Training Your Bot

You can add to your bot's corpus in two basic ways: conversations and training.

Training your bot allows you to have more control over the responses it gives to other people. It is also a faster, more efficient way to add to your bot's corpus than simply conversing with it.

Many-to-One Example

Let's say you are building a customer service bot, and you want to train it with frequently asked questions (e.g. "How do I open an account?") and answers to those questions ("Please visit any of our branches to open an account").

Because you don't know exactly how the user will phrase a given question, you'll also want to provide several variations of the same question (e.g. "How do I open an account?", "Where can I open an account?", "Open an account", "How do I start an account?", "How do I start a new account?"). All of these questions should all yield the same response: "Please visit any of our branches to open an account".

On the left side of the Training screen you would enter all the similar queries the user might make, and on the right side you would enter all the possible responses. In the case of our customer service bot, we would enter the following:

Queries Responses
How do I open an account? Please visit any of our branches to open an account
Where can I open an account?  
Open an account  
How do I start an account?  

This tells your bot that if the user asks any of the phrases on the left, the bot should respond with the phrase on the right.

Note that you may enter the questions in any order.

Many-to-Many

Let's say you wanted to create a bot that could hold a conversation about, say, Star Trek. You might enter queries and responses like this:

Queries Responses
Who is better: Kirk or Picard? Kirk is clearly superior to Picard!
Who would win: Kirk or Picard? Kirk wins any day, hands down
In a fight between Kirk and Picard, who would win? Kirk rules, Picard sucks!
Who do you like better, Kirk or Picard?  
Who is your favorite, Kirk or Picard?  

Whenever the user asks any of the questions on the left, the bot will respond randomly with one of the phrases on the right.

But what if the user asks a slightly different version of the question, e.g. "Is Kirk better than Picard?" BotMakr's algorithms will cause it to recognize that the user is asking one of the questions it was trained to recognize, and answer accordingly.

The more variations of a question you enter, the better your bot's chances of recognizing and responding to it meaningfully.