Probabilistic models of graphs for natural language processing


Sorcha Gilroy

The aim of the project is to carry out machine translation through graphical representations of sentences. This will include theoretical study of graph formalisms, specifically how to define graph grammars which can carry out translation. Formalisms that will be studied include hyperedge replacement grammars and combinatory categorial grammars. The project will also include empirical work on data sets which contain existing semantic representations of sentences. In particular, the Prague Czech English Dependency Tree Bank and the Abstract Meaning Representation data set.


Supervisors: Adam Lopez & Sebastian Maneth