Document-document correlations tell us whether two passages share vocabulary, ideas, and semantic structure.
Use a Document-Document search if:
You will be able to select one or more, or all, documents for your search. The LSA component will return results for all of the chunks in the selected document or documents.
The closer the cosine similarity approaches 1.0, the more closely related the two chunks are in terms of shared vocabulary and ideas.
If you want to narrow your search to one or more particular passages,use the Chunk-Chunk search instead. Chunk-Chunk correlations