CLARIN_MT_EN

This dataset includes trained Marian NMT models for Maltese-English translation, exploring the impact of BPE vocabulary size (5k, 10k, 20k) on translation quality. Models were trained on approximately 24,000 sentence pairs from the ELRC-SHARE Maltese-English parallel corpus. Each model was fine-tuned for 5 epochs and evaluated using BLEU and chrF. Additionally, selected samples were linguistically annotated using UDPipe and published as a searchable concordance corpus on AutoSearch. This release includes models, vocabulary files, tokenized corpora, and documentation for reproducibility.