BIA is committed to ensuring ongoing quality service during the COVID-19 pandemic.View

Foreign Language Document Review 101: Machine Translation or Not?

Foreign Language Document Review 101: Machine Translation

Should I Use Machine Translation for Foreign Language Document Review?

The scenario:  You’ve discovered that your new foreign language document review (FLDR) project involves tens of thousands of docs that are not in English. You’ve determined that the language is Japanese, informed your client, and gotten approval to proceed. Now what?

You might want to outsource altogether. (In a future blog post, we’ll hear from a Partner in a New York firm who entrusted BIA with one FLDR project and never looked back.) But if you are reading this blog, chances are you’re planning to manage it yourself. If you are cognizant of things like time and money, and if you have previous managed document review experience using AI or Technology Assisted Review (TAR) tools, one of your first questions might be:

Can I use online or automated translation tools to review the Japanese documents in my foreign language document review? Even if I can, should I?

The short answer is No.

Where Machine Translation Falls Short and Fails Long

Let’s cut-and-paste the Japanese sample text from our Foreign Language Document Review 101: Getting Started post that first revealed you were in FLDR territory.

ドック・レビューときたらやっぱりBIAが最高。

Here’s what you get when you run this text through some everyday tools:

A) Google Search: Ads for a USB dock on Amazon
(Is this case about USB docks?)

B) Bing Search: Film reviews for the movie Slumdog Millionaire.
(Weird. But okay, “dock” does sound a bit like “dog.”)

C) Google Translate or Microsoft Translator: “When it comes to dock reviews, BIA is the best.”
(Is BIA a boat company??)

D) Jisho.org (a trusted online machine translation app for Japanese-English): Ten possible options for the first word in the sentence, including dock (maritime), heavy heartbeat thumping (tick-dock), Boston Celtics coach Doc Rivers, and Otis Redding’s 1967 hit song “Sittin’ On The Dock of the Bay.”  
(Good song. Wait, what was I searching for again?)

Get the picture? Accuracy-wise, option C) hit the closest mark by far, but even that is slightly incorrect.

So, what’s the problem? 

A Mini Lesson on Japanese Word Borrowing

The Japanese language evolved from Chinese, eventually became its own entity, and later even contributed to the development of Korean. Japanese also borrows from many other languages, and it is famous for adopting other languages’ words to fit into particular social, cultural, and linguistic contexts. For example:

  • The Japanese word for “part-time job” is arubaito (from the German arbeit for “work”).
  • “Energy” is enerugī, but only if you are talking about the physical energy you get from a good night’s sleep and an hour on the treadmill. In Physics class, energy is always some variation on chikara, riki, or ryoku (力), which is also the term for “force” or “strength.”
  • Then there are the phrases made by splicing and dicing foreign words, like amefuto (“American football”), masukomi (“mass communications”), or my new favorite, arafō (“around forty”)—an acceptable answer to give when someone asks your age and you don’t want to be too precise (maybe you’re closer to arafī ..).

It so happens that for the phrase “document review,” Japanese directly borrows the English term doc review.  But on paper, the ways you would type the Japanese words for doc and dock are exactly the same.  This is the sort of contextual usage an experienced human translator would grasp right away, but machine translation technology can’t handle this yet. 

There, all that to get as far as Word #1 in our example sentence, dokku (ドック=“doc” or “dock”). 
Surface, meet Scratch. Trust me, it only gets more complicated.

If you were planning to just run all the Japanese text from your foreign language document review project through a machine translation tool, we’ve got news for you: the results will be inconsistent, nonsensical, and–worst of all–of little to no good use in your case.

For Foreign Language Document Review, Machines Just Don’t Cut It

Don’t get me wrong. For English document review at BIA, we love our TAR tools and analytics. The experts on BIA’s in-house managed review team use them, write about them, and teach our clients about their benefits every day.

But for foreign language document review projects, especially those involving APAC (or Asia-Pacific) languages, not only do the target languages have too little in common with English structurally, but the subtle nuances, cultural niceties, and unwritten inferences that you’ll find—even in a giant pile of corporate emails—are far beyond what a machine can interpret with the level of accuracy you or your client will need for your case.

Even with the most sophisticated machine translation tools, it’s important to keep in mind that the machine is not trained to give you an English version of information that is useful to your legal case. It is only trained to spit out a translation of the exact text entered, based on the algorithm used to design the software. (For much more on this topic and an excellent breakdown of foreign language document review, I recommend LLM Legal Review’s What Litigators & In-House Counsel Should Know About Foreign Language Document Review by Remu Ogaki, Esq. of The CJK Group.)

Another example: There was a period in my life as a doc reviewer (and frequent Japanese translator) when my least favorite word in the entire world was the Japanese word taiō (対応), pronounced Thai (as in “pad –”) and Oh (as in “Sandra –”). Taiō can mean support; handle; correspond; deal with; cope with; interact with; or often a conveniently vague combination of two or more of those options.

In a typical Japanese corporate email, it is very common for taiō to appear at both the beginning and the end, but with very different meanings in each instance. More specifically, the exact same word could be used to express “Thanks for getting back to me” and “I need you to deal with this.” For the purpose of doc review, this sort of nuance is something that only human translators with Japanese business language experience would be able to convey. (Here again, I highly recommend Ogaki’s article referenced above.)

Foreign Language Document Review: Save Time and Money—Use Smart Humans

Sure, machine translation has its benefits. The main benefit is that it’s extremely cheap. But as we all know, sometimes you get what you pay for. I once worked a project in New York where outside counsel had opted to have all the Japanese documents run through machine translation first, before onboarding a team of 100 Japanese document reviewers. I’m sure they thought they were saving money. However, the machine translations were so unintelligible that our first three weeks as reviewers were spent perusing and re-writing those “translated” docs from machine English gibberish to human English. At 65 bucks an hour, 60 hours a week (deposition deadline meant OT was approved immediately), and 100 reviewers on the team, that client spent over a million dollars in the first month of the project while we re-translated the translation.

From tips to tolls, laundry to lattes, it feels as though life is only getting more automated and more mechanical. It’s not that easy anymore to list areas in which the human brain far outperforms machines, but foreign language document review is on that list, especially when we’re talking APAC languages. At BIA, we use our teams of vetted foreign language document reviewers who often perform double or triple duty as translators, interpreters, and consultants to our law firm clients. For tips on how we go about choosing those reviewers (e.g. should they be attorneys, linguists, or both?), stay tuned for future blog posts.

A picture might be worth a thousand words, but in foreign language document review, the right humans can help you turn a thousand words into a picture that outside counsel can use to win the case. 

To enlist our smart humans for your foreign language document review, get in touch with BIA today.

Maureen Murchie

Maureen Murchie

Maureen Murchie has over 20 years of experience in Japanese document review, translation, interpreting, and consulting for major law firms and corporate clients. Having grown up in Japan and attended Japanese schools through high school, she first encountered foreign-language doc review as a college student in Texas where she worked an old-school “Bankers Box” review for Dallas firm Fee, Smith, Sharp & Vitullo. In addition to her responsibilities in the Sales & Marketing group at BIA, she has served as a recruiter, reviewer and team lead in the FLDR Division at BIA. Maureen holds degrees from Baylor University and the University of Illinois at Urbana-Champaign.