alansaid

Coherence (magic barrier) could be used as an evaluation measure. This would however need an extended data model. Probably it would make sense to wait with this till [Apache Commons...

Comment: The parsers in net.recommenders.rival.split.parser should be moved to examples. The reason for this is that they all correlate to an example dataset. There is also a parser in net.recommenders.rival.evaluation.parser...

Command line interface

4

Make Rival accessible and configurable through command line arguments to make the binary distribution usable without property files or configuration in Java.

per-user splitting

1

Implement a splitting strategy according to http://graphlab.com/files/lsrs2013/paper_12.pdf

Implement evaluation according to http://graphlab.com/files/lsrs2013/paper_12.pdf This will likely have to be an evaluation strategy taking a different strategy as input since the per-user evaluation does "standard" evaluation for each user...

The main pom file needs to be cleaned up (remove redundant things, etc).

Python interface through JCC or similar https://pypi.python.org/pypi/JCC/2.12

We need to set up proper logging and make sure we log things in new (and already implemented) classes.

Complete overhaul of tests. Increase coverage of tests.

Updated prepare script to work on "-current" version of DBpedia

5

This script will build the indexer, download the **current** version of DBPedia and run the indexer.