Annotation of WordNet Verbs with TimeML Event Classes

Georgiana Puşcaşu, Verginica Barbu Mititelu


Abstract
This paper reports on the annotation of all English verbs included in WordNet 2.0 with TimeML event classes. Two annotators assign each verb present in WordNet the most relevant event class capturing most of that verb’s meanings. At the end of the annotation process, inter-annotator agreement is measured using kappa statistics, yielding a kappa value of 0.87. The cases of disagreement between the two independent annotations are clarified by obtaining a third, and in some cases, a fourth opinion, and finally each of the 11,306 WordNet verbs is mapped to a unique event class. The resulted annotation is then employed to automatically assign the corresponding class to each occurrence of a finite or non-finite verb in a given text. The evaluation performed on TimeBank reveals an F-measure of 86.43% achieved for the identification of verbal events, and an accuracy of 85.25% in the task of classifying them into TimeML event classes.
Anthology ID:
L08-1563
Volume:
Proceedings of the Sixth International Conference on Language Resources and Evaluation (LREC'08)
Month:
May
Year:
2008
Address:
Marrakech, Morocco
Editors:
Nicoletta Calzolari, Khalid Choukri, Bente Maegaard, Joseph Mariani, Jan Odijk, Stelios Piperidis, Daniel Tapias
Venue:
LREC
SIG:
Publisher:
European Language Resources Association (ELRA)
Note:
Pages:
Language:
URL:
http://www.lrec-conf.org/proceedings/lrec2008/pdf/712_paper.pdf
DOI:
Bibkey:
Cite (ACL):
Georgiana Puşcaşu and Verginica Barbu Mititelu. 2008. Annotation of WordNet Verbs with TimeML Event Classes. In Proceedings of the Sixth International Conference on Language Resources and Evaluation (LREC'08), Marrakech, Morocco. European Language Resources Association (ELRA).
Cite (Informal):
Annotation of WordNet Verbs with TimeML Event Classes (Puşcaşu & Mititelu, LREC 2008)
Copy Citation:
PDF:
http://www.lrec-conf.org/proceedings/lrec2008/pdf/712_paper.pdf