Topic Modelling Used to Improve Arabic Web Pages Clustering

Topic modelling main purpose is to have machine-understandable and semantic annotation to textual contents of Web.It aim to extract knowledge rather than unrelated information. In this paper, we evaluate the impact of using topic model (which intended to represent the documents like a combination of topics where each topic is a mix of vectors) in improving documents clustering results. We have compared the results of clustering using PLSA or LSA. The experiments performed on a set of common newspaper websites that have highly dimensional data and we use Purity, Mean intra-cluster distance (MICD) and Davies-Bouldin index (DBI) for clustering evaluation. Thus, we acquired favorable clustering results, especially in the context of the Arabic language as PLSA were effective in minimizing MICD, expanding purity and bringing down DBI.