| Popis: |
Analyzing emerging topics from social media enables users to overview social movement and several web services to adopt current trends. Although existing studies mainly focus on extracting global emerging topics, efficient extraction of local ones related to a specific theme is still a challenging and unavoidable problem in social media analysis. We focus on extracting emerging social topics related to user-specified query words, and propose an extraction framework that uses a non-negative matrix factorization (NMF) modified for detecting temporal concentration and reducing noises. We conduct preliminary experiments for verifying our method using a Twitter dataset. |
