Gamification for Word Sense Labeling
Résumé
Obtaining gold standard data for word sense disambiguation is important but costly. We show how it can be done using a " Game with a Purpose " (GWAP) called Wordrobe. This game consists of a large set of multiple-choice questions on word senses generated from the Groningen Meaning Bank. The players need to answer these questions, scoring points depending on the agreement with fellow players. The working assumption is that the right sense for a word can be determined by the answers given by the players. To evaluate our method, we gold-standard tagged a portion of the data that was also used in the GWAP. A comparison yielded promising results, ranging from a precision of 0.88 and recall of 0.83 for relative majority agreement, to a precision of 0.98 and recall of 0.35 for questions that were answered unanimously.
Origine | Fichiers produits par l'(les) auteur(s) |
---|
Loading...