The Fundamentals of Chatbots — Terminology Explained
We talk a lot about chatbots here; why they’re so great, how they can benefit businesses and customers, and why we believe it’s imperative that we embrace advancing AI and conversational-based technology. But not everyone understands these technologies, or the terminology behind them. So if you find yourself scratching your head every time someone talks about ‘NLP’, or reaching for Google when you hear mention of ‘conversational UI’, do not fret. Read our simple guide on the fundamental chatbot lingo, and you’ll be sounding like an expert in no time.
So let’s start with the basics. What is a chatbot?
A chatbot is a computer program that is designed to simulate a conversation. These conversations are either going to be text-based (think instant messaging) or audio-based (virtual assistants like Apple’s Siri and Amazon’s Alexa). How do they work? Well, that would be the magic of AI and NLP, the next two terms we will tackle.
AI or Artificial Intelligence is, as simply as something so complex can be put, a computing process that displays intelligence and behaves similarly to a human. That’s still quite the notion to wrap your head around, so we’ll delve a bit deeper. When it comes to chatbots, the AI you experience is how the chatbot interprets what has been said to it, figuring out what the best response should be. The chatbot makes a decision by analysing the words that form a user’s query. For example, if you asked a hypothetical cinema chatbot ‘what is the price for an adult ticket?’, the bot would utilise AI to recognise the key words ‘price’ and ‘adult ticket’, helping it to deduce that the best response would be information on the pricing of adult tickets. How does it know that? NLP, or Natural Language Processing, a field of AI that’s focused on understanding language. It’s the difference between a chatbot seeing a sentence as a group of random words and being able to break it down to recognise meaning, entities and intent (we’ll talk about those shortly!), and it’s the foundation of any chatbot.
Just to confuse matters, you may see ML or Machine Learning or even Deep Learning in place of AI. AI is the branch of computer science which concerns itself with making computers exhibit human-like intelligence. Machine learning is just one of the techniques that are in use today to train artificially intelligent systems.
Next up, Intents and Entities, are the key parts of a sentence as seen by a chatbot. Intent is all about the requirement pertaining to a question. Going back to our ‘what is the price for an adult ticket?’ example, the intent would be enquiring about ticket prices. The intent is the reason a chatbot is being spoken to, and it is important that a chatbot can accommodate all of the common intents relating to its purpose. For every intent, a bot will have a response, and will use NLP to decide what the intention of a question is. Entities are extra pieces of information in a query that can help to narrow down the scope of an intent. In our example, we’re asking for an adult ticket price. Although the intent relates to ticket pricing, we have specifically asked for an adult ticket, and so we would use adult as an entity. Because we have specified an adult ticket, the response may be different to that given when just asking for general ticket pricing. The entity has provided our system with precise details about the intent, which will be vitally important for responding to the user in a more intelligent manner. Entities are often nouns, but this is not always the case.
Whether you are typing messages to a chatbot or talking with a voice-assistant, you are interacting with a computer via a Conversational User Interface (UI). Although we are all now accustomed to using a mouse or touchpad to click around in office suites or web browsers, or swiping the screens of our mobiles phones, conversation will always be a more natural and intuitive interface for humans. We communicate using speech from a very early age and soon learn how to use it to satisfy our curiosity with the world. The intricacies of language are hardly given a second thought when we ask for food or enquire as to why the sky is blue. But for computers, it has consistently been a syntactic minefield. Words can have one meaning in one particular context and an entirely different meaning in another. But with the recent advances in AI and Natural Language Processing we discussed earlier, it is now easier than ever to create interfaces to computer systems which are immediately instinctual to the untrained user. The only requirement for interacting with a chatbot is the ability to talk or type in our preferred language. Now that the problem of deciphering a user’s intent has largely been solved, we now have a new challenge; creating user experiences closely modelled on natural conversation. This is the aim of those developing conversational user interfaces.
Once a chatbot has been built and is ready to talk the talk, it needs to be trained. Chatbot Training is the process by which a bot improves it’s accuracy at recognising the intention of a user’s input. The most popular method utilised by chabot systems is to use lists of training phrases assigned to intents we expect the user to have. For example, our hypothetical cinema assistant will use training phrases such as ‘how much are tickets?’ and ‘what do you charge?’ as training phrases for the ‘request ticket price’ intent. The bot will then use those phrases as a guide to match a user’s input to the correct intent in the system. The accuracy of a chatbot is directly correlated with the size and variation of the provided training phrases. A smarter chatbot will be able to interpret more complex and detailed questions, so rigorous training is a crucial aspect of any chatbot’s development.
Hopefully you have a clearer understanding of what the most common phrases relating to chatbots mean. The next time you hear someone mention chatbots, you can dazzle them with your newfound knowledge. And the next time you talk to a bot, you can have a greater appreciation for what goes on behind the scenes!
Originally published at Natter.ai.
