The infovore framework is capable of converting Freebase to RDF, rewriting SPARQL queries using Freebase's name resolution system, and running tests to confirm correct operation of :BaseKB

You'll need a copy of the Freebase quad dump from

http://download.freebase.com/datadumps/

Infovore has been tested against the 2012-11-04 quad dump.

You'll need to select a base directory and an instance name for your copy of Freebase. You'll configure these using shell environment variables

$ export INFOVORE_BASE=/freebase $ export INFOVORE_INSTANCE=2012-11-04

you should install your data dump at

/freebase/data/2012-11-04/input/freebase-datadump-quadruples.tsv.bz2

Infovore will write temporary files to the work subdirectory of the instance directory and will write final output to the output subdirectory.

As currently configured, the instance directory grows to 80GB in the process of creating baseKBLite and baseKBPro. Future versions of Infovore may reduce disk consumption, but currently intermediate files are saved in case they are necessary for research and debugging

First you should build and run the script that installs the path and environment variables to run infovore

$ mvn clean install $ source hydroxide-apps/path.sh

then do

$ createPro.sh

to create :BaseKB Pro, a complete rendition of Freebase in RDF. Once you've created :BaseKB Pro, you can do

$ createLite.sh

to create :BaseKB Lite, a subset of :BaseKB Pro that is restricted to topics that exist in both DBpedia and Freebase.