The Performance Analysis of a Chi-square Similarity Measure for Topic Related Clustering of Noisy Transcripts