Semi-Supervised WSD in Selectional Preferences with Semantic Redundancy
Xuri Tang, Xiaohe Chen, Weiguang Qu, Shiwen Yu · 2012
This paper proposes a semi-supervised approach for WSD in Word-Class based selectional preferences. The approach exploits syntagmatic and paradigmatic semantic redundancy in the semantic system and uses association computation and minimum description length for the task of WSD. Experiments on Predicate-Object collocations and Subject-Predicate collocations with polysemous predicates in Chinese show that the proposed approach achieves a precision which is 8 % higher than the semanticassociation based baseline. The semisupervised nature of the approach makes it promising for constructing large scale selectional preference knowledge base. 1