Topic-Guided Automatical Human-Simulated Tweeting System

Social network grows increasingly popular in our life nowadays. It's an interesting intelligent behaviour to automatically post tweets in social network, which has not yet been explored. However, the associated researches are trapped in the problem of "multivalued mapping" where the agent should generate various appropriate tweets given a certain topic. In this paper, a human-simulated tweeting system is first designed to generate the multiple and appropriate tweets for given topics. In this system, a novel topic-image-tweet scheme is proposed with a Keyword-Based Retrieval Module (KBR-Module) and a Topic-Guided Image Captioning Module (TGIC-Module), where multiple topic-related images are searched in KBR-Module and encoded to generate the accurate tweets in TGIC-Module. The effectiveness of the proposed system and the superiority of our specific image captioning model are evaluated by sufficient quantitative comparisons and qualitative analysis in a real-world Twitter dataset.