A general feature space for automatic verb classification. Joanis, E., Stevenson, S., & James, D. Natural Language Engineering, 14(3):337–367, 2008. Also published by Cambridge Journals Online on December 19, 2006abstract bibtex Lexical semantic classes of verbs play an important role in structuring complex predicate information in a lexicon, thereby avoiding redundancy and enabling generalizations across semantically similar verbs with respect to their usage. Such classes, however, require many person-years of expert effort to create manually, and methods are needed for automatically assigning verbs to appropriate classes. In this work, we develop and evaluate a feature space to support the automatic assignment of verbs into a well-known lexical semantic classification that is frequently used in natural language processing. The feature space is general – applicable to any class distinctions within the target classification; broad – tapping into a variety of semantic features of the classes; and inexpensive – requiring no more than a POS tagger and chunker. We perform experiments using support vector machines (SVMs) with the proposed feature space, demonstrating a reduction in error rate ranging from 48% to 88% over a chance baseline accuracy, across classification tasks of varying difficulty. In particular, we attain performance comparable to or better than that of feature sets manually selected for the particular tasks. Our results show that the approach is generally applicable, and reduces the need for resource-intensive linguistic analysis for each new classification task. We also perform a wide range of experiments to determine the most informative features in the feature space, finding that simple, easily extractable features suffice for good verb classification performance.
@Article{ joanis1,
author = {Eric Joanis and Suzanne Stevenson and David James},
title = {A general feature space for automatic verb classification},
journal = {Natural Language Engineering},
volume = {14},
number = {3},
pages = {337--367},
year = {2008},
note = {Also published by <a
href=http://journals.cambridge.org/action/displayIssue?jid=NLE&volumeId=14&issueId=03>Cambridge
Journals Online</A> on December 19, 2006},
abstract = {Lexical semantic classes of verbs play an important role
in structuring complex predicate information in a lexicon,
thereby avoiding redundancy and enabling generalizations
across semantically similar verbs with respect to their
usage. Such classes, however, require many person-years of
expert effort to create manually, and methods are needed
for automatically assigning verbs to appropriate classes.
In this work, we develop and evaluate a feature space to
support the automatic assignment of verbs into a well-known
lexical semantic classification that is frequently used in
natural language processing. The feature space is
<em>general</em> – applicable to any class
distinctions within the target classification;
<em>broad</em> – tapping into a variety of semantic
features of the classes; and <em>inexpensive</em> –
requiring no more than a POS tagger and chunker. We perform
experiments using support vector machines (SVMs) with the
proposed feature space, demonstrating a reduction in error
rate ranging from 48% to 88% over a chance baseline
accuracy, across classification tasks of varying
difficulty. In particular, we attain performance comparable
to or better than that of feature sets manually selected
for the particular tasks. Our results show that the
approach is generally applicable, and reduces the need for
resource-intensive linguistic analysis for each new
classification task. We also perform a wide range of
experiments to determine the most informative features in
the feature space, finding that simple, easily extractable
features suffice for good verb classification performance.},
download = {http://www.cs.toronto.edu/~joanis/Papers/JoanisStevensonJames-nle13_3-2007.pdf}
}
Downloads: 0
{"_id":{"_str":"534282740e946d920a001b8a"},"__v":5,"authorIDs":["5457f20a2abc8e9f370008f4","54643dbb4f4818cf4f00057d"],"author_short":["Joanis, E.","Stevenson, S.","James, D."],"bibbaseid":"joanis-stevenson-james-ageneralfeaturespaceforautomaticverbclassification-2008","bibdata":{"bibtype":"article","type":"article","author":[{"firstnames":["Eric"],"propositions":[],"lastnames":["Joanis"],"suffixes":[]},{"firstnames":["Suzanne"],"propositions":[],"lastnames":["Stevenson"],"suffixes":[]},{"firstnames":["David"],"propositions":[],"lastnames":["James"],"suffixes":[]}],"title":"A general feature space for automatic verb classification","journal":"Natural Language Engineering","volume":"14","number":"3","pages":"337–367","year":"2008","note":"Also published by <a