Neo_Wordnet

This software imports WordNet RDF (version tested: 3.1) into neo4j 2.2.2. This is a sister project of Neo_Concept which imports ConceptNet into neo4j.

Pre-Requisites

• neo4j version 2.2.2 (needed for the new import tool)
• WordNet 3.1 RDF provides a list of assertions
• regular Python, no dependencies

Tested with neo4j-community-2.2.2.

• NEW!: Check out Neo_Merger to import Conceptnet and Wordnet into the same neo4j graph!

How-To

mkdir neo4j-kbs
cd neo4j-kbs
git clone https://github.com/redsk/neo_wordnet.git

# get latest wordnet rdf from http://wordnet-rdf.princeton.edu/wn31.nt.gz
mkdir wordnet
cd wordnet
wget http://wordnet-rdf.princeton.edu/wn31.nt.gz
gunzip http://wordnet-rdf.princeton.edu/wn31.nt.gz
cd ..

# Usage:
# python convertwn.py <input file>
python neo_wordnet/convertwn.py wordnet/nt31.nt

# I've filtered some relations and made some replacements to WordNet namespace (to shorten names)
# Have a look at the log file
less wordnet/WNimporter.log

# If you are interested in a getting the Set of the relations' types then do:
python neo_wordnet/relsSet.py
# From a python shell you can load that set by doing:
import cPickle as pickle
relsSet = pickle.load(open( "../wordnet/relsSet.p", "rb" ) )

# get latest neo4j (tested with neo4j-community-2.2.2)
curl -O -J -L http://neo4j.com/artifact.php?name=neo4j-community-2.2.2-unix.tar.gz
tar zxf neo4j-community-2.2.2-unix.tar.gz

# import WNnodes.csv and WNedges.csv using the new import tool
neo4j-community-2.2.2/bin/neo4j-import --into neo4j-community-2.2.2/data/graph.db --nodes wordnet/WNnodes.csv --relationships wordnet/WNedges.csv --delimiter "TAB"

# start neo4j
neo4j-community-2.2.2/bin/neo4j start

Go to localhost:7474 to see the graph. Create indexes on labels for performance reasons:

CREATE INDEX ON :`lemon#LexicalEntry`(id)
CREATE INDEX ON :`lemon#LexicalSense`(id)
CREATE INDEX ON :`wdo#Synset`(id)

You can now query the database. Example:

MATCH (c0:`lemon#LexicalEntry` { id:"wn/bus+driver-n" }), (c1:`lemon#LexicalEntry` { id:"wn/drive-v" }), path=allShortestPaths((c0)-[*..10]-(c1))
RETURN c0,c1,path