Summary of the paper
| Title | Parsing to Stanford Dependencies: Trade-offs between Speed and Accuracy |
|---|---|
| Authors | Daniel Cer, Marie-Catherine de Marneffe, Dan Jurafsky and Chris Manning |
| Abstract | We investigate a number of approaches to generating Stanford Dependencies, awidely used semantically-oriented dependency representation.We examine algorithms specifically designed for dependency parsing (Nivre,Nivre Eager, Covington, Eisner, and RelEx) as wellas dependencies extracted from constituent parse trees created by phrasestructure parsers (Charniak, Charniak-Johnson, Bikel, Berkeleyand Stanford). We found that constituent parsers systematically outperformalgorithms designed specifically for dependency parsing.The most accurate method for generating dependencies is the Charniak-Johnsonreranking parser, with 89% (labeled) attachment F1score. The fastest methods are Nivre, Nivre Eager, and Covington, used with alinear classifier to make local parsing decisions, whichcan parse the entire Penn Treebank development set (section 22) in less than 10seconds on an Intel Xeon E5520. However, this speedcomes with a substantial drop in F1 score (about 76% for labeled attachment)compared to competing methods. By tuning how much ofthe search space is explored by the Charniak-Johnson parser, we are able toarrive at a balanced configuration that is both fast and nearlyas good as the most accurate approaches. |
| Language | Grammar and Syntax |
| Topics | Parsing, Tools, systems, applications, Grammar and Syntax |
Full paper  |
Parsing to Stanford Dependencies: Trade-offs between Speed and Accuracy |
| Bibtex | @InProceedings{CER10.730, author = {Daniel Cer, Marie-Catherine de Marneffe, Dan Jurafsky and Chris Manning}, title = {Parsing to Stanford Dependencies: Trade-offs between Speed and Accuracy}, booktitle = {Proceedings of the Seventh conference on International Language Resources and Evaluation (LREC'10)}, year = {2010}, month = {may}, date = {19-21}, address = {Valletta, Malta}, editor = {Nicoletta Calzolari (Conference Chair), Khalid Choukri, Bente Maegaard, Joseph Mariani, Jan Odjik, Stelios Piperidis, Mike Rosner, Daniel Tapias}, publisher = {European Language Resources Association (ELRA)}, isbn = {2-9517408-6-7}, language = {english} } |