Essentially application of PCA to text data, where we usually skip the pre-processing step.

We use it for measuring document similarity. Use angle between vectors representing documents.

Intuition on applying PCA to text