I recently attended a Facebook group live Q&A session called #AskMeAnything. An expert answered members’ questions about machine translation (MT). I wanted to listen and learn from prospective customers. All did not go as planned…
The first few questions showed a huge disparity in members’ perceptions. Answers bounced from question to question. The term MT was being used with five different meanings in five different contexts within one conversation!
- MT as a technological component
- MT as a system
- MT as a process
- MT as the product from a system
- MT as the poor quality product from a system
I was confused! Really! I make software that others call “MT” but I was confused. If I didn’t know what he was talking about, how can someone who’s not on the technology side of this business learn anything?
Then a member asked, “If a client asks whether you use MT, what do you suggest replying?” I broke my silence and asked the obvious question, “What is the definition of ‘machine translation?’ I think in a contract, it’s important to define the terms.” The expert (who owns a small agency) replied, “In an agency context I’m not sure it is a good idea to pick apart legal documents from agencies.” ––– What?!
How can we adhere to a contract if the terminology isn’t clearly defined? But this Q&A session wasn’t negotiating a contract. It was supposed to be an opportunity to learn. We’re in a business where words have meaning (Pssst… really!). I realized I had unintentionally picked a fight. I did my best to quietly exit the session and decided to write this post.
Terminology Is King Of The Hill
Let’s think about this terminology challenge in a context outside of translation services. An internal combustion engine (“motor” for short) is our example. That motor can be engineered to power a lawnmower or a Mercedes or a bus. So, let’s zoom in on a simple lawnmower.
I filled my lawnmower with gas, started the motor and cut the lawn. I really liked the look of my lawn, but dreaded the next task – to rake the cuttings. First, I’ll drive to the grocery in my Mercedes to buy a cold beer and hope I’m not hit by a bus.
Hold on… we have seven terms:
- motor – technological component
- lawnmower – system
- cut – process
- lawn – the product from a system
- clippings – poor quality product
- Mercedes – changed system context altogether
- bus – another system context change
Rewriting our example with the MT uni-term approach to terminology, we have:
I filled my motor with gas, started the motor and motored the motor. I really liked the look of my motor, but dreaded the next task – to rake the motors. First, I’ll drive to the grocery in my motor to buy a cold beer and hope I’m not hit by a motor.
Despite these confusions, I won’t offer definitions as long as prevailing MT environment fields the questions with the same respect as Donald Trump fielding fake news questions and I must defend myself like Bill Clinton did with “It depends upon what the meaning of the word ‘is’ is.”
Instead, I offer this presumed meaning from the 1966 ALPAC report (page 19) that seems to be the prevailing definition among MT experts:
“Machine Translation” presumably means going by algorithm from machine-readable source text to useful target text, without recourse to human translation or editing.1966 ALPAC report (page 19)
Let’s rephrase that without the government-speak:
MT is software that automatically converts source language text into useful target language text without human translators.
Not Machine Translation
We don’t refer to a lawnmower as a motor; a lawnmower has a motor.
We don’t refer to Microsoft Word as a spelling checker; MS Word has a spelling checker.
We don’t refer to memoQ as machine translation; memoQ has segment assembly (which is an embedded MT component).
Is it ignorance or confusion that drives us to refer to some applications as MT when it’s simply an embedded technology? Yet CAT tool vendors (like memoQ and Trados Studio) do not use the term MT when they add features that convert source language text into useful target language text. Is this because many translation services contracts prohibit the use of MT and don’t want to suffer from the ban?
Like it or not, we live in the Trump era where the act of asking a simple question provokes indignation and rebuke. Our closer look at the prevailing meaning of MT makes it easy to understand why experts avoid discussing the subject in the company of professional translators. Until such time we can agree on a meaning of “machine translation” and we can openly discuss MT as professionals, I maintain that “Quite simply, Slate is not machine translation.” Semantics? Yes. We’re all in the business of semantics!