Ask HN: Finding useful relationships in a collection of documents?

It seems like every project ends up with a collection of documents or nodes of some kind. Most of the time these documents have meta data with them (the most famous being 'tags' of some sort).

What are some good resources or common practices for finding, grouping, or generally providing useful relationships from this data?

- Related documents

- Suggested tagging (based in previous documents)

- k-means grouping for categorizing

- "Tag clouds"

- etc...

If this was a relational database, the meta storage would probably consist of just three tables:

TABLE documents { id, body }

TABLE meta {id, meta}

TABLE document_tags {document_id, meta_id, value }

(if `value` was likely to be repeated you could make a forth table)