AI Powered Human Translation Startup Unbabel Uses Data As Differentiator

Published on:

We spoke to Vasco Pedro, CEO of Unbabel, one of Business Insider's "startups to watch in 2017", to find out about their success as well as the current use of AI & NLP in translation.

How would you describe Unbabel in a few words?

Unbabel is a technology company whose goal is to enable anyone in the world to understand and be understood in any language.

AI Powered Human Translation Startup Unbabel Uses Data As Differentiator

Unbabel CEO Vasco Pedro

Unbabel combines human editing & machine learning, how did you come up with this process and what is the benefit of it?

Both me (CEO) and João (CTO) have PhDs in Natural Language Processing (NLP). It was clear from the beginning that neither Machine Translation or Human Translation alone would solve the fundamental problem of a multilingual world. Machine Translation doesn’t have the quality required and Human Translation is not scalable. The benefit of the process is that you have a highly scalable translation with human quality.

 

AI along with deep and machine learning are trends that have been around for a while now, with many startups diving into these topics. What makes Unbabel stand out of the crowd? What is the key differentiator?

Unbabel is vertically integrated, from the base machine translation to the tools for Unbabelers to work on, to the integrations that customers use on a daily basis. One great effect of this is that we have incredible data that no one else does. And in machine learning, data is the big differentiator.

Are you using Unbabel internally? How does that affect the viewpoints in the development team?

We do use it internally and, in fact, we are big believers of dogfooding. It is crucial for the development team to get a users perspective on the products. It has enriched their understanding of the problems and leads to better solutions.

Unbabel has raised $8mln, has been awarded “Most Innovative Startup” (2015), and serves clients like oculus, Trello, and Eventbrite – sounds like an entrepreneur’s dream come true. What is the biggest challenge that the company has faced?

Translation is a gigantic market, but with a thousand possible use cases. One of the biggest challenges for Unbabel was figuring out a scalable business model and maintaining the necessary focus on our engine of growth.

AI and NLP, two of the fields you’re active in, begin to play a bigger role in media as well. Kris Hammond from Narrative Science goes as far as to predict that “A machine will win a Pulitzer one day”, from your experience, how far away is that day and how could media nowadays benefit from AI & NLP?

Media is already benefiting from NLP and AI. The New York Times has had a great NLP team for years. I think we are far away from having a Pulitzer-winning machine though. In my mind, that will require us getting close to Singularity, and I think we are 20+ years away. Until then, humans will be required in the loop.

What is the most memorable moment throughout the history of Unbabel?

I would have to say that getting into Y Combinator would rank pretty high in memorable moments. We were completely jet lagged and getting that acceptance call was just amazing. Though, since we were so tired, it was a pretty short celebration.

What do you do differently at Unbabel culture wise?

We surf. We believe in paying off our emotional debt within the company. Surfing keeps us connected, so once a week we go out at 7 am and get in the water and have a blast. You should never forget to have fun 🙂

What’s one piece of advice you can give to fellow founders for their startup?

Understand your market well, make sure you talk to the people currently working on a solution, talk to your competitors. Then believe in yourself and persevere. Don’t let go until you get to where you want to go.

 

__________

Sharing is caring!