Machine Translation Meets Augmented Reality

As with many of you, my Facebook and Twitter friend feeds are littered with overlapping circles of contacts. Some I know well, have worked with in the past, and are great friends personally. Others I have met only once or twice.  Though as a rule I have none of my current work colleagues as Facebook friends. (Sadly, I had to unfriend someone recently as we started working together again, but we both agreed it was better that way.)  Occasionally two circles overlap directly (the time one FB friend was tagged in a photo from another FB friend that I didn’t think knew each other), or sometimes the circles overlap around topics. That is what happened this week.

My former translation services colleagues were buzzing about a new App for iPhone called Word Lens that allegedly allows you to point your phone at a street sign for example and receive an instant translation of what it says. Wow! For anyone who has traveled in a foreign country, what a truly amazing, and useful app. If only it worked (more on that in a moment).

At the same time, folks I met sort of tangentially in life who are closely tracking the rise of Augmented Reality were quick to point out that this was one of the first truly relevant AR Apps that they had seen. For those who are not familiar, AR is a movement to overlay useful information about things you might be looking at through your phone. For example you are walking through Hyde Park in London, come to the “Physical Energy statue,” point your phone at it and using maps and global positioning, on screen details about the statue come up like that it was sculpted in 1904 by GF Watts.

If the hype is true, these kind of Apps will be amazingly helpful for travelers. I say “if,” Apps plural, and “will be” because Word Lens wasn’t the first App to the party, and hopefully isn’t the best.  This article on MSNBC suggests that the translation is pretty bad with some pretty funny examples.  And in the comments (I love the reading the comments!) another firm called LinguaSys touts their App as being way better. Is that just sour grapes and poor marketing? I am not sure, the market will surely decide.

There are bound to be others firms as well, I haven’t done an exhaustive search on the subject. Here are a few obvious firms who could jump into this, and may have already:

>Google themselves could easily ramp this up using the fairly robust Google Translate service.  (pretty good for gisting by the way)

>IBM has spent a lifetime on Machine Translation, seems like they could easily over engineer something faster than you can say “Smarter Planet!”.

>What about Wikipedia? Where are they on mobile?  Couldn’t they crowdsource the top 1000 or so most common street signs/words by language and then write an App that queries that pool?

>The translations agencies could easily get in on this as well.  Lionbridge for one has a massive database of previously translated terms numbering in the billions, seems like they could roll an app out pretty easily.

The list is long I imagine. But on this one I am siding with my Augmented Reality connections. As someone who has stared at the signs in foreign countries in baffled and slightly panicked amusement THIS is a good idea.

Let’s hope it gets real.

International SEO Requires Planning and Discipline

Some time ago at my prior company I wrote about the challenge with Global SEO. Namely, how best to translate the key terms that your users search on to find your web site.  Things really haven’t advanced significantly since then.

However this article from ChiefMarketer.com shows its more valuable than ever.  In it there are some good updated statistics on the value of global search and search engine diversity.

But to truly take advantage of the local opportunity, first make sure your site is visible to the local Google/Search Engines in the countries you operate in.  The easiest way to do this is to be sure that the XML site map that you submit to the Search Engines contains links to your local sites, otherwise you are likely dark. At my prior company we fought with this issue for months, and were stunned at the simplicity of the answer once we stumbled on it.

Once you are showing up, you need to think about what the right terms are in local language. There are numerous examples of how a straight translation of your key English terms misses the market completely in local language.  Unfortunately a key word research tool isn’t necessarily going to tell you that answer. Ideally, you would use the glossary you built in the translation process as the starting point. Then work with your in-country team to validate the “adjacent terms.” These are the local idioms that would be the more popular, and more likely, terms for your buyer to search on. Its hard work, but skip this step at your own peril.

And if you are launching a PPC program in local language, make sure you have completely rebuilt your ad copy in-language. Translating English is going to be practically useless, particularly when considering the character limits of each row.

The good news is if you put the time in up-front, you will immediately see measureable results both in terms of increases in traffic to your site, as well as conversions and revenue.

 

 

 

Is “One Hour Translations” a Game Changer for Global Marketing?

I am fascinated with this interview with the CEO of One Hour Translation on the Global by Design blog. And not just because of our American fixation on speed– Think “1 hour Dry Cleaning” or “10 minute oil changes”, or Pizza that is “delivered in 30 minutes or less or it’s free.”

No, I am fascinated because localization is really hard work, and this company threatens to make it easy. Particularly for the marketing organization at a mid-sized company, this really has the potential to be a game-changer. Here’s why.

I have my own experiences with the translation process both from inside and outside the industry.  I have written about some of them here in this blog. And some of the whitepapers I wrote on the subject live on in my former company’s whitepaper archive. I can summarize the challenges for a global marketing team thusly:

  1. It takes time: Sucking content out of whatever systems you use, getting it out to translation, reviewed, pushed back into your systems, reviewed and published… All adds up, it’s painful
  2. Volume rules, and you don’t have a lot of it. Localization is a volume game, and on the marketing side, volumes can often be low, but the amount of work feels the same if the job is 10,000 words or a million words.
  3. There is never enough budget: You budget $10k, it will cost $15k. If you budget is $15k, it will cost $20k. And the corollary to rule #3 is…
  4. The countries don’t want to pay for it: It’s amazing. When the “bill comes due,” often a team materializes out of thin air to handle the translation in-house. It’s one of those things that falls rapidly out of the budget cycle when the details become real.

