question Classification, Relation Extraction and Semantic Role Labeling, support our theoretical findings and demonstrate the algorithm performance. |
Year | Venue | Keywords |
---|---|---|
2010 | CoNLL | algorithm performance,diverse classification task,theoretical framework,gradient-vector component,theoretical finding,relation extraction,comparative experiment,feature selection,selection algorithm,syntactic tree kernel,semantic role,reverse feature engineering |
Field | DocType | Citations |
Kernel (linear algebra),Pattern recognition,Feature selection,Computer science,Selection algorithm,Tree kernel,Feature engineering,Artificial intelligence,Syntax,Semantic role labeling,Machine learning,Relationship extraction | Conference | 11 |
PageRank | References | Authors |
0.55 | 33 | 2 |
Name | Order | Citations | PageRank |
---|---|---|---|
daniele pighin | 1 | 289 | 18.72 |
Alessandro Moschitti | 2 | 3262 | 177.68 |