Monday, 15 December 2014 | 10:00 to 11:00

Speaker: Valerio Basile, University of Groningen

Title: Linguistic Games for Semantic Annotation: the Case of Wordrobe

The semantic annotation of large quantities of text is always a difficult task, because both the nature of the annotation itself (classifications are often debated) and the sheer quantity of manual work involved. The Groningen Meaning Bank [1] is a corpus of written English freely distributed and comprising several layers of annotation, from morphology to syntax to logical form expressing the semantics of the documents. In order to build the GMB, we developed a wiki-like interface for expert linguists. At the same time, we launched Wordrobe, a collection of online linguistic games aimed at the ordinary player, not necessarily an expert linguist. By answering quiz-like questions, players of Wordrobe from all over the world contribute to enlarge and refine the annotation of the GMB as they collect points and occasional prizes. So far, nine games have been published on Wordrobe, exploring phenomena such as POS-tagging, Word Sense Disambiguation, Named Entity Recognition, Anaphora Resolution and others. The success of Wordrobe, in terms of both quantity and quality of the annotation extracted from the games [2], shows the benefits of applying Gamification techniques to Computational Linguistics, and paved the road for future work in this direction.

[1] V. Basile, J. Bos, K. Evang, N. Venhuizen (2012): Developing a large semantically annotated corpus. Proceedings of LREC 2012.

[2] N. Venhuizen, V. Basile, K. Evang, J. Bos (2013): Gamification for Word Sense Labeling. Proceedings of the 10th International Conference on Computational Semantics (IWCS 2013), pp 397–403, Potsdam, Germany




Sala Consiglio, FBK, Povo (TN)