ConSOM: A conceptional self-organizing map model for text clustering

In the novel conceptional self-organizing map model (ConSOM) proposed for text clustering in this paper, neurons and documents can be represented by two vectors: one in extended concept space, and the other in traditional feature space, and weight modification of neuron vector is guided by combination of similarities in both traditional and extended spaces. Experimental results show that by utilizing concept relevance knowledge effectively, ConSOM performs better than traditional “SOM plus VSM” mode in text clustering due to its semantic sensitivity.