Identifying Users' Gender via Social Representations

Gender prediction has evoked great research interests due to its potential applications like targeted advertisement and personalized search. Most of existing studies rely on the content texts. However, the text information is hard to access. This makes it difficult to extract text features. In this paper, we propose a novel framework which only involves the users' ids for gender prediction. The key idea is to represent users in the embedding connection space. We present two strategies to modify the word embedding technique for user embedding. The first is to sequentialize users' ids to get the order of social context. The second is to embed users into a large-sized sliding window of contexts. We conduct extensive experiments on two real data sets from Sina Weibo. Results show that our method is significantly better than the state-of-the-art graph embedding baselines. Its accuracy also outperforms that of the content based approaches