Research And Implementation Of Social Hashtag Recommendation Algorithm Based On Attention Mechanism

Hashtag is a topic tagging method commonly used in social platforms.It can effectively improve the efficiency of information organization and information retrieval,thereby improving the convenience and usability of social platforms.Hashtag labeling is a tedious and time-consuming process,so most users of social platforms are reluctant to actively hashtag social media information.In view of this problem,using existing Hashtag tagging information,how to effectively recommend Hashtags has become one of the current hotspots of social platform research.Most Hashtag recommendation methods focus on the task of recommending text content,while Hashtag recommendation for images or text and images has relatively little work.In social platforms,images often contain rich semantic information,and making full use of this information can greatly improve the recommendation performance of Hashtags.Therefore,it is particularly important to effectively mine the association between image information or information combined with text and hashtags.Based on this,for the Hashtag recommendation problem in two different social scenarios,this paper proposes multiple recommendation algorithms based on the deep learning attention mechanism,and designs and completes a personalized forum system.The specific work is as follows:(1)For Hashtag recommendation in Instagram social,this paper designs two Hashtag recommendation network frameworks based on attention mechanism: Dual-Attention and Co-Attention models.These two frameworks use pre-trained VGG-16 models in two different scenes to extract image features,and use the attention mechanism to learn the coefficients of features.Comparative experiments on the open data set HARRISON verify the feasibility and effectiveness of the Dual-Attention and Co-Attention models.(2)Aiming at the multimodality of social information in twitter social application,a multimodal hashtag recommendation model based on multilevel attention mechanism isproposed.It can overcome the shortcomings of single-mode information that can not fully represent the content of microblog which combine image and text information together,and make full use of multilevel attention to obtain low-level and high-level features.Experiments on Twitter dataset show that the multilevel attentions can improve the performance of Hashtag recommendation tasks.(3)This paper finally designed a personalized campus forum system,and applied the Hashtag recommendation algorithm designed in this article to the forum's algorithm module,completed the design of the overall system framework and software design logic,and designed the corresponding database module,and finally completed the forum System development,testing and other work.