Evaluation of clustering algorithms for Polish Word Sense Disambiguation

Word Sense Disambiguation in text is still a difficult problem as the best supervised methods require laborious and costly manual preparation of training data. Thus, this work focuses on evaluation of a few selected clustering algorithms in task of Word Sense Disambiguation for Polish. We tested 6 clustering algorithms (K-Means, K-Medoids, hierarchical agglomerative clustering, hierarchical divisive clustering, Growing Hierarchical Self Organising Maps, graph-partitioning based clustering) and five weighting schemes. For agglomerative and divisive algorithm 13 criterion function were tested. The achieved results are interesting, because best clustering algorithms are close in terms of cluster purity to precision of supervised clustering algorithm on the same dataset, using the same features.