This paper presents a novel architecture titled “Recurrent Capsule Network” for text classification. The effectiveness of the proposed architecture is evaluated on “Toxic Comment Classification Challenge Dataset, Kaggle” provided by Jigsaw. We have also evaluated the performance of proposed recurrent capsule networks on other text classification benchmarks such as AG News, IMDB Large Movie Review and Yelp Reviews - Polarity and Full.

Thank You!

Important Links

Tackling Toxic Online Communication with Recurrent Capsule Networks
GitHub Repository
Toxic Comment Classification Challenge - Kaggle
Dymanic Routing between Capsules