Cluster-Based News Representative Generation with Automatic Incremental Clustering

Nowadays we are facing much abundant information, especially news, and makes us confused in sorting out the information, so that it wastes our time in filtering that information. Though the news often contains similar contents that should save our time for reading. In this paper, we propose a new approach to provide aggregation mechanisms from cluster-based news and produce representative news, using our proposed Automatic Incremental Clustering. This approach presents a mechanism for clustering incremental news data and dynamically providing an automatic creation of new clusters. This approach consists of six main functions, which are (1) Data acquisition with incremental news sources from several news service providers, (2) Keyword extraction for term representation of news data, (3) Metadata aggregation for creating vector space of terms, (4) Automatic clustering for initiating news cluster generation, (5) Automatic incremental clustering for clustering incoming news data to pre-determined clusters or creating a new cluster of news data, and (6) News representation for selecting the most representative news of data clusters. For experimental study, we involved 95 news data service providers with 751 news data for for creating initial clusters with automatic clustering and 110 news data for incremental automatic clustering. Our approach performed 85.14% accuracy for incremental automatic clustering, and is able to dynamically create new clusters for incremental news data.