Optimizing Group Discussion Generation Using K-Means Clustering And Fair Distribution