User profiling on Tumblr through blog posts
Kriti Pandey, Ayush Mittal · 2016
Tumblr, a microblogging service, has garnered a lot of prominence in recent past with 269.3 million blogs and 125.9 billion posts which makes it a hub for creative content. Despite this drastic growth of Tumblr since its launch, not a lot of scholarly work has been done on it. This work is an attempt to comprehensively profile the blogger in terms of age, gender and relationship status on Tumblr. The analysis conducted in this paper gives an approach for interest categorization of blogs based on topical analysis and writing characteristics. A normalized weighted representation of the most talked about categories for a blog was inferred based on the correlation between topical words and defined category related terms. The trained models predict age, gender and relationship status of the user with an accuracy of 82.37%, 79.61% and 76.52% respectively. To the best of our knowledge, this is the first attempt to profile users on Tumblr.