A Document Clustering Algorithm Based on Semi-constrained Hierarchical Latent Dirichlet Allocation

摘要

The bag-of-words model used for some clustering methods is often unsatisfactory as it ignores the relationship between the important terms that do not cooccur literally. In this paper, a document clustering algorithm based on semi-constrained Hierarchical Latent Dirichlet Allocation (HLDA) is proposed, the frequent itemsets is considered as the input of this algorithm, some keywords are extracted as the prior knowledge from the original corpus and each keyword is associated with an internal node, which is thought as a constrained node and adding constraint to the path sampling processing. Experimental results show that the semi-constrained HLDA algorithm outperforms the LDA, HLDA and semi-constrained LDA algorithms.