Word Sense Disambiguation Based on Distance Metric Learning from Training Documents

摘要

Word sense disambiguation task reduces to a classification problem based on supervised learning. However, even though Support Vector Machine (SVM) gives the distance from the data point to the separating hyperplane, SVM is difficult to measure the distance between labeled and unlabeled data points. In this paper, we propose a novel word sense disambiguation method based on a distance metric learning to find the most similar sentence. To evaluate the efficiency of the method of word sense disambiguation using the distance metric learning such asNeighborhood Component Analysis and Large Margin Nearest Neighbor, we make some experiments to compare with the result of the SVM classification. The results of the experiments show this method is effective for word sense disambiguation in comparison with SVM and one nearest neighbor. Moreover, the proposed method is effective for analyzing the relation between the input sentence and all senses of the target word if the target word has more than two senses.