Can anyone explain how the LDA-topic model assigns words to topics? I understand the generative property of the LDA model but how does the model recognize that "Labrador" and "dog" are similar words/ in the same cluster/topic? Is there kind of a similarity measure? The learning parameters of LDA are the the assignment of words to topics, the topic-words probabilities vector and the document-topic probabilities vector. But HOW is it learned?
1 Answer
$\begingroup$
$\endgroup$
You are right, LDA is not very intuitive. It involves a lot of mathematics and concepts. However this video should help you
Also this article
“Intuitive Guide to Latent Dirichlet Allocation” by Thushan Ganegedara https://link.medium.com/texozcnAc6