How Google Translate works: Neutral Machine Translation

Everyone is curious to know how google translate works.

Since the launch of Google translate, the service has gone on to be the number one go-to source for translations.

Google fought off competitions from the likes of Babel Fish in the early days to dominate the translation service. Today, it boasts of more than 500m users daily worldwide offering about 103 languages.

How Google Translate works

In 2006, Google’s translation algorithm originally centred on statistical machine translation.

It worked by translating the required word, phrase or sentence into the English language as a baseline language before translating it into the target language and cross-referencing the target text with millions of documents taken from official United Nations and European parliament transcripts.

The original translations from google were not accurate but it revealed the intent of the original passage in a broader view.

However, it did not deliver accurate, coherent translations as would have been expected from a human expert.

This concept is contrary to the initial assumptions of how Google translate works.

There is a popular misconception that google hired experts to compile grammars for each mother tongue due to the ever-evolving nature of languages.

READ  How to download and use Google Translate Offline

However, Any attempt to do so would have quickly become obsolete.

In November 2016, Google announced the transition of its translation service from a statistical machine translation to a neutral machine translation premise.

Neutral Machine translation is a deep learning practice that changed Google’s way of translating by comparing whole sentences from a broader range of linguistic sources.

This increased the accuracy of translations resulting in full context than sentence clauses in isolation.

By comparing Spanish-to-English translations with Japanese-to-English translations, it deduces and identifies the relationship between Spanish and Japanese and makes translations between both languages.

Google can spot recurring patterns between words in different languages by processing these calculations over and over again.

This is how google translate works, feel free to go through other posts on our blog.

One Comment on “How Google Translate works: Neutral Machine Translation”

Leave a Reply

Your email address will not be published. Required fields are marked *