Chatbots are the cutting edge of digital customer services and there’s a good chance you’ve already had a few interactions with them, perhaps even without knowing. The difference between a chatbot and a standard web interface is that instead of simply pressing buttons and ticking boxes to find your way to specific content, information or contact details, you are having a conversation in a chat window with a non-sentient (so far!) AI.
There are all sorts of different chatbots that work in different ways, from the simple to the super-sophisticated, and they are becoming more advanced all the time. It is predicted that by 2020, around 85% of all chat interactions on the web will be with non-human agents!
How does a chatbot work?
The simplest form of the bot is rule-based and is simply a pre-programmed piece of software designed to respond to pre-set options (buttons or selection boxes) with predefined answers. It’s a very simple device that can be extremely useful for handling a lot of simple support questions or perhaps guide users towards product categories or human contacts as required. Understanding the different varieties helps with understanding how chatbots work and how they operate.
More advanced Chatbots use machine learning to improve their support capabilities. Machine learning is a computer’s ability to analyze inputs, data and patterns of behavior to actually learn by itself without the help of a human programmer (or with as little human input as possible). Machine learning bots will develop learned responses to questions depending on previous experiences. Certain trigger words or phrases will start to correlate with specific answers and responses and the bot will eventually decide for itself how best to respond to a human user.
NLP (natural language processing) is how chatbots process human questions and requests for help or support. Used by web or chat window-based bots and by speech activated chatbots (like Amazon’s Alexa or Google’s Assistant or Microsoft’s Cortana), NLP allows the bot to interpret the more intrinsically human aspects of language such as intent and context. By breaking a question or sentence down into distinct elements, a bot can infer the intended meaning and provide a more helpful answer. Allied to machine learning processes this should, in time, lead to a more experienced and, therefore, more helpful digital assistant.
In NLP processes, phrases are broken down into constituent parts:
The Intent: What is the user trying to do? What is the problem they are trying to solve?
The Entity: What details are associated with the intent?
For example, with the question: Hey chatbot, what is the time in London right now?
The Intent is “time” and the Entities are London” and “right now”. With this structure, the chatbot can infer what the user would like to know and answer appropriately.
Where are you likely to find chatbots?
Increasingly, chatbot platforms like Facebook Messenger and WhatsApp, web apps and even voice-activated devices such as Alexa or Google’s assistant either in household devices or on a mobile phone are all typical platforms for chatbots. We may start to see them turning up in doctor’s waiting rooms, high street shops and perhaps pouring our drinks at the bar in the future. The combination of low cost, efficient help allied to a friendly conversational interface is proving to be a successful and compelling combination.