NLP to Boost Artificial Intelligence

Lately, we are being bombarded by news regarding hands-free speakers that users can control with their voice. All the major players in the market had launched their own over the past months, the last one Apple with its HomePod.

In the beginning, it was all about human-to-human interaction, with the rise and massive growth of Social Media. Then Internet of Things enabling communication between machines became the trend. But lately, thanks to the push of some companies in the field of Artificial Intelligence, it’s time for human-to-machine communication. Hands-free speakers, equipped with virtual assistants, are a good example.

However, virtual assistants like chatbots still have a long way to go to understand user requests completely, and therefore to be profitable.

During WWDC 2017 Apple announced the launch of Core ML, a Machine Learning API to make AI faster on all your devices, including hands-free speakers. This announcement is great news for the industry, however, in the NLP part, all the services seem to be in beta.

NLP is the right tool to use if you want your software applications to process human language, no matter if it’s text or voice. However, inside NLP there are different services to use depending on how capable we want our app to be. 

Let's see a few examples:

We have previously talked in this blog about the importance of lemmatization, and the benefits it provides compared to stemming. Particularly in highly-inflected languages such as French or Spanish in which ambiguous words are quite common. You can see a few examples below:

lemma fraces.png

However, how can we determine which one is the right lemma? Is it “livrer” or “livre”? POS tagging is the complementary tool to use in this case.  Let’s take another example, the Spanish word “cara” has two different meanings “face” and “expensive”.  By using lemmatization we will achieve the following result:

 

screenshot-parser.bitext.com-2017-06-22-16-53-53.png

 

 

 

 

 

 

Which one should we use? Depend on the context, and that is why we need POS Tagging, to know exactly which meaning the user is looking for. Let's take the sentence "Tengo una mancha en la cara"

 

screenshot-parser.bitext.com-2017-06-22-16-57-27.png

As you can see by determining the POS we can extract which is the proper lemma to use. 

There are more NLP tools we can use to make our apps understand better human language, stay tuned because we will expand on them in upcoming posts. For now,  download our lemmatization examples in different languages.

Download our whitepaper

Subscribe to email Updates