Graph Based Algorithm for Automatic Domain Segmentation of WordNet

We present a graph based algorithm for automatic domain segmentation of Word-net. We pose the problem as a Markov Random Field Classification problem and show how existing graph based algorithms for Image Processing can be used to solve the problem. Our approach is unsuper-vised and can be easily adopted for any language. We conduct our experiments for two domains, health and tourism. We achieve F-Score more than.70 in both do-mains. This work can be useful for many critical problems like word sense disam-biguation, domain specific ontology ex-traction etc.