2014 Sixth International Symposium on Parallel Architectures, Algorithms and Programming (PAAP) (2014)
July 13, 2014 to July 15, 2014
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/PAAP.2014.30
This paper presents a simple and effective approachto improve dependency parsing by exploiting multiplefeature-sets. Traditionally, features are extracted by applyingthe feature templates to all the word pairs(first-order features)and word tuples(second-order features). In this pa per, we showthat exploiting different feature templates for different wordpairs and word tuples achieves significant improvement overbaseline parsers. First, we train a text chunker using a freelyavailable implementation of the first-order linear conditionalrandom fields model. Then we build a clause-chunk tree for agiven sentence based on chunking information and punctuationmarks. Finally, we extract features for dependency parsingaccording to multiple feature-sets. We extend the projectiveparsing algorithms of McDonald and Carreras for ourcase, experimental results show that our approach significantlyoutperform the baseline systems without increasing complexity. Given correct chunking information, we improve from baselineaccuracies of 91.36% and 92.20% to 93.19% and 93.89%, respectively.
Feature extraction, Heuristic algorithms, Accuracy, Inference algorithms, Approximation algorithms, Training, Tagging
Xianchao Zhang, Dong Du, Xinyue liu, Wenxin Liang, "Multiple Feature-Sets Method for Dependency Parsing", 2014 Sixth International Symposium on Parallel Architectures, Algorithms and Programming (PAAP), vol. 00, no. , pp. 57-62, 2014, doi:10.1109/PAAP.2014.30