- Modern translation systems are in high demand because over five billion people are using the internet worldwide and speak 7,151 languages.
- The NLLB-200, a single AI model that translates over 200 distinct languages with cutting-edge accuracy, was just unveiled by Meta as a significant development in this project.
The social metaverse firm, Meta, gets closer to achieving its goal of “giving people the power to establish community and bring the globe together” with every innovation. The company recently announced a research breakthrough in its No Language Left Behind (NLLB) project. The project aims to create high-quality machine translation capabilities for most of the world’s languages.
Less communication with more languages
It’s not surprising that modern translation systems are in high demand, given that over five billion people are using the internet worldwide who speak 7,151 languages. However, translation systems that aim to overcome linguistic obstacles in the consumption of digital material are constrained by the lack of linguistic data. Although Google Translate is a sophisticated, multilingual neural machine translation service, it can only translate 133 different languages.
Another translation application from one of the biggest technological corporations in the world, Microsoft Bing Translator, supports just over 100 languages. Many low-resource languages (mainly in Africa and Asia) are not supported in these systems, even though only 23 of the 7,151 world languages are spoken by over half of the world’s population. This suggests a restricted interaction between speakers of these languages and the desired content.
AI and translation for businesses
Translation is one of the interesting ways Artificial Intelligence (AI) is redefining human connection and its effectiveness. The market for machine translation and the application of AI in translation was assessed at USD 800 million in 2021 and is expected to reach USD 7.5 billion in value by 2030.
According to Global Market Insights, a critical factor in the growth of the machine translation market is businesses’ increasing desire to enhance client experiences. This is supported by research from Gartner, which shows that translation is a widespread enterprise concern. It has become more relevant in four major synchronous and asynchronous use cases like real-time multimedia, online customer sales and support, multimedia, and documents and texts.
Organizations that want to expand their presence globally need to implement inclusive translation solutions to address the increasingly complex needs of a global customer base. This is where Meta comes in.
A development in accurate machine translation
The NLLB project is Meta’s ambitious effort to create a universal language translator that can translate between any language regardless of the linguistic data available to the AI. It was initiated more than six months ago. The NLLB-200, a single AI model that translates over 200 distinct languages with cutting-edge accuracy, was just unveiled by Meta as a significant development in this project.
This paradigm encourages the professional translation of lesser-used languages, particularly those from Asia and Africa. For instance, the model facilitates the translation of 55 African languages that currently have minimal resource availability. This is a 46% improvement over what is possible with current translation technologies.
According to Meta, this model produces an average 44% gain in the overall bilingual evaluation understudy (BLEU) scores throughout the 10,000 directions of the FLORES-101 benchmark and outperforms current translation systems by more than 70% for various African and Indian languages.
To give a sense of the scale, Zuckerberg reveals that “the 200-language model has over 50 billion parameters, [trained] using [Meta’s] new Research SuperCluster (RSC), which is one of the world’s fastest AI supercomputers. The advances here will enable more than 25 billion daily translations across our apps.”
Despite this success, Meta understands that without creative cooperation, the project’s goals will not be met. It made the NLLB-200 model open source and offered to fund up to USD 200,000 to nonprofit organizations so they could use the NLLB-200 in their operations. These actions allow other researchers to broaden their linguistic reach and develop more inclusive technologies.
The broad ramifications of this methodology for the more than 25 billion translations across Meta’s platforms will hasten the development of stronger communities and cross-linguistic and cross-cultural cooperation. According to Zuckerberg, “Communicating across languages is one superpower that AI provides, but as we keep advancing our AI work, it’s improving everything we do — from showing the most interesting content on Facebook and Instagram, to recommending more relevant ads, to keeping our services safe for everyone.”
Wikipedia will also use this technology to translate its media content into more than 20 languages with few resources.
Mark Zuckerberg, founder, and CEO of Meta, said, “We just open-sourced an AI model we built that can translate across 200 different languages — many of which aren’t supported by current translation systems. We call this project No Language Left Behind, and the AI modeling techniques we used are helping make high-quality translations for languages spoken by billions of people around the world.”