AI Translation: Impact Could AI Save Europe's Rare and Endangered Languages?
It will soon be more accessible to see Facebook and Instagram posts in 200 lesser-spoken languages around the world. Meta’s No Language Left Behind (NLLB) project announced in a paper published this month that they’ve scaled their original technology. The project includes a dozen “low resource” European languages, like Scottish Gaelic, Galician, Irish, Lingurian, Bosnian, Icelandic, and Welsh.
The Importance of Native Speaker Consultation
Experts suggest that to improve the service, Meta should consult with native speakers and language specialists as the AI translation tool still requires refinement. Meta trains its AI with data from the Opus repository, an open-source platform with a collection of authentic text for various languages. This data is used to create a multilingual language model (MLM) that enables the AI to translate between different languages without relying on English data. The NLLB team evaluates the quality of translations and has seen a 44% improvement in accuracy since their first model was released in 2020.
Challenges and Potential Solutions
While the advancements in AI translation are impressive, challenges remain in ensuring accurate translations, particularly for rare and endangered languages like Scottish Gaelic. Native speakers, such as William Lamb, professor of Gaelic ethnology and linguistics, emphasize the importance of direct consultation with those who speak these languages fluently. Lamb notes that the quality of translations in Scottish Gaelic is not yet optimal due to the data sources used.
Collaboration between linguists, language specialists, and Big Tech companies like Meta is seen as essential for refining the data sets and improving the accuracy of AI translations. Alberto Bugarín-Diz, a professor of AI, advocates for a combination of expertise from humanities and technical backgrounds to enhance the quality of language data used by AI systems.
Community Involvement and Future Prospects
By integrating AI translation technology with the NLLB project, Meta aims to make lesser-spoken languages more accessible on social media platforms. However, community involvement and collaboration with native speakers are crucial for the continued refinement and perfection of these AI models. Experts agree that ongoing improvements and engagement with communities speaking endangered languages are essential for preserving linguistic diversity and heritage.
ALSO, READ: India’s Legal Battle Against Highrich Online’s Crypto Ponzi Scheme Unfolds.
Despite the progress in natural language processing, ongoing efforts are needed to ensure the accuracy and relevance of AI translations, especially for languages at risk of extinction.
