You are in : Home » Results & submissions » Select UOA » 11 - Computer Science and Informatics » View submission: University of Birmingham » Outputs » Detail

Output details

11 - Computer Science and Informatics

University of Birmingham

Return to search Previous output Next output

Output 153 of 157 in the submission

Output title

Very High Accuracy and Fast Dependency Parsing is not a Contradiction

Type

E - Conference contribution

DOI

Name of conference/published proceedings

International Conference on Computational Linguistics

Volume number

Issue number

First page of article

ISSN of proceedings

Year of publication

2010

URL

http://delivery.acm.org/10.1145/1880000/1873792/p89-bohnet.pdf?ip=147.188.193.240&acc=OPEN&key=782525BEA428F891C47A0316A5533171BE67861B672150770DF987DAA11C9304&CFID=222971427&CFTOKEN=12587301&__acm__=1370523636_c16e7e8fc854608dfc4d5bda5376c0b1

Number of additional authors

Additional information

<22>Dependency parsing is one of the most researched techniques used to obtain syntactic structures in Natural Language Processing. This paper combines a Hash Kernel with Bohnet's syntactic analyzer that ranked first in the CoNLL Shared Task 2009 for English and German (second when averaging across the seven languages; thirteen teams). This innovative parsing approach resulted in 10% error reduction, faster processing times and lower memory requirements. The system is freely available (http://code.google.com/p/mate-tools/), was downloaded >800 times and it is therefore well cited. Further, it formed the basis for a patent that is currently under evaluation together with Sony-Research.

Interdisciplinary

Cross-referral requested

Research group

A - Artificial Intelligence and Intelligent Robotics

Citation count

Proposed double-weighted

Double-weighted statement

Reserve for a double-weighted output

Non-English

English abstract