Extracting tree fragments in linear average time

Abstract

This report details the implementation of a fragment extraction algorithm using an average case linear time tree kernel. Given a treebank, the algorithm extracts all fragments that occur at least twice, along with their frequency. Evaluation shows a -fold speedup over a quadratic fragment extraction implementation. Additionally, we add support for trees with discontinuous constituents.

Cite this paper

@inproceedings{Cranenburgh2012ExtractingTF, title={Extracting tree fragments in linear average time}, author={Andreas van Cranenburgh}, year={2012} }