indexing research papers with extracting different metadata and relationships — beyond full text chunking and embedding