Automatic Document Clustering: Amharic News Items

Document clustering has a vital role in information storage and retrieval. In order to retrieve set of documents, it would be easy if the collection was clustered. This research is an investigation to the development of a prototype system that can automatically cluster a collection of Amharic news items. There are different techniques to automatic document clustering. The variants of hierarchical and partitional techniques are the most widely used methods. This research followed partitional clustering methodology specifically k-means and find optimum cluster of documents using statistical approach. Both internal and external qualities of clusters were evaluated using news items from Ethiopian News Agency (ENA). Overall average pair- wise similarity and overall purity were used to measure the internal quality of clusters. The values of these parameters were found to be 0.0797 and 0.7840 respectively. External quality was also measured using f-measure and its value was found to be 0.7705