What the promise of One Hour Translation does is to streamline the whole process for you. (Full disclosure, I have never used the service so I have no idea if their quality is any good, if their TM process works as advertised etc.) Instead of having to negotiate every job, have a file analyzed for the number of words, check for matches against the TM database, manage the glossary, deal with a project manager as the go-between to the faceless translator, it’s all right there for you. A flat 7 cents a word, send off the job, and back it comes. Interact directly with the translator if needed, adjust quality on the fly. Poof, “so easy a caveman could do it,” with or without an “Easy Button.” And for a marketing team that is rolling out slowly in a few countries, or has just a periodic volume of work, this model will be very attractive.

There are surely a number of problems with this as well, starting with the languages supported, the inability to build a true glossary, reuse, translation memory, and scalability, to say nothing of the costs. But if you are sophisticated enough to manage those things individually, or have such a high volume of content that scalability is a concern, well you are not going to be using this service plain and simple.

For the rest of us, it will be a development worth keeping a close eye on.

 

When Rolling out Global Web Sites, Who Manages Localization Process?

I have already chronicled our challenges managing the translation review process for our global web site roll-out. But another issue that crops up is who is really managing the translation agency and the internal review teams? Large organizations will often have a dedicated localization manager. But most mid-sized and small organizations likely can’t staff that role full-time.

So what then?

In prior companies I worked at, the person managing the web site took it on, and relied on the project manager from the translation agency to make sure it goes smoothly.  Problem with this model is that the agency isn’t much help on the internal side of things, where projects often get stuck.

We were fortunate to find a contractor who could fill that internal project manager. He had been at an agency and could help bridge the gap between what the agency needed and what we needed to accomplish. And since he worked for us, he was able to push the agency when it was required, and yet still partner with the internal folks to keep things on track.

This wound up being a crucial role for us as the scope of the project constantly shifted, and the internal strategy shifted around. If you have someone in the role, good for you, make sure you have the tools in place to support them. If you don’t, make sure that everyone knows what aspect of the project they own.

Launching a Global Web Site, Continued

I am pleased that we were able to launch our first local site this weekend serving the Czech Republic.  If interested, you can see it here: www.ness.com/cz

It was hard work for sure, (as I wrote about earlier this year) and comes out many months later than we first anticipated, but it is still satisfying to see it live. Still to come, the team in the Czech Republic now has to add additional country specific content and make sure it represents their interests in the context of Ness as a whole.

We aren’t done yet either, we are getting set to launch Slovakia, shortly followed by Romania, Hungary, UK and India. Israel is in the works as well.

Though it’s still a work on progress, a few lessons learned so-far:

  1. Deadlines help. At Lionbridge we managed to launch 25 sites in just 4 months. But there we had no choice but to launch all of our sites at once and I vowed to never do it that way again.  But the flip side is that deadlines can help drive towards a completion.  In absence of them, things can linger forever.
  2. Be willing to compromise: Sometimes it’s important to give a little to get a lot. For example, we had envisioned utilizing a single website localization partner for all languages. But it turned out that we had an in-house Czech team that could handle the project.  It worked out just fine.
  3. Be flexible:  If we stuck to our original plan put in place in late 2008, we might never have gotten it launched. We listened to the needs of the local teams and adjusted course along the way. Surely made the whole process much easier.

Stay tuned for future launches in the coming weeks, which represents a flexible compromise brought on by an impending deadline!

In-Country Review is hard

Some of you reading this are not sure what I mean by “In-country Review.” Those of you who do are nodding your head right now. In short, In-Country Review“ is the process companies use to review translated content before going live, either in product software or web content, or whatever.  I am knee deep in my own challenge as we speak, and it’s not fun.

The irony is that at my last company we talked extensively about the bottleneck that often craters large-scale translation projects. I even wrote whitepapers on the topic. In fact, one of them is good enough to still be available online.

So how come we still go t stuck?

Good question. We did follow a number of best practices:

  1. We established a glossary of common terms and went through a translation review and approval process before beginning
  2. We trained the reviewers on what to expect when they got the content
  3. We started translation (and review) well in advance of the end-systems being ready to accept translated content realizing this would take some time)
  4. We reminded the reviewers of their role when we got the content for review
  5. We broke up the content in batches to make it easier and not feel as daunting.
  6. We paused translation on the second batch so we could get comments back from batch 1 and incorporate those prior to sending batch 2
  7. And we delivered the content “side-by-side” to facilitate easy review (no easy task when the language is Hebrew!)

And yet, we are still stuck.  Two things happened to us that happen to everyone I think.

First: Our players changed.  Not surprising in the time from kickoff through development and translation to review, the people involved in-country shifted. One person left, another started, another went out on maternity etc.  As a result, the current reviewers have not “lived the project” from the beginning.

Second: Review is hard. There is a certain math that comes into play here. A good translator can only translate x number of words a day, a good reviewer can only review x number of pages a day. No matter how fast you think you are going, there is an actual maximum speed. And since these are never full-time jobs, the “crush of the now’ always destroys proposed timeframes.

Third: We are all human.  No matter how many times you go through a glossary process, or a training process, or remind people that you are reviewing for accuracy only not for nuance, you always get trapped “finessing” the language.  I know I do it when I get content from my global peers that was translated into English, why shouldn’t I expect them to do the same?

So what to do when you are stuck? Also a good question. Time will tell if we did the right things or not but here are the steps we are taking

  1. Acknowledge that we are stuck (read the above!)
  2. Marshall as many resources a possible to lower the total commitment of each person on the project
  3. Identify “cant miss” deadlines and lay-out the clear project plan for achievement
  4. Escalate as required to establish the delay as a business problem to unstick additional resources, or gain leniency on deadlines.

My last piece of advice on this topic: when starting the project, calculate how much time you think it should take for review and double it. Communicate THAT to the team on the project. But TRIPLE it for your own internal planning processes. After that, hang on because you still might be off by an order of magnitude!

@ajdun