Multi-view subspace text clustering

Text clustering has become an important challenge in artificial intelligence since several applications require to automatically organize documents into homogeneous topics. Given the availability of several text representation models, text documents can be organized through a multi-view text clustering approach. In this context, we propose a new subspace multi-view text clustering method (MVSTC). The proposed method offers a rich representation of text by integrating several models to detect different aspects of text such as syntactic, topic, and semantic features. MVSTC is capable of discovering latent correlations between documents by projecting the data onto a topological map. MVSTC seeks a subspace representation based on a low-rank and sparse representation to capture the global and local structure of multi-view textual data. Extensive experiments on real text data sets demonstrate that our method outperforms the existing multi-view clustering methods in terms of several evaluation metrics.