Semantic Neural Machine Translation using AMR — mentioned by orange3xchicken

Tue, 26 May 2020 16:14:07 GMT

It looks like they go from tree -> sequence via prefix notation. I'm curious why Lample decided on this seq2seq approach when it seems that there might be models which could be more naturally applied to the tree structure directly [1, 2] [1] https://arxiv.org/abs/1902.07282 (an AST translation system) [2] https://arxiv.org/abs/1609.02907 (GCNN)

Paper and discussion history. A mention is not an endorsement.

Semantic Neural Machine Translation using AMR · HackerPapers discussions

Semantic Neural Machine Translation using AMR — mentioned by orange3xchicken