What’s the State of the Language Art

A new batch of machine translation tools powered by artificial intelligence are already translating millions of messages daily. Proprietary ML translation solutions from Google, Microsoft and Amazon are in daily use. Facebook takes its path with an open-source approach. What works best for translating software, documentation, and natural language content? And where’s driving the automation of AI-powered neural networks?

development of neural machine translation
William Mamane, head of digital marketing at Tomades, a professional language services agency, was skeptical about machine translation. “Our company has been around for 12 years, with over 50,000 business clients. We have supported the value of ‘human translation’ and still do.

However, we have seen steady progress in the quality of machine translation. Currently, machine translation does not compete with a good mother tongue linguist, but there is still a place for AI and machine translation in the value chain of translation services.”

To trace this evolution, go back to the origins of AI as applied to machine translation. At a basic level, MT uses algorithms to convert words in one language to another. This proves to be insufficient to translate successfully.

An understanding of complete phrases is essential for both the source and target languages. We can understand MT as decoding the source language and recording its meaning in the target language.

There are various approaches to solving this challenge, some involving applying statistics to choose the best translation for a given phrase. Others apply structured rules to select the most likely meaning. But in complex language forms such as fiction or other types of literature, even the best machine translation engines don’t seem natural.

Machines do better with language structured for specific uses. These include weather reports, financial reports, government protocols, legal documents, sports results. Language and idioms are limited in these cases. There are formulaic linguistic structures and formats.

from algorithm to system
Here machine translation is already in daily use. Even after so much thought, that doesn’t diminish the need for human beings to be editors and proofreaders. They need to identify proper names, resolve ambiguities, and understand idioms. However, supervisory, editorial, or audit roles are less demanding and less time-consuming than translation.

On the web, automatic translation began in the 1990s with Xerox’s Systran and AltaVista’s Babelfish. Both used statistical methods and rules to translate the short text. The popularity of both was shocking. In 1996, AltaVista reported that BabelFish made half a million requests a day.

Even in 2012, Google processed translations that filled a million books per day. And that was before the translation revolution that took place in the last five years. More information on the early history of MT is here.

neural machine translation
Neural Machine Translation (NMT) uses artificially produced neural networks. This deep learning technique, when translating, looks at whole sentences, not just individual words. Neural networks require a fraction of the memory required by statistical methods. They work much faster.

Deep learning or artificial intelligence applications for translation first appeared in speech recognition in the 1990s. The first scientific paper on using neural networks in machine translation appeared in 2014. Many advances were made in the field after this article.

An NMT system first appeared in 2015 at OpenMT, a machine translation competition. Since then, the contests have been filled almost exclusively with the NMT tool.

The latest NMT approaches are called a bidirectional recurrent neural network, or RNN. These networks connect an encoder that produces a source sentence for a second RNN, called a decoder. A decoder predicts the words that should appear in the target language. Google uses this approach in NMT which operates Google Translate.

Microsoft uses RNN in Microsoft Translator and Skype Translator. Both aim to realize the long-held dream of simultaneous translation by Harvard’s NLP group, which recently released an open-source neural machine translation system, OpenNMT. Facebook is involved in extensive experiments with open source NMT, learning from the language of its users.

Google Translate is a free multilingual machine translation service developed by Google to translate text.

Google Translate is a free multilingual machine translation service developed by Google to translate text. It provides a website interface, mobile apps for Android and iOS. Its API helps developers to build browser extensions and software applications. Google Translate supports over 100 languages, living and dead.

It was serving over 500 million people daily as of May 2017. As of 2018, it was translating over 100 billion words per day.

Leave a Comment