Multi-Modal Mention Topic Model for mentionee recommendation

As one of the most commonly used communication functions in Twitter-like social media systems, mention is playing an important role in users’ online interactions. With the dramatic increase in the number of social media users, the problem of mentionee recommendation, i.e., recommending mentionees (mentioned users) when mentioners (mentioning users) attempt to mention others, has received considerable attention in recent years. While an increasing line of work has studied this problem, the existing efforts focus only on the contribution of the non-visual data like post text. In fact, many social media posts contain not only textual but also visual contents like images, and these two heterogeneous data sources both describe users’ mentioning tendencies. In this work, we proposed a novel generative model, named Multi-modal Mention Topic Model (MMTM), to tackle the mentionee recommendation problem by learning users’ semantic patterns and the correlations between contents in different modalities of users’ multi-modal mentioning documents in a unified way. Extensive experiments were conducted on a real-world dataset to evaluate the performance of our method. The experiment results demonstrated the superiority of our method in terms of making more effective recommendations compared with other state-of-the-art methods.