chunked/nltk_chunking/model_doc_lxmert/chunk_4.txt

We thus propose the LXMERT (Learning Cross-Modality
Encoder Representations from Transformers) framework to learn these vision-and-language connections.