Google Translate, towards better understanding


Have you tried booking an autorickshaw in Chennai? If you can’t speak Tamil, you are better off using Ola or Uber. However, sometimes you would like to hop into an autorickshaw without having to wait for a cab. For instance, getting an autorickshaw from Elliot’s Beach to Indira Nagar station in Chennai would be a lot quicker than booking a cab. But as a customer who can’t speak Tamil, you might be charged any amount between ₹100 and ₹300 for the three-kilometre ride. ‘Autos, autos everywhere. But not a ride for me.’

Recently, I used Google Translate’s audio translation feature to communicate with autorickshaw drivers. What initially seemed quite bizarre to both of us, ultimately was quite amusing and rather effective. It helped book an autorickshaw ride on metre plus ₹10 extra. That is a deal I’ll take any day.

Then and now

In a country where 23 languages are constitutionally recognised, technology like Google Translate offers a great way to bridge the linguistic barrier and understand each other.

Launched in 2006, Google Translate initially translated text into English, before translating into the selected language using predictive algorithms. In 2010, it was added as an app on Android devices and in 2011 on iOS. The tech had improved so much that the platform was seen as a portable personal interpreter. By now, Google Chrome could pronounce the text, recognise words in the picture and spot unfamiliar text and languages.

In 2016, the tech underwent a major back-end change when Harold Gilchrist, a software engineer, led Google’s research team to develop the Neural Machine Translation system (NMT), to increase fluency and accuracy in translations. Instead of translating word by word, NMT allows translating “sentences at a time, rather than just piece by piece. It uses this broader context to help it figure out the most relevant translation, which it then rearranges and adjusts to be more like a human speaking with proper grammar,” writes Barak Turovsky, Product Lead, Google Translate, in the blog ‘Found in translation: More accurate, fluent sentences in Google Translate’.

Today, the Google Translate app offers improved translations using the camera, audio and handwriting, thanks to NMT’s deep learning capabilities. As a result, more people can access the Android operating system, Google Maps and Google Pay services.

If you haven’t tinkered with the Google Translate app in a while, I recommend testing your portable personal interpreter’s skills. The improved speed and accuracy will surprise you.

Calibrating expectations

In 2017, Pixel Buds gave a glimpse of the future with its real-time translation feature. Last year, it added Conversation mode in the Google Translate app, making the real-time translation feature accessible for everyone. However, it’s important to keep in mind that the technology is still improving. In my experience, the translation works best when slangs are not used and complete sentences are formed. For instance, when I asked ‘kem cho?’ in Gujarati, it translated that to ‘why are you’ and suggests ‘hello’ as an alternative translation. But when I asked, ‘tamey kem cho’ it translated accurately to ‘how are you?’

Before 2018, I relied on text-based translations for communication. However, improved voice translations come in handy in a situation when the other person can’t read. But there is a caveat — many users have pointed out audio translations might be perceived as rude. To avoid this, one can always give a heads-up by ‘Mistranslations in the audio play are unintentional’. And don’t worry, the app translates it accurately.

Leave a Comment

Your email address will not be published.

You may also like

Hot News