Transitioning from AMR to UMR in Turkish: A Semantic Representation Perspective

Abstract Meaning Representation (AMR) is a widely adopted semantic representation framework originally developed for English. Adapting AMR to morphologically rich and typologically diverse languages like Turkish poses unique challenges due to its agglutinative structure and free word order. Uniform Meaning Representation (UMR), a multilingual extension of AMR, aims to provide a more consistent and language-independent semantic annotation. In this study, we present the first guideline for converting Turkish AMR annotations into UMR, marking the first work on UMR adaptation for the Turkish language. We applied this guideline to a 100-sentence subset of The Little Prince AMR corpus using manual annotation by three experts. We discuss the guideline development process, annotation experience, and initial observations on the differences and potential advantages of UMR compared to AMR for Turkish. Our work provides an important foundation for advancing semantic parsing research and cross-lingual representation for Turkish.