Leveraging Hashtag Networks for Multimodal Popularity Prediction of Instagram Posts

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 7191–7198 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 7191 Leveraging Hashtag Networks for Multimodal Popularity Prediction of Instagram Posts Yu-Yun Liao Graduate Institute of Linguistics National Taiwan University r09142002@ntu.edu.tw Abstract With the increasing commercial and social importance of Instagram in recent years, more researchers begin to take multimodal approaches to predict popular content on Instagram. However, existing popularity prediction approaches often reduce hashtags to simple features such as hashtag length or number of hashtags in a post, ignoring the structural and textual information that entangles between hashtags. In this paper, we propose a multimodal framework using post captions, image, hashtag network, and topic model to predict popular influencer posts in Taiwan. Specifically, the hashtag network is constructed as a homogenous graph using the co-occurrence relationship between hashtags, and we extract its structural information with GraphSAGE and semantic information with BERTopic. Finally, the prediction process is defined as a binary classification task (popular/unpopular) using neural networks. Our results show that the proposed framework incorporating hashtag network outperforms all baselines and unimodal models, while information captured from the hashtag network and topic model appears to be complementary. Keywords: Hashtag Network, Instagram Popularity Prediction, Multimodal Deep Learning 1. Introduction Over the past few decades, social media have become one of the most powerful and effective devices for spreading information. Companies and influencers around the world share millions of ideas and launch advertisement campaigns everyday through social media. In particular, Instagram, a photo and video sharing platform launched in 2010, has achieved widespread popularity among younger generations and Internet influencers these days. As a result, Instagram have gradually gained tremendous commercial and social importance as it unveils novel business strategies and marketing possibilities for not only companies and organizations, but also influencers, politicians, entertainers, and even country governments. Due to the increasing significance of Instagram, multimodal prediction of popular content on Instagram has become of great interest to researchers in the field of digital marketing or social media analytics. On Instagram, users can utilize a digital arsenal of photos, videos, filters, captions, hashtags, or stories to create appealing content for their audience. Each of the modalities in an Instagram post often require different techniques to represent; for example, natural language processing for captions, computer vision for images, both of which involve knowledge in data analysis, neural networks, or deep learning. However, among different modalities, hashtags are often ignored or over-simplified as features such as hashtag length, number of hashtags in a post or hashtag frequency. Some even consider hashtags as part of the post caption, failing to represent hashtags as individual and informative features. With the above being said, the present study proposes a multimodal framework considering post caption, image, and incorporating hashtag networks for predicting popular Taiwanese influencer posts on Instagram. Unlike existing Instagram multimodal frameworks, this study considers the co-occurrence relationship between all hashtags in the corpus and constructs hashtags as a homogeneous …

Content truncated to 500 words