The biggest time saver in bot building since the introduction of AI
This week Chatlayer launched a major upgrade to its platform. The Chatlayer 3.0 release brings many improvements. We now have built-in voice bot functionality, versioning, synonym entities, a code editor, an iFrame module and advanced NLP analytics that tell you where and how you can improve your NLP even before you launch your bot. But 3.0 comes with yet another unique feature: language independent NLP. The biggest time saver in bot building since the introduction of AI.
What’s in it for me?
Language independent NLP makes it a lot easier for you to build bots in more than one language. If you build a bot in for example English and you properly train it’s NLP in that language, it will already understand more than 100 other languages out of the box.
Yes, you read that correctly. Our NLP is actually language independent. Here is an example for our Choo Choo bot. If you have ever followed our tutorial, you know this bot that lets you order train tickets. Since the tutorial is in English, you might think your bot only understands English. But that is no longer the case. Go into the NLP – Test tab and type the following sentence in English: “I want to book a train ticket”.
As you can see in the screenshot above the model is 99% confident that you want to book a train ticket. Now repeat the same for all the other languages you might think of. You can use Google Translate or even better DeepL to give you the correct translations for languages you don’t speak.
- French – “Je veux réserver un billet de train” – 99% confident
- German – “Ich möchte eine Zugfahrkarte buchen” – 95% confident
- Spanish – “Quiero reservar un billete de tren” – 99% confident
- Russian – “Я хочу забронировать билет на поезд” – 99% confident
- Arabic – “أريد حجز تذكرة قطار” -a 99% confident
- Chinese (Simplified) – “我想订火车票”– 98% confident
As you can see from the examples above, once you have trained your NLP in one language, you can instantly use it for other languages. We support over 100 languages. If you are interested, you can find the entire list of supported languages here.
If you have ever built a bot before, you know that training and maintaining the NLP is the part that requires the most effort. And once you make your bot multilingual, this effort is doubled every time you double the amount of languages your bot supports. We say: no more! Thanks to our language independent NLP you need to do this work only once, for one language. For the other languages it just works out of the box. You only need minimal effort to adapt your NLP to very language specific expressions.
If you build a bot in one language you spend 80% of your time building the conversational flows and 20% of your time on training the NLP. Previously you had to redo the effort of training the NLP for each language. Thanks to our language independent NLP this is no longer the case. If you build a bot that supports 5 languages, you no longer spend 5 times the NLP effort. You will save 75 to 80% on the NLP effort or around 40% of the time spent building a bot.
But how does this work?
If you have ever worked with Google Translate you know that the system is not perfect. And even the more powerful DeepL makes weird mistakes from time to time. So how is it possible that our NLP can understand so many languages? And that without specifying the language up front?
The difference with our NLP and a translator is that our engine doesn’t translate into another human language. Our engine translates whatever a user says to a machine language. And our NLP uses this machine language to determine their intent.
What is special about our NLP is that you don’t have to specify which language you are using. Our NLP maps it correctly automatically, whatever the language. So the era of one NLP model for every language is finally over!
A bot that can understand any language is no longer fiction, it is now a reality. This breakthrough is not only interesting from an AI perspective, more importantly, it is the biggest time saver ever introduced in bot platforms since the introduction of AI.
In practice the effort required to build a bot in five languages is reduced by about 40% and the maintenance effort of a bot in five languages is reduced by up to 75%.