NLP in NLP

Working with NLP has helped me admire the fact that the algorithms have advanced a lot further from text parsing to conversational communication. Human languages are naturally chaotic, and NLU plays a big role in making better sense of it. The very same sentence could mean entirely different things in different contexts and it gets a lot murkier when human emotions interplay with the phrases. Let alone NLP, even humans might find it hard to understand the nuances in certain situations - married men will be able to relate to this. No - I was not fighting with my wife while I wrote this, though natural events that occurred in the recent past might have had influences, which is not relevant here. The mere fact is human language and interactions are way more complex than what is “said” and/or “heard”.

Let us take a few examples. When we started learning a language as a child, we never knew that, often, a smile followed by a promise, negates the promise. Eventually, we learned it the hard way after an array of unfulfilled promises. The next time we would look out for that wired smile - Aha, caught you, not going to get away with it this time. We learned the language from our experience - we learned the correlation between communication and the execution of the communication. Communication for us is a combination of words and gestures.

Yes, gestures speak louder than words in some cases and that is precisely why, not only words, posture, expressions, gestures, context - everything is important in communication.

When a boy tells a girl: “I am seriously in love with you”, will she consider it serious, just because he said so? Maybe maybe not, but NLU will. She would consider the other aspects- the past, whether he was looking in her eyes, was he emotional, was he holding her hand (with permission) – a plethora of parameters. How will NLU do all those things, if it has only one input - the sentence? We expect NLU to predict the intent, entities and sentiment, from the very sentence that we just rejected. Yes, most probably NLU will tell her that he loves her more than anything else in the world, with an error rate of 50% and she could ignore the rate (as we often do), because the machine just predicted her life partner. It is AI, it knows better and faster than humans, right?

As I was mentioning earlier, NLU can only understand what we trained it for, with very limited aspects of communication. Now that technology has advanced a lot on image detection, gesture recognition, IOT, NLU, NLG, Speech recognition, so on and so forth, imagine what we can achieve if we can combine all these technologies to help NLU to understand human language/communication better. That way NLU will be able to understand or at least attempt to when someone says “Yes” and nods his head to a “No!”, or even when someone tells "Trust me!", fingers crossed. To cross one's fingers is a hand gesture commonly used to wish for luck. The use of the same gesture is often considered by children as an excuse for telling a white lie. By extension, a similar belief is that crossing one's fingers invalidates a promise being made. These could be difficult propositions even for humans.

Speaking of that, which leads to the most discussed issue in NLP, Sarcasm. The key to detecting sarcasm lays in understanding the characteristics of the person and the facts related to an event, this allows the detection of contradiction. To achieve this we need one more attribute of communication and Knowledge. Knowledge about the person, how he reacts and behaves, and a contextual memory of the experiences.

When I say "Thank you for sending my baggage to Hyderabad.", yes it is a good thing, right?. How about "Thank you for sending my baggage to Hyderabad and flying me to Kolkata.", this additional information makes the sentence "Ironic".

But further more information like "Thankyou for sending my baggage to Hyderabad and flying me to Kolkata. Brilliant service".

Yes, now it is sarcastic and trending on Twitter because NLU couldn't understand the sarcasm. DO you know what was the reply from the Chatbot, "Glad to hear that, keep flying with us.". After that reply I will definitely fly regularly with you, apparently, NLU is not going to understand this statement too, probably will reply "Thank you for your loyalty". Sarcasm is a key problem that needs to be solve else it will affect the brand and business.

NLU alone can't achieve greater goals, need to advance further to integrate Neuro-Linguistic Programming(NLP) in Natural Language Processing(NLP), which will help Machines to understand just like how we understand each other, even better without language barriers.

Yes, I have seen terminator, so I do not recommend giving moving parts to this one even after we integrate Emotional Intelligence. For better tomorrow, Stay safe.