Abstract:Recent years have witnessed the flourishing of social media platforms (SMPs), such as Twitter, Facebook, and Sina Weibo. The rapid development of these SMPs has resulted in increasingly large scale multimedia data, which has been proved with remarkable marketing values. It is in an urgent need to classify these social media data into a specified list of concerned entities, such as brands, products, and events, to analyze their sales, popularity or influences. But this is a rather challenging task due to the shortness, conversationality, the incompatibility between images and text, and the data diversity of microblogs. In this paper, we present a multi-modal microblog classification method in a multi-task learning framework. Firstly features of different modalities are extracted for each microblog. Specifically, we extract TF-IDF features for each microblog text and low-level visual features and high-level semantic features for each microblog image. Then multiple related classification tasks are learned simultaneously for each feature to increase the sample size for each task and improve the prediction performance. Finally the outputs of each feature are integrated by a Support Vector Machine that learns how to optimally combine and weight each feature. We evaluate the proposed method on Brand-Social-Net to classify the contained 100 brands. Experimental results demonstrate the superiority of the proposed method, as compared to the state-of-the-art approaches.

Identifying User Interests based on Microblog Classification

Measuring and Visualizing Interest Similarity between Microblog Users

A Content-Based Intelligent Ranking Model for Microblog

Inferring User Interests in Microblogging Social Networks: A Survey

Automatic Identification of Chinese Weblogger's Interests Based on Text Classification

Mining the Interests of Chinese Microbloggers Via Keyword Extraction

Topic model-based micro-blog user interest analysis

Research and Application of Personalized Modeling Based on Individual Interest in Mining

A comparative study of users' microblogging behavior on sina weibo and twitter

Mining Microblog User Interests Based on TextRank with TF-IDF Factor

Gender Classification of Chinese Weibo Users

A New Users Interests Prediction Method Based on Social Hub in Microblog

A Simple Integration of Social Relationship and Text Data for Identifying Potential Customers in Microblogging.

User Interests-Based Microblog Tracing Algorithm

Identifying Important Users in Sina Microblog

Multi-modal microblog classification via multi-task learning

Beyond Your Interests: Exploring the Information Behind User Tags.

The Multi-Tag Semantic Correlation Used for Micro-Blog User Interest Modeling.

SpamDia: Spammer Diagnosis in Sina Weibo Microblog

Analysis and Identification of Spamming Behaviors in Sina Weibo Microblog

Mining the Personal Interests of Microbloggers via Exploiting Wikipedia Knowledge