Dialectal and Low-Resource Machine Translation for Aromanian

TL;DR

This paper develops a neural machine translation system for Aromanian, creating the largest parallel corpus and analyzing models to support language preservation and computational linguistics.

Contribution

It introduces the largest Aromanian-Romanian corpus and compares multiple translation models, along with auxiliary tools for language processing.

Findings

Created a 79,000 sentence pair corpus for Aromanian-Romanian

Compared several translation models optimized for Aromanian

Provided publicly available datasets and tools for language preservation

Abstract

This paper presents the process of building a neural machine translation system with support for English, Romanian, and Aromanian - an endangered Eastern Romance language. The primary contribution of this research is twofold: (1) the creation of the most extensive Aromanian-Romanian parallel corpus to date, consisting of 79,000 sentence pairs, and (2) the development and comparative analysis of several machine translation models optimized for Aromanian. To accomplish this, we introduce a suite of auxiliary tools, including a language-agnostic sentence embedding model for text mining and automated evaluation, complemented by a diacritics conversion system for different writing standards. This research brings contributions to both computational linguistics and language preservation efforts by establishing essential resources for a historically under-resourced language. All datasets, trained models, and associated tools are public: https://huggingface.co/aronlp and https://arotranslate.com