adaptNMT: an open-source, language-agnostic development environment for neural machine translation

Lankford, Seamus; Afli, Haithem; Way, Andy

doi:10.1007/s10579-023-09671-2

Cited by 1 publication

(3 citation statements)

References 41 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Clearly, the system translating in the GA → EN direction performs better, when evaluated using both automatic and human evaluation, than its counterpart when translating in the opposite direction. These results are consistent with our previous work, which also show better GA → EN translation performance [5]. This performance difference is attributed to the morphologically rich nature of the Irish language, which relies heavily on inflection, derivation, and its case system.…”

supporting

confidence: 92%

“…A significant aspect of our research involves creating applications and models to address language technology challenges. Similar to our previous work, which focused on developing NMT models [5], we hope this paper will be particularly helpful for those new to MT wishing to learn more about fine-tuning MLLMs.…”

Section: Introductionmentioning

confidence: 83%

“…The highest-performing EN → GA system in the LoResMT2021 Shared Task was submitted by the ADAPT team [39]. The model was developed with an in-house application, adaptNMT [5] using a Transformer architecture. It performed well across all key translation metrics (BLEU: 36.0, TER: 0.531 and ChrF3: 0.6).…”

Section: Translation In the En ↔ Ga Directionsmentioning

confidence: 99%

See 2 more Smart Citations

adaptMLLM: Fine-Tuning Multilingual Language Models on Low-Resource Languages with Integrated LLM Playgrounds

Lankford,

Afli,

Way

2023

Information

Self Cite

View full text Add to dashboard Cite

The advent of Multilingual Language Models (MLLMs) and Large Language Models (LLMs) has spawned innovation in many areas of natural language processing. Despite the exciting potential of this technology, its impact on developing high-quality Machine Translation (MT) outputs for low-resource languages remains relatively under-explored. Furthermore, an open-source application, dedicated to both fine-tuning MLLMs and managing the complete MT workflow for low-resources languages, remains unavailable. We aim to address these imbalances through the development of adaptMLLM, which streamlines all processes involved in the fine-tuning of MLLMs for MT. This open-source application is tailored for developers, translators, and users who are engaged in MT. It is particularly useful for newcomers to the field, as it significantly streamlines the configuration of the development environment. An intuitive interface allows for easy customisation of hyperparameters, and the application offers a range of metrics for model evaluation and the capability to deploy models as a translation service directly within the application. As a multilingual tool, we used adaptMLLM to fine-tune models for two low-resource language pairs: English to Irish (EN↔ GA) and English to Marathi (EN↔MR). Compared with baselines from the LoResMT2021 Shared Task, the adaptMLLM system demonstrated significant improvements. In the EN→GA direction, an improvement of 5.2 BLEU points was observed and an increase of 40.5 BLEU points was recorded in the GA→EN direction representing relative improvements of 14% and 117%, respectively. Significant improvements in the translation performance of the EN↔MR pair were also observed notably in the MR→EN direction with an increase of 21.3 BLEU points which corresponds to a relative improvement of 68%. Finally, a fine-grained human evaluation of the MLLM output on the EN→GA pair was conducted using the Multidimensional Quality Metrics and Scalar Quality Metrics error taxonomies. The application and models are freely available.

show abstract

supporting

confidence: 92%

Section: Introductionmentioning

confidence: 83%