Share on facebook
Facebook
Share on twitter
Twitter
Share on linkedin
LinkedIn

In part 1 of this series, we introduced some core concepts of AI. In this post we look at how NLP (Natural Language Processing) is applied today.  

Speech processing is the when machine receives the vibrations of sound waves, converts them into numbers that are then transformed into words.

Exciting as this may be for speech to text transcription, the real action is when all these words are represented as numbers and vectors of numbers.  A vector can be thought of in the following way. Any single word can be described or defined using a unique combination of words and that combination of words is its vector.  And when we can uniquely describe a word and its place in a sentence, many possibilities open up.  Suddenly we understand a word and we can guess its function.  When you ask a smart speaker to book an appointment, the NLP in the background identifies which words refer to the time and the place and accordingly puts them into the appropriate slot in your diary. 

At present, as powerful as it is, NLP can really be thought of as a way for a human being to access some information.  It is not a conversation in any real human (or perhaps animal) sense.  At present we can envision a near future where smart factory’s and workplaces’ embedded speakers allow more efficient location of items or perhaps a technical manual that is available on-call via headset.  Perhaps in the future it can help us understand the complex insurance and credit card applications! 

So, when we think of NLP it is best to understand it as a program that allow us to access and direct information – i.e.. “put this there or bring me this”.  And because it is a machine, it can find what you want from hundreds of thousands of documents or websites with speed and accuracy. 

Want to go further, read more on the current state of NLP here

Ready to Learn more?

Get in touch with our team of experts to find out how we can help you drive ROI through our AI solutions

Share this: