Topic Modeling is a NLP algorithm that $($only one answer is true$)$:

Uses cross$-$validation to sample the text for the training set.

Uses conditional probabilities to update the weights$/$probabilities of topic assignement to documents.

Is a simulation where words $($n$-$grams$)$ are assigned to segmentation.

Uncovers $10$ topics with $10$ words each in a corpus of documents.