Identifying the Style of Chatting

Group chat brings great convenience to people's online social interaction. When identifying users in online social networks, individual speaking style of the member in group chat particularly plays an important role. In this paper, we propose a novel and efficient group chat hashing framework termed Group Chat Style Hashing(GCSH), which is the first work to utilize personal hashes in group chats to identify the speakers. GCSH combines a supervised VAE and a discriminator to extract accurate identity-related features and applies a new sample aggregation hash algorithm to generate an exclusive identity style hash for each one in the group chat. Our model fuses more context information to directly get the representative hashes, achieving an end-to-end identification model and sharply reducing the matching cost compared with previous text hashing methods. We conduct extensive experiments to prove the effectiveness and the generalization ability of our method across datasets.