Real time machine translation: The present and future

An inappropriately titled new story and live demonstration nevertheless showcases how far technology has evolved when it comes to machine translation guided by context and nuance as much as words themselves.

The first time I saw a technology demonstration in this regard was at Strong Angel III in San Diego, in 2006. The real time translation of broadcast video was at the time, and frankly, even now, incredible stuff.

Now, Microsoft is demonstrating real time translation of voice calls made over its Exchange Server platform. So while this is NOT a translating phone as the news story on Download Squad and the technology demonstration itself tries to make out, it is nevertheless extremely impressive.

The applications go well beyond the corporate world where this technology will be first deployed. Let’s be clear, significant perils of machine translation are well known and unlikely to go away for the next few years. And Microsoft is not the only one experimenting with this cutting edge technology. However, even the sophistication of the translation demonstrated by Microsoft in this video was unimaginable a few years ago.

But then again, didn’t DARPA – that secretive agency in the US – promise us by this year “software that can almost instantly translate Arabic and Mandarin Chinese with 90 to 95 percent accuracy“?

Implications for humanitarian aid and peacebuilding?
There are of course several exciting possibilities for the application of this technology in the domains of humanitarian aid and peacebuilding. If the inaccuracy of the translations are accepted by all parties, these platforms even without 100% accuracy become valuable tools in cross-cultural conflict resolution, conflict transformation and post-disaster, to strengthen interactions between foreign and local aid workers as well as victims on the ground.

A very different approach by Skype to almost real time translation years ago failed, but as a model still interests me. Could this model be revisited for example in situations like Chile and Haiti this year?

The other impeding factor is the lock-in to Microsoft Exchange Server infrastructure, well beyond the price point and sustainability for NGOs that matter most in disaster response and peacebuilding. Can we envision a service model where this is moved to the cloud and on-demand, can connect entities together? An associated fee is easier to find funding for, and without any server infrastructure or complicated software to set up, this then becomes a more helpful service.

Microsoft’s video is interesting for another key feature I’ve not seen yet on any comparable system. Machine retranslation, in this case from English to German and then to English on one end of the conversation, all in real time. This allows the speaker to judge how well his statement has been translated, helping to address false impressions created by erroneous translations.

I wonder when we will really move into real time translation on mobiles? Today, we are playing Quake III on mobiles. When I was growing up, Quake III required a pretty serious gaming rig.

25 years hence, will it really matter if we are really monolingual? By then, would multilingual social and political interactions, and even person to person conversations, be mediated by some form of mobile, real time, contextual translation?

One thought on “Real time machine translation: The present and future

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s