Generation of Abstract Meaning Representations by Hyperedge Replacement Grammars – A Case Study
Abstract: The abstract meaning representation (AMR) is a graph-based semantic representation of a natural language sentence that has been the target of numerous studies – in particular, a collection of such graphs over a small domain including the words boy, girl, want and believe has been presented. We investigate the suitability of two hyperedge replacement grammar based formalisms for the generation of AMRs in the form of a case study using the aforementioned boy-girl domain. The first formalism is predictive top-down parsable grammars and the second one is restricted directed acyclic graph grammars; we argue that the latter is more suitable in this limited case. Furthermore, we claim that the domain must be extended in order to draw further conclusions, and that other formalisms should then be considered since both grammar types evaluated here show signs of not being able to handle the generation of AMRs in general. Moreover, we provide a correctness proof for one of the grammars used in this study.
AT THIS PAGE YOU CAN DOWNLOAD THE WHOLE ESSAY. (follow the link to the next page)