Deep NLP models benefit from underlying structures in the data---e.g., parse trees---typically extracted using off-the-shelf parsers. Recent attempts to jointly learn the latent structure encounter a tradeoff: either make factorization assumptions that limit expressiveness, or sacrifice end-to-end differentiability. Using the recently proposed SparseMAP inference, which retrieves a sparse distribution over latent structures, we propose a novel approach for end-to-end learning of latent structure predictors jointly with a downstream predictor. To the best of our knowledge, our method is the first to enable unrestricted dynamic computation graph construction from the global latent structure, while maintaining differentiability.
@article{arxiv.1809.00653,
title = {Towards Dynamic Computation Graphs via Sparse Latent Structure},
author = {Vlad Niculae and André F. T. Martins and Claire Cardie},
journal= {arXiv preprint arXiv:1809.00653},
year = {2018}
}