Extract form input from PDFs and group keywords into subtopics with Latent Dirichlet Allocation (LDA).