Current structure of the *_data tables:
1. Blob
	word  - contains word to look up
	idx - blob containing id's of all documents containg that word
		along with number of occurances
2. Column
	word_id - contains id of word to look up
	doc_id - contains id of doc
	count - contains number of occurances
3. Phrase
	word_id - contains id of word to look up
	doc_id - contains id of doc
	idx - contains locations of occurances