Paper Title
PERSONALIZED HASHTAG RECOMMENDATION BASED ON HIERARCHICAL USERS AND GROUP POSTS CO-ATTENTION WITH GENERATIVE ADVERSARIAL NETWORKS

Abstract
Abstract - Hashtags can help people find what they are looking for quickly and increase engagement on social platforms. Currently, tag recommendation is still a difficult task. Different users care about different types of information due to different preferences. Different users may tag similar posts with different hashtags. Under such circumstances, how to utilize multimodal data effectively and learn the tagging habits of users has become a key part of personalized tag recommendation, which has not been considered by existing methods.This study proposes a personalized hashtag recommendation method based on hierarchical users and group posts co-attention and conditional generative adversarial networks. In addition to considering post text and images, we also use a novel user tagging habit model and group content feature extraction model to enable our proposed method to predict tags more effectively. The proposed method adopts a new co-attention mechanism to enhance the important features and correlations of multimodal data. We adopt a generative adversarial network (GAN) based model to make use of the powerful ability and predict the tags more effectively. We evaluate the effectiveness of the proposed method on the real Instagram dataset. The experiment results demonstrate that our proposed method outperforms several representative methods in the field of personalized tag recommendation. Keywords - Personalized Hashtag Recommendation, User Tagging Habit, Multimodal Feature Extraction, Co-Attention Mechanism, Generative Adversarial Network