Title:  Graph- based parsing into logical forms

Speaker: 孙薇薇

Time:  24 DEC, 2019, 15:10-18:00

Place:  Classroom 406, Second Teaching Building, PKU

Abstract：

        In this talk, I will introduce English Resource Semantics (ERS), the semantic representations produced by the English Resource Grammar. ERS includes not only basic predicate-argument structures, but also information about various complex phenomena such as the distinction between scopal and non-scopal arguments, conditionals, comparatives, and many others.  High-performance parsing models for such richly detailed meaning representations can provide an application-independent intermediate layer of interpretation and thus significantly enrich the toolkit of automatic linguistic analysis for semantically-sensitive NLP tasks. I will introduce the linguistic and formal designs of ERS, in order to make this resource more accessible. After a short tour of the fundamental theory, I will introduce recent progress on neural ERS parsing. In particular, I will discuss the parsing techniques developed under the umbrella of graph-based meaning representations, including composition-, factorization- and transition-based models.