WIT: Wikipedia-based Image Text Dataset for Multimodal Multilingual Machine Learning.

Krishna Srinivasan, Karthik Raman, Jiecao Chen, Michael Bendersky, Marc Najork

Browse the full SIGIR paper archive.