Simple similarity service based on chapter 3 of http://infolab.stanford.edu/~ullman/mmds.html.
Backed by Redis but easily adaptable to other stores.

Very, very, very early days.

Building:

mvn package

The test scripts described below assume you have a redis going on localhost.

Initializing:

scripts/tool Initialize <bands> <rows> <minCount>

- only run this *once*, before you load in any data
- bands and rows define the similarity threshold, as outlined in the book
- minCount is how many items a set has to have before it is considered for similarity with others
- a good default set of values:
    scripts/tool Initialize 5 25 5

Loading:

scripts/tool Load <file.tsv>

- run as many times as you like with different data
- expects two tab-separated columns: set key, then item key

Dumping:

scripts/tool Dump

- this will dump out all pairs of similar sets
- format is TSV, columns are similarity, size of intersection, item1, item2