Word Similarity Based on an Ensemble Model Using Ranking SVMs
Hui Liu, Ruzhan Lu · 2008
A novel ensemble model is suggested to measure the similarity between two words. The authors apply ranking support vector machines to combine the results of existing similarity models. Both training and test data are extracted from the standard Miller & Charles dataset randomly. Evaluations by cross validation show that the ensemble model outperforms known similarity models for not only English words, but also Chinese words.