Title: Context-Free Graph Grammars for Semantic Analysis, Generation and Machine 
Translation
Speaker: Daniel Bauer (Columbia University)
Place: Science Center. Rm 4102, CUNY Graduate Center. 5th Ave & 34th St.

Abstract:

Graphs are omnipresent in natural language processing where they represent deep
syntax, semantics (including coreference), and many other aspects of language.
Yet, unlike for trees, formal methods to manipulate graphs and translate
between graphs, trees, and strings are scarcely discussed. In our work, we
investigate semantic construction using Hyperedge Replacement Grammars (HRG), a
context free grammar formalism that describes languages of graphs. HRGs are
representationally expressive and maintain attractive algorithmic properties.
An extension of HRG to Synchronous Hyperedge Replacement Grammars (SHRG) allows
translation from strings or trees to graphs and back. Thus SHRG can be used to
build semantic parsers into (forests of) graph-shaped meaning representations
as well as natural language generators. We discuss semantic analysis and
generation and present two methods to automatically induce weighted SHRGs from
semantically annotated corpora. We employ these techniques in a new statistical
machine translation system that uses an intermediate graph-shaped meaning
representation and evaluate semantic parsing, generation, and end-to-end
translation. This work is the outcome of a summer internship at USC/ISI and
joint work with Jacob Andreas (Cambridge), Karl-Moritz Hermann (Oxford), Bevan
Jones (Edinburgh), and Kevin Knight (USC/ISI).

Bio:

Daniel Bauer is a third year doctoral student at Columbia University, where he
works with Owen Rambow on semantic parsing and turning textual descriptions
into 3D scenes. In his research he aims to combine syntactic parsing, semantic
parsing, knowledge representation, and inference to make language processing
systems more aware of domain and world knowledge. Daniel received a MSc in
Language Science and Technology from Saarland University and a BSc in Cognitive
Science from the University of Osnabrück, Germany. He spent this summer at
USC/ISI, working with Kevin Knight and David Chiang on semantics based machine
translation. Website: http://www.cs.columbia.edu/~bauer/