LDA is an unsupervised learning topic probability generation model. The input is the document collection and the number of topics, and the output is the topic presented in the form of probability distribution. It is often used for topic modeling, text classification, and opinions. Mining and other fields. It assumes a premise: the document is equivalent to a bag-of-words, the words in the bag are independent and interchangeable, without grammatical structure and order. The basic idea is: each document (Document) is composed of multiple topics (Topic), and each topic has multiple corresponding words (Word) to describe.