Crowdsourcing for linguistic inclusivity: leveraging Community Feedback for Faroese Machine Translation

There is a critical gap in language technology for low resource languages like Faroese. This situation is worsened by the growing centrality of large language models (LLMs) in various sectors. The scarcity of data, infrastructure and evaluation metrics hinders the development of fundamental applications such as machine translation. In this project, we aim at addressing Faroese-English machine translation by: (1) augmenting existing parallel sentences by leveraging few shot abilities of large language models, (2) fine-tuning a multilingual model to perform Faroese-English translation and (3) designing an evaluation platform to collect community feedback. This feedback would fuel further development of the model, evaluation metrics and dataset, enriching them with Faroese cultural elements.