Cross Validated is a question and answer site for people interested in statistics, machine learning, data analysis, data mining, and data visualization. It only takes a minute to sign up.
Sign up to join this community
I have my set of documents and have extracted the unique words from them, including a count of the number of times each word appears in the document. But it would seem from the documentation on the Python library I'm using, that word count within documents is immaterial, and that co-occurrence is the main thing.
Am I correct in that understanding?
Note that LDA has the assumption that each topic, or a language model, is a multinomial distribution over all words. For a document, multiple occurrence of the same word is a draw from this multinomial distribution. So, in theory, word counts matter.
However, in terms of inference, if you use Gibbs sampling, you probably need to re-sample topic assignments for multiple occurrence. But if you use variational inference, you just do it the same across multiple occurrence.
