Technology Takeaways from GALA 2017

Under News | Posted by Richard Ochab

At GALA 2017 (left to right): Bill Rivers, JNCL Executive Director; Jesper Sandberg, GALA Chairman; and Jiri Stejskal, CETRA President & CEO.

CETRA was once again represented at the 9th annual conference of the Globalization and Localization Association (GALA), which took place in Amsterdam March 26-29, 2017. This was our eighth appearance at what has become the premier event of the language industry, launched in Cancun in 2009 and alternating locations between Europe and the US ever since. Of great interest among the variety of topics were new advances in translation and interpretation technologies.

Neural Machine Translation – A Black Box

In his talk, John Tinsley of Dublin-based Iconic Translation Machines presented Neural Machine Translation as having brought on a “paradigm shift in computational linguistics.” While it is not quite commercially ready, what is quite fascinating about Neural Machine Translation is that it acts like a black box and the developers themselves are not quite sure how it works. In contrast to the now firmly-established Statistical Machine Translation, it is language-independent and cannot be customized for a specific domain.

A takeaway from this session was the observation that we are no longer stuck in the debate over “man vs. machine” (computers replacing human translators), but have moved on to considering the “man and machine” combination, acknowledging that machine translation is complementary to human translation. A secondary point is that the traditional comparison of machine translation with human translation (a popular source of mirth showing ridiculous machine-made mistranslations) is essentially flawed, because the quality of human translation also varies significantly among translators.

So what does it mean for the future of machine translation? Again according to Tinsley, in the next few years we will see hybrid systems of ruled-based, statistical, and neural machine translation engines, but in the long term there is the possibility of developing one giant neural network which will include all thinkable language combinations and cover all thinkable domains.

Interpretation Technology Solutions – An Alphabet Soup

Hélène Pielmeier of Common Sense Advisory delivered her presentation remotely via interpreting delivery platform and provided an overview of some old and some new typology for interpretation technology solutions. Here are the most common abbreviations:

IMS: Interpreting Management System designed to schedule and manage interpreting assignments.

IDP: Interpreting Delivery Platform supporting delivery of spoken word language services.

OPI: Over-the-Phone Interpreting using call center technology to provide audio remotely (but no longer exclusively over the phone as a variety of mobile apps have entered the OPI market).

VRI: Video Remote Interpreting providing both audio and visual rendition of both sign and spoken languages.

RSI: Remote Simultaneous Interpreting for audio only or audiovisual simultaneous interpreting.

MI: Machine Interpreting providing interpretation services without human intervention.

The title of Hélène’s presentation – “A Cure without a Disease? Sustainability in Interpreting Technology Solutions” – indicates that some of the new technologies were developed with a “build it and they will come” attitude and only time will show which of these will succeed. For example, some technologies enable the user to see the interpreter on video which adds a layer of complication but typically does not bring any value to the process. Similarly to the situation with translation technologies supplementing the work of human translators, traditional on-site interpretation which currently represents about 70% of the $10 billion market is not being replaced but the demand is actually growing, albeit at a slower pace than the technology solutions described above, with VRI showing the fastest growth.

Blog Author:
Jiri StejskalJiri Stejskal
President & CEO