Facebook just introduced a new advance for its automatic translation tools, which is aimed at improving communications between users around the globe. It is based on the company’s M2M-100 multilingual machine translation (MMT) model. This model is capable of translating a total of 100 languages, without relying on English data.
English training data is the most widely available, but the M2M-100 model doesn’t rely on it, and therefore can better preserve meaning of various languages. According to Facebook, this advanced model outperforms English-centric systems by 10 points on the BLEU metric for evaluating machine translations.
Basically, with this new AI technology, the initial step of first translating a language to English is eliminated and therefore, in this way, this system is better at maintaining the original meaning of the text, and would ultimately maximize communication among people.
Moreover, Facebook’s translation tools also play a major role in maximizing business opportunities, as they facilitate cross border commerce into regions where there are language barriers. Due to lack of effective communication, businesses face difficulty when trying to expand their market in a particular region. Along with Facebook’s increasing focus on eCommerce, the translation tools could help businesses a great deal to tackle this problem.
Furthermore, Facebook is open-sourcing its work on its currency project, which is aimed at developing Facebook’s own unique currency that would be different from local banks, and therefore any business anywhere would be able to sell to anybody across the world on Facebook’s marketplace. This project would facilitate the platform’s improved translation tools more broadly.
Facebook’s primary motive as a social platform is to help connect the world, and the company doesn’t seem to hold back from developing innovative ways to achieve its goal in a better way each time, and continues to provide significant new opportunities to maximize global communication.