Language Technologies Thesis Defense

  • Ph.D. Student
  • Language Technologies Institute
  • Carnegie Mellon University
Thesis Orals

Parsing and Generation for the Abstract Meaning Representation

A key task in intelligent language processing is obtaining semantic representations that abstract away from surface lexical and syntactic decisions.  The Abstract Meaning Representation (AMR) is one such representation, which represents the meaning of a sentence as labeled nodes in a graph (concepts) and labeled, directed edges between them (relations).  A traditional problem of semantic representations is producing them from natural language (parsing) as well as producing natural language from them (generation).  In this thesis, I present algorithms for parsing and generation for AMR.

In the first part of the thesis, I present a parsing algorithm for AMR that produces graphs that satisfy semantic well-formedness constraints.  The parsing algorithm uses Lagrangian relaxation combined with an exact algorithm for finding the maximum, spanning, connected subgraph of a graph to produce AMR graphs that satisfy these constraints.

In the second part of the thesis, I present a generation algorithm for AMR.  The algorithm uses a tree-transducer that operates on a spanning-tree of the input AMR graph to produce output natural language sentences.  Data-sparsity of the training data is an issue for AMR generation, which we overcome by including synthetic rules in the tree-transducer.

Thesis Committee:
Jaime Carbonell (Chair)
Noah Smith (Chair, now University of Washington)
Chris Dyer (Chair, now Google DeepMind)
Dan Gildea (University of Rochester)

Copy of Thesis Document

For More Information, Please Contact: