AI and neural networks for digital female translators

by Silvia Giancola

As part of my continued commitment to professional development, I recently attended a webinar entitled “Demystifying AI and NMT for Language Professionals”, organized by Women in Localization Eastern Canada.

Women in Localization (WL) is a global non-profit community set up in California, with thousands of female professionals signed up. It’s divided into regional divisions known as chapters. I’m part of the Italian chapter, as are many of my colleagues.

The aim of WL is to develop a global community capable of promoting professional equality in the translation and localization sector, as well as supporting women through all stages of their careers.

The organization pursues this goal through webinars, such as the one I attended, which was on a very relevant topic: artificial intelligence (AI) and neural machine translation (NMT).

Victoria Pratt (SDL) led us through the world of AI, beginning with the definitions provided by Google, Merriam-Webster and IBM for the concept.

According to Google, AI refers to systems of predictive analysis which are potentially applicable to any field.

The Merriam-Webster dictionary defines it as a branch of computer science dealing with the simulation of intelligent behaviour in computers, or the capability of a machine to imitate intelligent human behaviour.

IBM, meanwhile, simply defines it as “augmented intelligence”.

AI is undoubtedly a huge concept, in large part represented by machine learning. It also includes an array of subdomains, including neural networks, which are the key component in machine translation.

Neural networks are algorithms that can detect links and create connections and models that can then be applied to complex problems, such as translation between one language and another.

To be successful, these models need to prove useful – for things such as by saving time, generating more information, integrating into consolidated processes or being easy to use. Ultimately, they need to boost productivity – otherwise they’re not the right choice.

The secret recipe to making this type of model useful requires three ingredients: data (which must be good quality, rapidly accessible and in plentiful supply), science (you need state-of-the-art technology and scientists and analysts who know the subject matter well) and a mature, informed localization industry (that can understand the role neural networks can play and is aware of the needs of the specialists involved).

AI that augments human translation

Jane Hendricks (SDL), the second speaker at the event, quoted the consultancy firm Gartner, according to which “advances in AI are offering enterprises new opportunities to reduce costs and improve the availability of translation services. In many cases, these augment the activities of human translators”. This was the first time I’d seen something like this put down in black and white. And I have to say I agreed with every word.

I loved the definition of human translators as “data stewards”. I think portraying linguists as custodians of data – and thus of words and language itself – is both fascinating and powerful.

We might not always feel like it, but human translators have a fundamental role to play in this revolution, because improvements to Machine Translation (MT) can only be ensured through continuous, detailed feedback from the translators and post-editors who are constantly working with, improving, supplementing and enriching machine output. This allows the industry to create specialist translation engines for all sectors and language combinations.

Translation engines require constant input to learn how to translate better and to deal with factors such as gender bias.

Just as the industry was revolutionized 20 years ago with the advent of the first Computer-Assisted Translation (CAT) tools, the sector is now embarking on another adventure. I can remember the fears of some older colleagues back then, many of whom felt uneasy – some even terrified – in the face of this radical change. The vast majority weren’t ready for it.

My hope is that today, translators – whatever their age – have the necessary soft skills to adapt and respond to these latest changes. Change is the rule nowadays, not the exception, and those who remain stuck in their ways are sadly destined to become obsolete in a hurry.

Way2Global strives to be one step ahead of the latest trends and innovation and has been investing in researching and developing the best translation technology for many years, in addition to focusing heavily on developing its staff and collaborators – as any good B Corp should.

Sure in the conviction that new technology represents an extension of human capability – not a replacement – Way2Global continues to promote Language Industry 4.0, a significant movement that aims to turn AI and digital in general into drivers for social impact, female empowerment and new responsible business models.

By creating an integrated, advanced, cohesive system, we can deliver benefits for all and ensure that innovation really makes sense.