This model was trained using the MTUOC training framework and the Marian-NMT toolkit. A general Catalan-English model was first created using data from HPLT and NLLB, which comprised 16,037,694 sentence pairs after cleaning, rescoring, and language verification.

Afterwards, we curated a domain-specific corpus for Law consisting of magazines from the Hemeroteca Científica Catalana of the Institut d’Estudis Catalans, teaching materials from the Universitat Oberta de Catalunya and Wikipedia articles. The specialised dataset contained 524,092 sentence pairs, which were used to fine-tune the general model for the Law domain.

This model can be used for machine translation from Catalan to English, specifically optimised for texts in domain of Law and Legal Studies.

You can download this model and run it with MTUOC-server, a dedicated framework for translation models which includes Marian NMT. MTUOC-server handles preprocessing, tokenization, and translation requests.

The server is available at MTUOC-server GitHub, where installation instructions, usage examples, and configuration guides are provided. Once the server is running, you can send sentences in Catalan and receive English translations generated by the model.

At the time of submission, no measures have been taken to estimate bias or toxicity embedded in the model. We are aware that the model may be biased, and research on this topic may update this model card in the future.

The general model was trained on a combination of datasets including Opus and HPLT, which were cleaned, rescored using SBERT, and language-verified. Catalan segments were adapted to the new orthographic standard from the Institut d’Estudis Catalans, producing 16,037,694 parallel segments.

The domain-specific fine-tuning for Law used corpora from:

For a total of 524,092 sentence pairs.

Data preparation: All datasets are deduplicated and filtered to remove sentence pairs with a cosine similarity below 0.75. Similarity scores were computed using a MTUOC rescoring system for the Wikipedia, Opus, and HPLT datasets, and LaBSE for the others.

All Catalan segments were then normalised to follow the IEC's updated grammar guidelines using an MTUOC script.

Tokenization: All data is tokenized using SentencePiece with a shared encoder/decoder vocabulary of 64k tokens.

Metrics: BLEU, chrF2, TER on a domain-specific evaluation corpus.

The evaluation results for the Law domain show a markedly positive scenario. The fine-tuned model Marian FT 2 56 achieves: 68.5 BLEU, 82.4 chrF2 and 21.8 TER, and clearly surpasses the general base model Marian GEN 2 (63.5 BLEU, 25.6 TER), and Google Translate (64.2 BLEU, 24.8 TER).

Author: Antoni Oliver and Gemma Segués, Universitat Oberta de Catalunya. Contact: aoliverg@uoc.edu, gseguesm@uoc.edu Copyright: © 2025 Antoni Oliver and Gemma Segués, Universitat Oberta de Catalunya. Funding: Supported by Institut d’Estudis Catalans through the project Eines d’intel·ligència artificial per al foment de la comunicació científica en català.