Abstract
The possible combinations of features traditionally used for sentence similarity amount to a very large feature space. Considering all possible combinations and training a support vector machine on the resulting meta-features in a two step process significantly improves performance. The proposed method is trained and tested on the SemEval-2012 Semantic Textual Similarity (STS) Shared Task data, outperforming the task's highest ranking system. © 2013 Springer-Verlag.
Author supplied keywords
Cite
CITATION STYLE
Shareghi, E., & Bergler, S. (2013). Feature combination for sentence similarity. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 7884 LNAI, pp. 150–161). https://doi.org/10.1007/978-3-642-38457-8_13
Register to see more suggestions
Mendeley helps you to discover research relevant for your work.
