Semi-supervised document clustering algorithms based on seeds and LDA
Pin Zhou · Jisuanji gongcheng yu sheji · 2014
To reduce the data sparsity of semi-supervised clustering algorithms,two semi-supervised document clustering algorithms based on latent dirichlet allocation( LDA) which adopted seeds were proposed,namely Seeded-LDA and Constrained-LDA. SeededLDA uses seeds that were obtained from document labels to initialize parameters,Constrained-LDA constrained subsequent cluster assignment during the clustering process and made it same with labels. Experiments on realistic document datasets showed that the proposed algorithms had better clustering results and lower data sparsity compared with other semi-supervised clustering algorithms based on K-Means algorithm.