Unsupervised Discourse Constituency Parsing Using Viterbi EM
2020-01-01TACL 2020Code Available0· sign in to hype
Noriki Nishida, Hideki Nakayama
Code Available — Be the first to reproduce this paper.
ReproduceCode
Abstract
In this paper, we introduce an unsupervised discourse constituency parsing algorithm. We use Viterbi EM with a margin-based criterion to train a span-based discourse parser in an unsupervised manner. We also propose initialization methods for Viterbi training of discourse constituents based on our prior knowledge of text structures. Experimental results demonstrate that our unsupervised parser achieves comparable or even superior performance to fully supervised parsers. We also investigate discourse constituents that are learned by our method.