Spanish Abstract Meaning Representation: Annotation of a General Corpus

TL;DR

This paper presents the creation of a large, annotated Spanish AMR corpus, adapting the formalism for Spanish and providing gold-standard annotations to improve cross-lingual AMR parsing.

Contribution

It introduces the first sizable, general Spanish AMR annotation project, extending English AMR with Spanish-specific features and releasing a new annotated corpus for diverse genres.

Findings

Created 586 gold-standard Spanish AMR annotations

Extended English AMR with Spanish semantic features

Analyzed annotation disagreements and implications for cross-lingual AMR

Abstract

The Abstract Meaning Representation (AMR) formalism, designed originally for English, has been adapted to a number of languages. We build on previous work proposing the annotation of AMR in Spanish, which resulted in the release of 50 Spanish AMR annotations for the fictional text "The Little Prince." In this work, we present the first sizable, general annotation project for Spanish Abstract Meaning Representation. Our approach to annotation makes use of Spanish rolesets from the AnCora-Net lexicon and extends English AMR with semantic features specific to Spanish. In addition to our guidelines, we release an annotated corpus (586 annotations total, for 486 unique sentences) of multiple genres of documents from the "Abstract Meaning Representation 2.0 - Four Translations" sembank. This corpus is commonly used for evaluation of AMR parsing and generation, but does not include gold AMRs; we hope that providing gold annotations for this dataset can result in a more complete approach to cross-lingual AMR parsing. Finally, we perform a disagreement analysis and discuss the implications of our work on the adaptability of AMR to languages other than English.