September 20, 2018 | Written by: Danny May
Categorized: AI/Watson | Consumer Products
Share this post:
While on a business trip to China, my passport was stolen. I wasn’t sure what to do, so I approached a policeman for help.
The poor performance of the translation app I downloaded led to the policeman going for his gun and my colleague running away in fear. Instead of saying “hello, how are you”, I said “hello, I love you.”
I knew I couldn’t be the only person who had had problems while traveling because I didn’t speak the local language. Something better needed to be done.
Enabling better beginnings with Watson APIs
Before I founded Lingmo, I was a plumber. After my close call in China, I spent a lot of time studying everything I could about translation services, AI and machine learning.
In my research, I found that the two biggest problems with translation services were accuracy and latency time. I knew AI could help with both. So, we sold our house and I created Lingmo.
My team and I began by using our own machine learning, algorithms and data—with terrible results. It took roughly 50 seconds for a translation.
That’s when we turned to IBM. Once we starting working with Watson Language and Speech APIs, we found that system training went 50% faster. We also increased accuracy, including with dialects and slang. The more data we entered, the better the results.
One of our key differentiators that helps with accuracy is that we don’t do word for word translation because this often leads to unintelligible results. Instead, we use AI for speech recognition that puts what’s said into linguistic context. This cuts down the possibility of error tremendously.
Expanding services and offerings
In June 2017, we launched a device called Translate One2One, which is the first independent translation interface. This would have been impossible to put into devices had we not addressed accuracy and latency time issues with AI.
Talk2You, which we released in April 2018, is the world’s first for instant translation voice messaging. It works like WeChat, WhatsApp or other instant messaging platforms where you can leave a voice message for someone, but ours also enables translation from one language to another through the process.
These types of services are real game-changers for CEOs who, for example, can now instantly deploy messages to teams across the world.
In addition, we custom build our code to suit large organizations like airlines, where specific terminology is needed.
Making a difference, one language at a time
Our systems are being used in more than 36 countries. At the moment we offer translation in ten languages: Arabic, Chinese, English (U.S. and U.K. dialects), Italian, Japanese, Spanish, Brazilian, Portuguese and Korean. This covers roughly 85% of the world’s population.
We launched Lingmo because we wanted to empower people—whether you are one of the most well-known CEOs in the world, or if you’re someone who just wants to go on a once-in-a-lifetime trip. Whether for work or leisure, removing linguistic barriers is our goal.
For more details on Lingmo, read the case study.
Listen to Danny May discuss Lingmo’s goal of empowering people by removing linguistic barriers: