Supporting Clustering with ContrastiveLearning (SCCL) is a novel framework to leverage contrastive learning topromote better separation . SCCL significantly advances the state-of-the-art results on most benchmark datasets with 3%-11% improvement on Accuracy and 4%-15% improvements on Normalized Mutual Information . The framework leverages the strengths of both bottom-up instance discrimination and top-down clustering to achieve better intra-clusters and inter-cluster distances when evaluated with ground truth cluster labels . The results are based on short text clustering and short text clusters on a benchmark dataset with the help of ground truth clusters .

Author(s) : Dejiao Zhang, Feng Nan, Xiaokai Wei, Shangwen Li, Henghui Zhu, Kathleen McKeown, Ramesh Nallapati, Andrew Arnold, Bing Xiang

Links : PDF - Abstract

Code :

Keywords : clustering - clusters - ground - cluster - results -

Leave a Reply

Your email address will not be published. Required fields are marked *