Abstract Meaning Representation (AMR) is a semantic graph formalism designed to capture sentence meaning using a directed graph. Many systems treat AMR-to-text generation as a sequence-to-sequence problem, drawing upon existing models. The largest AMR dataset (AMR 3.0) provides a sequence format which is considered equivalent to the graph format. However, due to the position-sensitive nature of sequence-to-sequence models, graph traversal order affects system performance. In this work we explore the effect that different, valid orderings have on the performance of sequence-to-sequence AMR-to-text systems and find that changing the traversal order can result in a BLEU score drop of up to 17.5 on a state-of-the-art system.

This paper introduces the Persian Abstract Meaning Representation (AMR) guidelines, a detailed guide for annotating Persian sentences with AMR, focusing on the necessary adaptations to fit Persian’s unique syntactic structures. We discuss the development process of a Persian AMR gold standard dataset consisting of 1562 sentences created following the guidelines. By examining the language specifications and nuances that distinguish AMR annotations of a low-resource language like Persian, we shed light on the challenges and limitations of developing a universal meaning representation framework. The guidelines and the dataset introduced in this study highlight such challenges, aiming to advance the field.