href=http://journals.cambridge.org/action/displayIssue?jid=NLE&volumeId=14&issueId=03>Cambridge Journals Online</A> on December 19, 2006","abstract":"Lexical semantic classes of verbs play an important role in structuring complex predicate information in a lexicon, thereby avoiding redundancy and enabling generalizations across semantically similar verbs with respect to their usage. Such classes, however, require many person-years of expert effort to create manually, and methods are needed for automatically assigning verbs to appropriate classes. In this work, we develop and evaluate a feature space to support the automatic assignment of verbs into a well-known lexical semantic classification that is frequently used in natural language processing. The feature space is <em>general</em> – applicable to any class distinctions within the target classification; <em>broad</em> – tapping into a variety of semantic features of the classes; and <em>inexpensive</em> – requiring no more than a POS tagger and chunker. We perform experiments using support vector machines (SVMs) with the proposed feature space, demonstrating a reduction in error rate ranging from 48% to 88% over a chance baseline accuracy, across classification tasks of varying difficulty. In particular, we attain performance comparable to or better than that of feature sets manually selected for the particular tasks. Our results show that the approach is generally applicable, and reduces the need for resource-intensive linguistic analysis for each new classification task. We also perform a wide range of experiments to determine the most informative features in the feature space, finding that simple, easily extractable features suffice for good verb classification performance.","download":"http://www.cs.toronto.edu/~joanis/Papers/JoanisStevensonJames-nle13_3-2007.pdf","bibtex":"@Article{\t joanis1,\n author\t= {Eric Joanis and Suzanne Stevenson and David James},\n title\t\t= {A general feature space for automatic verb classification},\n journal\t= {Natural Language Engineering},\n volume\t= {14},\n number\t= {3},\n pages\t\t= {337--367},\n year\t\t= {2008},\n note\t\t= {Also published by <a\n\t\t href=http://journals.cambridge.org/action/displayIssue?jid=NLE&volumeId=14&issueId=03>Cambridge\n\t\t Journals Online</A> on December 19, 2006},\n abstract\t= {Lexical semantic classes of verbs play an important role\n\t\t in structuring complex predicate information in a lexicon,\n\t\t thereby avoiding redundancy and enabling generalizations\n\t\t across semantically similar verbs with respect to their\n\t\t usage. Such classes, however, require many person-years of\n\t\t expert effort to create manually, and methods are needed\n\t\t for automatically assigning verbs to appropriate classes.\n\t\t In this work, we develop and evaluate a feature space to\n\t\t support the automatic assignment of verbs into a well-known\n\t\t lexical semantic classification that is frequently used in\n\t\t natural language processing. The feature space is\n\t\t <em>general</em> – applicable to any class\n\t\t distinctions within the target classification;\n\t\t <em>broad</em> – tapping into a variety of semantic\n\t\t features of the classes; and <em>inexpensive</em> –\n\t\t requiring no more than a POS tagger and chunker. We perform\n\t\t experiments using support vector machines (SVMs) with the\n\t\t proposed feature space, demonstrating a reduction in error\n\t\t rate ranging from 48% to 88% over a chance baseline\n\t\t accuracy, across classification tasks of varying\n\t\t difficulty. In particular, we attain performance comparable\n\t\t to or better than that of feature sets manually selected\n\t\t for the particular tasks. Our results show that the\n\t\t approach is generally applicable, and reduces the need for\n\t\t resource-intensive linguistic analysis for each new\n\t\t classification task. We also perform a wide range of\n\t\t experiments to determine the most informative features in\n\t\t the feature space, finding that simple, easily extractable\n\t\t features suffice for good verb classification performance.},\n download\t= {http://www.cs.toronto.edu/~joanis/Papers/JoanisStevensonJames-nle13_3-2007.pdf}\n\t\t \n}\n\n","author_short":["Joanis, E.","Stevenson, S.","James, D."],"key":"joanis1","id":"joanis1","bibbaseid":"joanis-stevenson-james-ageneralfeaturespaceforautomaticverbclassification-2008","role":"author","urls":{},"metadata":{"authorlinks":{}}},"bibtype":"article","biburl":"www.cs.toronto.edu/~fritz/tmp/compling.bib","downloads":0,"keywords":[],"search_terms":["general","feature","space","automatic","verb","classification","joanis","stevenson","james"],"title":"A general feature space for automatic verb classification","year":2008,"dataSources":["n8jB5BJxaeSmH6mtR","6b6A9kbkw4CsEGnRX"]}