DemoHash: Hashtag recommendation based on user demographic information

Abstract

Social network services have become widely used, and hashtags, which are implicitly involved in delivering specific information, have shown to greatly improve user engagement. A number of prior studies have attempted to recommend appropriate hashtags for each social media user considering his/her posts by consequently extracting the important features from text and images. To develop this multi-dimensionality with hashtag recommendation, user demographic information also plays a significant role in the manner of personalized hashtag recommendation. Thus, this paper proposes the demographic hashtag recommendation (DemoHash) model to utilize users’ demographic information extracted from their selfie images, in addition to textual and visual information. The experimental results with the datasets from Instagram show that our proposed model achieves a greater performance with F1-score, Precision, and Recall than the existing hashtag recommendation methods by average of 4.19%, 18.45%, and 3.91%, respectively. Our approach effectively combined the content-based as well as user-oriented modeling for personalized hashtag recommendation.