You know language technologies, and, in particular, machine translation, are in the midst of a hype cycle when Big 4 consulting firm Deloitte makes it a central theme in an artificial intelligence (AI) whitepaper. The 28-page paper published by the Deloitte Center for Government Insights explores how the US government can leverage AI to save on cost.
Deloitte says “AI programs can play games, recognize faces and speech, learn, and make informed decisions” and are improving “at an exponential rate”, which has resulted in everything from “self-driving cars to swarms of autonomous drones, from intelligent robots to stunningly accurate speech translation.” All that awesome tech could save the US government up to USD 41.1bn, the report estimates.
Deloitte proceeds to elevate machine translation to a stand-alone category under cognitive technologies, along with rules-based systems, computer vision, machine learning, robotics, and natural language processing. Typically, machine translation is seen as a sub-category of machine learning and natural language processing.
MT’s Obvious Implications
Deloitte references the recent rapid progress in machine translation by using neural networks, saying that “significant advances have been made in (MT) in only the past year.” So why does machine translation matter, according to Deloitte? The report says “machine translation has obvious implications for international relations, defense, and intelligence, as well as, in our multilingual society, numerous domestic applications.”
To illustrate how the US government can leverage emerging AI technologies in very practical terms, Deloitte offers “four automation choices”: relieve, split up, replace, and augment. As an example of relieve, the report cites the Associated Press’ use of AI to write corporate earnings reports so journalists can focus on in-depth reporting.
“An augmented approach to translation increases productivity and quality while leaving the translator in control of the creative process and responsible for aesthetic judgments”
And example of splitting up is the US Customs and Immigration’s use of chatbots to answer simple questions. The postal service’s use of handwriting recognition to sort mail by ZIP code is advanced as an example for “replace.” For augment, the report cites the deployment of IBM Watson to support oncologists in cancer diagnosis.
Translation Takes the Spotlight
However, Deloitte did not choose oncology or chatbots as a real-life example of AI in action across the four choices. Instead, the consultancy focused “on a single government job, translator, and one cognitive technology: machine translation.”
A relieve approach in translation, according to the report, could “involve automating lower-value, uninteresting work and reassigning professional translators to more challenging material with higher quality standards, such as marketing copy.” Using marketing translation as an example of a field demanding “higher quality standards” seems a little random and many legal, patent, or financial translator would beg to differ.
As an example of splitting up, Deloitte uses post-editing of machine translation, an approach which has been around for more than a decade. The report is more in tune with the industry in this category, correctly pointing out that many professional translators disdain PE and consider it “linguistic janitorial work.”
The replace approach is self-explanatory. The human translator goes away, the machine takes over. As a text type that would lend itself for pure machine translation Deloitte highlights the “technical manual.” Any technical documentation manager at a manufacturer who might incur product liability issues for mistranslations, might object.
Finally, the augment approach. In fact, with technologies such as adaptive machine translation developed by startups such as Lilt (and now also SDL), this approach is widely seen as where human translation is headed over the coming five years.
For this approach, the Deloitte report is spot on: “Translators use automated translation tools to ease some of their tasks, such as suggesting several options for a phrase, but remain free to make choices. This increases productivity and quality while leaving the translator in control of the creative process and responsible for aesthetic judgments.”
While the report takes a very US-government centric approach to quantifying the savings, there is likely also a very receptive audience across the Atlantic in multilingual Europe.