Chapter 10 – Latent Dirichlet Allocation

1. LDA assumes that a document is the result of topics and that a topic is a distribution of words. By discovering the hidden topics and words, LDA can assign a document to topics with probabilities. LDA considers hidden topics as templates in a printing shop. Each topic template contains a set of words. An article is “generated” from a topic template or a mixture of topic templates.
2. The name Latent Dirichlet Allocation describes its technical approach. It contains the word latent because it finds the hidden topics in the latent space. The word Dirichlet (pronounced as Deer-e-kh-let) refers to the assumption that both the distribution of topics in a document and the distribution of words in a topic follow Dirichlet distributions. Allocation means that a mixture of topics and words is generated from the topic templates and allocated to a document.
3. The distribution of topics and the distribution of words in a topic...