It is now common to have a conversation with others on social media. Many research have been taken to see the community structure on social media, but there are few studies that apply link-based community (link community) detection on a large social network. Link community detection allows users to belong to more than one community. We improve the method of existing link community detection of Ahn et al., which extracts many small communities. We evaluate existing and proposing methods by network indexes, and we characterize link communities from users' biographies. We found that link communities sharing users have similar characteristics from biographies.