Create fast graph language models from converted PDF documents for knowledge extraction and Q&A.