Reimagining the Wheels – Translation by SpeakLike

In this Reimagining the Wheels article, I will spotlight on a cool translation service that I’ve come across call SpeakLike. Here is how the company explains its own technologies (emphasis added),

SpeakLike’s Human Assisted Language brings human understanding to translation this way: You send your text, a human translator checks and corrects the machine-translated text in real-time, then the people you are talking with see it in their own languages, quickly and correctly.

Because SpeakLike seamlessly integrates automated and human translation, you’re able to chat right away, without needing to set up costly, time-consuming conference calls as you do using traditional interpreting services. And SpeakLike finally gives you a real alternative to ‘free’ machine-based translation, which is highly inaccurate and often results in costly mistakes.

With innovative technology backed by the human touch of SpeakLike, you’ll be part of a world where people who speak different languages speak one language: SpeakLike.

I first heard of SpeakLike after its CTO Graham Neumann demonstrating it at Demo Camp Calgary #10. After watching Graham’s demo video, I sensed their web-based translation technology involving human translators can be potentially promising.

So I reached out to Graham and we set up a time this past Monday and chatted extensively about their technologies and some challenges I saw. You can take a look of a brief demo video here.

The following are my brief impressions,

1. I see interesting potentials (which is why I include it in my Reimagining the Wheels series) together with some challenges as well.
2. It is nice to know that SpeakLike has paying customers already and they are happy with the services.
3. Based on a very brief chat, Graham seems to be laying the technological foundation well and making it scalable. Although I haven’t dig deep but I think Graham’s prior experience should lead to a well architected software framework.
4. At the same time, I see human resources (good communities of translators) being may be the more fundamental part to the success of SpeakLike. And I think more work needs to be done in this area.
5. In some sense (this may sound strange coming from someone who is a trained computer scientist), the cool technologies are really nice but others can copy it given time and effort. I would argue that the more durable competitive advantage will be the people in a cohesive and thriving translator communities that SpeakLike will need to work hard in building. And this may be harder to build and maintain than the community of photographers in the case of iStockphoto.

I have been fascinated with English/Chinese written and simultaneous verbal translation since the 80s. To me, correct and usable communication is the key to human exchanges and smooth business transactions. And to translate something well is definitely part “science” and part “art”.

Over the years, I have seen some poorly written English business emails by educated Chinese workers in mainland China. So I think there is a potential market in China (if the price is right and the service is good). The potential is there but the challenge is to tap into these needs and approach it with understanding of the Chinese market.

As I wrote in this August 2006 blog entry, I saw machine translation being provided by the likes of Babel Fish and Google Translate as being inadequate. In fact, if I were less generous, I might argue that these “translations” actually do a disservice unless the users are heavily discounting the validity of the machine translated text like me.

Unfortunately, I doubt that many people are as careful as I am and many may give too much trust to these machine translated text. I heard that SpeakLike uses machine translator as a starting point. While I understand the desire for lower cost, at the same time, I worry about the potentially inaccurate or poor quality results being used as a starting point for SpeakLike human translator. The best chef in the world can’t cook good food from spoiled ingredients. It is hard to judge without more data. It will definitely be fun to dig deeper.

Finally, to be fair with SpeakLike, they are still a very young company and I don’t think the challenges they have are unsolvable. Given time, I look forward to improved and better translation results.

Here is a YouTube video of CEO Sandy explaining SpeakLike at DEMO 2008.

Here is a YouTube video of CTO Graham talking about SpeakLike at DEMO 2008.