Facebook always comes with features that surprise its users.
Recently Facebook has developed artificial intelligence, which is capable of accurately translating between any pair of 100 languages. Yes, you get it right.
Now there is no need for you to feel like that the language is not according to the nation where you are living.
There is no need for you to rely on the first translation into English in any case.
We all know that multiple systems are there that translate the content into English first and then you will be able to get the results.
But this problem will not be going to arise anymore.
Artificial intelligence outperforms such Systems by 10 points on a hundred point scales with the help of academics so that they can automatically evaluate the quality of machine translation.
The translations that will be going to produce by the model were assessed by humans and they almost score 90% accurately.
The system of Facebook was trained on the data which was a set of 7.5 billion sentences that are paired together from the web across a hundred languages.
It is quite astonishing for all the users out there that they will be able to get the results directly.
There are not all the languages included in it or have an equal number of sentence pairs. But you will be able to get answers easily.
Now users are feeling happy that English as a middleman has been gone and there is no need for them to engage in a necessary Hustle.
Globally, there are plenty of regions where people speak to languages but these are not English.
This has been mentioned by Angela fan of Facebook AI who moves ahead with this work.
This model came into existence by focusing on languages that are commonly translated to and from each other.
These are the grouping languages into 14 separate collections and all of them are based on Geography and cultural similarities as well.
This is to ensure that high-quality translation of a more commonly used connection must be there so that the model must be trained more accurately.
It is quite interesting to note that for some language pair the new system shows significant improvement as compared to the previous translation quality available.
For example, when you are translating from Spanish to French it becomes quite a stronger and good translation because Spanish is the second most spoken language worldwide this clearly indicates that researchers had access to a large amount of training data.
When they are doing the translation between English and Belarusian then also the quality has been improved over the existing results because artificial intelligence learns from translating Russian and their similarities with Belarusian as well.
This system is not yet used on the social network site but Facebook is planning to put it to work soon so that it will be able to handle the 20 billion translations made every day when people click translate on a post written in more than 160 languages.
Future work will be done in other languages as well. It was also mentioned by Fan. Moreover, especially the languages where they do not have a lot of data like South East Asian and African language, there will be going to work on the same as well.
The work breaks away from the English of scientific models and tries to build up more diverse multilingual ones.
This also has been mentioned by a person from the center at Dublin City University.
It is quite refreshing, but also some human assessment needs to be focused when making it happen.
In case it is not focusing on common people, then a user will not be able to get the results as they are expecting.
Moreover, Castillo also mentioned that the human assessment only looked at a small fraction of examples and making it hard to know whether the accurate judgment is there or not about high performance or not.
She also mentioned that there is a problem that evaluation will be going to then by bilingual volunteers apart from considering the professional translators.
Non-professional lack knowledge of translation so to get accurate changes it is important to approach those people.
Some people also mention that they are looking forward to the integration of this AI soon into Facebook so that the middleman English will get terminated.
They also claim you have a slight improvement in English centric model.
It is a bit empty but also mentioned that with time they will be ascertained results which led them to grab all the information as they are expecting.
There will be no problem that will arise for the user in any case at all.
Also, soon it will be going to appear on site so that users will be able to get the results as they are expecting and there will be no such trouble, in any case, anymore.