Unsupervised Discourse Constituency Parsing Using Viterbi EM
Unsupervised Discourse Constituency Parsing Using Viterbi EM
复制标题
DOI：
10.1162/tacl_a_00312
复制
发表时间：
2020-04
影响因子：
10.9
通讯作者：
Noriki Nishida;Hideki Nakayama
中科院分区：
文献类型：
--
作者：
Noriki Nishida;Hideki Nakayama
本文介绍了一种无监督语篇成分分析算法。我们使用Viterbi EM和基于边界的准则来以无监督的方式训练基于SPAN的语篇分析器。我们还提出了基于我们的文本结构先验知识的语篇成分维特比训练的初始化方法。实验结果表明，我们的无监督分析器的性能与完全监督分析器相当，甚至更好。我们还调查了通过我们的方法学习的语篇成分。
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.