Let’s Chat Bot Lingo

As with any topic, delving into an online search expedition to better understand the concept of chatbot technology can leave you feeling tongue-twisted and tangled in a technical term frenzy.

From NLP and NLU to machine learning, deep learning, AI and API, it can all get a little confusing. Enter any of those terms into your browser and you’ll find several definitions that are challenging to fit into one breath.

However, knowing the general meaning and differences between these terms will give you a helpful overview and understanding of how bots get so chatty.

Below are some of the most common terms related to chatbots and what they mean — in non-Programmer friendly speak.

Natural Language

Quite simply, a natural language is any language that has developed over time through verbal or written communication between humans. Natural language is less structured than the deliberately created languages typically used to communicate with and to program machines.

Rather than clicking a series of buttons on a website to search for what’s showing at the movies for example, you could ask a chatbot in a familiar chat scenario — just like you’d send an instant message to someone, by typing, “What’s showing at the movies tonight?”

From Whatsapp to Facebook Messenger or even SMS, instant messaging is something that people — no matter what level of tech-savvy — are very familiar with, making natural language and chatbots very powerful tools to encourage customer engagement with brands.

Natural Language Processing (NLP), Natural Language Understanding (NLU) & Natural Language Generation (NLG)

NLP, NLU and NLG are closely related terms and often get used incorrectly. A helpful way of thinking of it is that a machine must first do some language processing, followed by language understanding before it can then generate a response back to a human.

Natural language processing is the entire end-to-end process that allows humans and computers to speak to each other, naturally. Part of that process though — the most important part — is the natural language understanding (NLU). This is the interpreting and responding to the natural language.

Let’s go back to our movie chatbot — let’s call it CineBot. You would ask CineBot what’s showing that night, CineBot would use algorithms to analyze your question (the technical stuff we’re not covering here today), interpret what you’ve asked and formulate a relevant response, likely a list of movies showing. That entire process is referred to as NLP, the part where it interprets what you’ve said is NLU and the formulation of a well-phrased response is NLG.

Natural language understanding is a very complex part of natural language processing and can be tricky when it comes to interpreting different pronunciations or sentence structures that wouldn’t usually be a problem in human-to-human conversation. The better the NLU, the more accurately a chatbot can interpret and respond to what’s being said to it and ultimately, the more human it will feel.

The one-liners:

NLP is the full end-to-end process of a computer receiving, interpreting and responding to natural language.

NLU refers specifically to the interpretation of natural language.

NLG refers specifically to the formulation of a response.

Machine Learning

This is when a computer is programmed (with all sorts of nifty algorithms) to automatically learn through its experiences and apply what its learned. Over time, it gets to know you and other users’ preferences, your linguistic quirks, previous interactions, etc. — just like a human relationship.

A perfect example of Machine Learning (ML) is how an online store or music-streaming site would make recommendations based on what you and people like you have previously selected or searched.

The one-liner: Computers programmed to automatically learn through experience.

Deep Learning

Deep learning is a modern, sophisticated aspect of machine learning that has been shown to work well for natural language processing. To learn how to react or make predictions, the machine employs a relatively large artificial neural network with many artificial neurons, arranged in many ‘stacked’ layers. The more layers in a network, the ‘deeper’ it is said to be.

The one-liner: Computers using their own ‘deep’ neural system or ‘brain’ to figure out how to react and behave.

Artificial Intelligence (AI)

Artificial intelligence (AI) or machine intelligence, is the perceived ability of machines to acquire and then apply knowledge. Interestingly, as machines acquire the ability to do more tasks, these are often removed from the list of tasks considered to require intelligence. For example, the ability of machines to beat grand masters at games like Chess and more recently, Go. This has lead to a common remark that “AI is whatever machines haven’t done yet.”

Application Programming Interface (API)

An API is the doorway through which one gains access to NLP and NLU functions of companies like Feersum Engine and Google, in order to build a chatbot.

Not too bad is it? Hopefully that helps clear things up a little for you and adds a good boost to your bot knowledge bank.