Latent Dirichlet Allocation (LDA) is a common method of topic modeling. That is, if I have a document and want to figure out if it's a sports article or a mathematics paper, I can use LDA to build a system that looks at other sports articles or mathematics papers and automatically decides whether this unseen document's topic is sports or math.
To LDA, a document is just a collection of topics where each topic has some particular probability of generating a particular word.
I can teach you about LDA in Machine learning.
Thanks