by Chatbot Concierge | Apr 24, 2018
What is a chatbot?
A chatbot is a program that performs automated tasks and can hold a conversation with a human.
Anyone can ask a chatbot to reserve a hotel room or to help with directions to the best coffee house in town. The bot responds with a selection of room choices or directions to the secret, speakeasy-like café with killer espresso.
Cool, right?
How do chatbots work?
There are two kinds of chatbot applications.
Rule-based chatbots - can are programmed to respond to certain commands.
These bots get smarter if programmers update them.Artificial Intelligence (AI) and Machine Learning (ML) chatbots - attempt to understand and reply using natural language.
These bots get smarter as they have more and more conversations.
Rule-based chatbots are scripted and have a very clear structure to the conversations they can have. They can provide the right answers to specific questions or instructions.
People who interact with these bots must be taught how to use them. This is usually handled with up front onboarding instructions, either from the bot itself or provided by the programmers.
An employee engagement bot likely understands different ways to say “kudos” or “well done”, but it may not know them all.
For example, you could say “kudos to Jason for his work on our last sales call” - the bot will understand that you want to thank Jason for his efforts.
If you were to say “three cheers to Jason for his work…” the bot may not know what you want it to do. Yell three times?
So, scripted chatbots are only as smart as the programming they’ve been given.
That doesn’t mean scripted chatbots are stupid, though. Remember that they have a very specific purpose and can be updated and trained continuously.
But there is another way.
AI chatbots are able to understand a wider range of natural language and questions.
Using the employee engagement bot from above, if that same bot were built using AI and Natural Language Processing (NLP) it would figure out that “three cheers” is another way of saying “well done” and respond properly.
A bot using Machine Learning (ML) will learn from that interaction. The next time it is presented with that same input, the bot will know precisely how to handle it.
What is Natural Language Processing?
Human language is messy.
We may not always finish our thoughts. Texting has led us to use half-sentences and halting instructions.
That’s where NLP comes to a programmer’s rescue. Imagine trying to brainstorm each and every way someone could possible respond to a question or give an instruction.
Starbucks baristas train customers how to order their coffee correctly the next time because there are a million and one ways to order your orange mocha frappuccino.
Longest drink that can be inserted into the Starbucks ordering system.
Instead of training chatbot users how to ask questions the right way, NLP helps the bot understand and respond to different ways of saying the same thing.
An NLP engine like Dialogflow works to identify the context, intent and entities in a user’s input.
These are the relevant bits of information provided by a user when they type or speak an instruction. The NLP engine uses a number of libraries to match a person’s input to words and speech patterns it already knows.
The NLP engine then returns an answer based on what it understood from the instructions.
For more details on NLP, check out this article from Neospeech here.
Bots are built with a purpose
Assistant bots like Amazon's Alexa or Apple's Siri have a very broad mandate. That is to be as helpful to as many users as possible, performing lots of tasks reasonably well.
A hotel will likely build a bot to help guests book a room or participate in loyalty programs.
A orthodontist may create a bot that can answer common questions and pre-qualify patients for specific treatments.
The Chatbot Concierge bot shows off a bit of what’s possible for hotels looking to get more direct bookings and offer guests a unique way to learn more about the property and the area.
Do I need to be a scientist to use AI?
Bottom line: no. You don't need to be an expert at artificial intelligence (AI) to get started making amazing conversational experiences built with AI.
Tools like Dialogflow and LUIS have made it relatively easy to work natural language processing into a chatbot.
It's more important that you set user expectations about what the bot can and cannot understand.
Our experience at Chatbot Concierge has been that people immediately want to ask random questions.
What’s the weather today?
What day is it?
When is Thanksgiving?
What are you wearing? (our chatbots are dressed in formal evening wear for work, if you must know).
By starting a conversation with a simple onboarding script to tell anyone how to use the chatbot, and what to expect in return, you can avoid the immediate user response:
"This thing is broken. It doesn't work."
How to interact with a chatbot
You can start a conversation with a chatbot simply by sending it a message.
Your message can be text, voice, image, emoji or silly gif.
Try these bots
Go ahead and do it already. You know you wanna try some.
SAY HELLO :)
CHATBOT CONCIERGE BUILDS CHATBOTS FOR BUSINESSES.
Get in touch!
Now that you’ve read the use cases…try us. Chatbot Concierge is helping build the conversational web. We just want to keep in touch with you on facebook messenger. So click and say HI!
