Clustering Based Approach for Novelty Detection in Text Documents

Abstract

As the information is overloaded over the internet accessing of information from the internet according to a given query provides redundant and irrelevant information. It is necessary to retrieve relevant and novel information from a given query by the user. With the result of this the user will require minimum effort to access the information need. In this work we proposed a clustering based approach for novelty detection which will provide the relevant and novel documents for the information need. Based on the user query the incoming stream of documents will be clustered using k-means algorithm. Then the cluster heads are selected from the various clusters with the minimum distance. These cluster heads are the novel documents from a collection of documents from different clusters having the large distance. The proposed technique can be further used in the field of information retrieval.