A Study Towards Spanish Abstract Meaning Representation

Abstract

Taking into account the increasing attention that researchers of Natural Language Understanding (NLU) and Natural Language Generation (NLG) are paying to Computational Semantics, we analyze the feasibility of annotating Spanish Abstract Meaning Representations. The Abstract Meaning Representation (AMR) project aims to create a large- scale sembank of simple structures that represent unified, complete semantic information contained in English sentences. Although AMR is not destined to be an interlingua, one of its key features is the ability to focus on events rather than on word forms. They do this, for instance, by abstracting away from morpho-syntactic idiosyncrasies. In this thesis, we investigate the requirements to – and we come up with a proposal to – annotate Spanish AMRs, based on the premise that many of these idiosyncrasies mark differences between languages. To our knowledge, this is the first work towards the development of Abstract Meaning Representation for Spanish.