He has a PhD in German Studies from Northwestern University in Chicago. After working at a number of universities until 2010, he moved to Hamburg, Germany where he developed a successful academic translation business. Christopher founded skribling.com in 2014 and LanguagePilot GmbH in January 2018.
Here we present bytes from an exclusive interview with Christopher Reid:
Q. What’s your background and what is LanguagePilot?
A. My name is Chris and I am the founder and CEO of LanguagePilot in Germany. We developed an AI-based tool that analyses a text to determine its suitability for “neural” machine translation. Neural machine translation is relatively new and it is a more powerful way to translate text than Google Translate, which people know from the Internet. Basically, our tool helps people save time and money by analysing their text and offering processing recommendations. It tells people how much of their text can be translated accurately by machine translation and how much would then need to be proofread by a human being so that the text sounds like it was written by the native speaker.
Q. You have developed this product. If you are using technology like AI, machine learning or deep learning to train your models, do you think it would be a threat if Google or Microsoft also came up with a similar solution?
A. What is interesting is that we are developing something which shines a light on the black box of machine translation. Google is, of course, interested in quality estimation, but mainly to improve their own machine translation. There is no external instance or outside actor that is actually looking at the quality in an objective way. We are therefore taking an entirely different approach; we are holding the machines translation companies (like Google, Amazon, and DeepL) accountable for the quality of their translations and, in the process, giving people more confidence when using these tools.
Q. Assume that Google is at least interested in improving its machine translation. They have tons of cutting-edge things they can do to capture the market. It’s all about training the model and all about the accuracy. So, assuming Google and others can offer high-quality machine translation, what should I do if I want to use machine translation as a content writer and content aggregator?
A. First of all, we predict the accuracy of a particular machine translation tool with regard to your specific text. We will tell you which machine translation options are best, whether it’s Microsoft’s Bing or DeepL. Our tool tells people their text’s complexity relative to these neural machine translation tools so that they know exactly what to do with it. You can see what is the ideal the machine/human mix for getting the best outcome.
Q. How does this ultimately help content writers and aggregators? How can it allow them to publish in multiple languages and even improve their content, but also save time?
A. Typically, if you want to translate a text, you have to go to a human translator. Your preferred translator, however, may already have too many jobs on their plate and may not be able to translate your text until the following week. On the other hand, with our tool and its processing recommendations, you will feel confident using machine translation on your own. You can then have your texts translated right away. And because you’ll only need someone to proofread your text, it can be ready much faster.
Q. What is your next step with LanguagePilot? What is your larger vision for your application?
A. Above all, we need to continue to train the tool. To do this, we need more data. When we have more data, we can train the tool for virtually hundreds of language combinations and subjects. And, of course, we can improve the tool’s accuracy. Eventually, we want to offer our tool in emerging markets, like India, where machine translation is becoming more and more common.
Indians who are interested in translating a document may not do it because of the expense or the time it takes. Machine translation is a much cheaper and faster option and our tool helps people to take advantage of it. If we can tell someone that 80% of their text can be accurately translated by a machine and also which machine translation tool is best, then it will make the whole process much easier.
Q. Who do you think is the ideal customer for you? And who else might use your tool?
A. The ideal customer right now is anyone who is already using machine translation. But, more specifically, there are freelance translators and translation agencies that use machine translation APIs. Every day they experience the pain of not knowing how well their customer’s text will be translated by a machine-translation engine. As a result, they have to make a decision under great uncertainty. An agency might ask: “Can we use machine translation here or do we need to outsource the text to a human translator?” The answer can have a big impact on their pricing and deadlines.
Christopher Reid and LanguagePilot have developed the TCA (text complexity assessment tool) to solve the problem of predicting machine-translation quality. The company aims to become a world leader in language technology and services. We wish him all the best for a successful launch.