Concerning different approaches to automatic PoS tagging: EngCG-2, a constraint-based morphological tagger, is compared in a double-blind test with a state-of-the-art statistical tagger on a common disambiguation task using a common tag set. The experiments show that for the same amount of remaining ambiguity, the error rate of the statistical tagger is one order of magnitude greater than that of the rule-based one. The two related issues of priming effects compromising the results and disagreement between human annotators are also addressed.
Description
Fantastic. A double-blind experiment to test the accuracy of a tagger believed too good to be true.
%0 Conference Paper
%1 samuelsson1997cls
%A Samuelsson, Christer
%A Voutilainen, Atro
%B Proceedings of the 8th conference of the EACL
%C Morristown, NJ, USA
%D 1997
%I ACL
%K CG DASP303 HMM POS-tagging apertium pri statistical toread trigram
%P 246--253
%R http://dx.doi.org/10.3115/979617.979649
%T Comparing a Linguistic and a Stochastic Tagger
%U http://portal.acm.org/citation.cfm?id=979649
%X Concerning different approaches to automatic PoS tagging: EngCG-2, a constraint-based morphological tagger, is compared in a double-blind test with a state-of-the-art statistical tagger on a common disambiguation task using a common tag set. The experiments show that for the same amount of remaining ambiguity, the error rate of the statistical tagger is one order of magnitude greater than that of the rule-based one. The two related issues of priming effects compromising the results and disagreement between human annotators are also addressed.
@inproceedings{samuelsson1997cls,
abstract = {Concerning different approaches to automatic PoS tagging: EngCG-2, a constraint-based morphological tagger, is compared in a double-blind test with a state-of-the-art statistical tagger on a common disambiguation task using a common tag set. The experiments show that for the same amount of remaining ambiguity, the error rate of the statistical tagger is one order of magnitude greater than that of the rule-based one. The two related issues of priming effects compromising the results and disagreement between human annotators are also addressed.},
added-at = {2009-10-23T21:21:48.000+0200},
address = {Morristown, NJ, USA},
author = {Samuelsson, Christer and Voutilainen, Atro},
biburl = {https://www.bibsonomy.org/bibtex/290e598c3d63351a2526164a09dc5a94a/unhammer},
booktitle = {Proceedings of the 8th conference of the EACL},
description = {Fantastic. A double-blind experiment to test the accuracy of a tagger believed too good to be true.},
doi = {http://dx.doi.org/10.3115/979617.979649},
interhash = {6ad955c358049393f79ab671f408f5d7},
intrahash = {90e598c3d63351a2526164a09dc5a94a},
keywords = {CG DASP303 HMM POS-tagging apertium pri statistical toread trigram},
location = {Madrid, Spain},
pages = {246--253},
publisher = {ACL},
timestamp = {2009-10-23T21:21:48.000+0200},
title = {{Comparing a Linguistic and a Stochastic Tagger}},
url = {http://portal.acm.org/citation.cfm?id=979649},
year = 1997
}