Build Chatbot Interactions

Markov Chain: Train Lita to Speak in Someone Else’s Voice

In this section, you’ll use a Markov chain to respond to users’ questions. Markov chains are a technique for probabilistically generating content based on observations of past content. In a chatbot context, that means feeding a large corpus—lots and lots of words, such as a book or a collection of emails—into an analyzer and observing which words and phrases are most likely to come after any given word. As a small example, consider the following two sentences:

The cow jumped over the moon.

The quick brown fox jumped over the lazy dog.

If you were trying to implement a naive predictive text generator using only those two sentences as inputs, you could see that the word “jumped” is followed ...