README File content

This README file describes the WebChild property commonsense database: whose construction is explained in the WSDM 2014 paper: webchild.pdf , webchild-slides.pdf

WebChild contains property commonsense knowledge of a concept with respect to several (19) relations.
The data (160MB) can be downloaded at property-data.txt.


property.txt

This file contains several property relations in the following format:
   Column    |       Type        | Description
-------------+-------------------+-----------
 x_disambi   | character varying | disambiguated subject e.g. plant#n#2
 attr        | character varying | relation e.g hasColor
 y_disambi   | character varying | disambiguated property e.g. green#a#1
 x           | character varying | ambiguous subject e.g. plant
 y           | character varying | ambiguous object e.g. green 
 freq        | integer           | triple's frequency in external data e.g. 100
 numsources  | smallint          | support of distinct sources e.g. 3
 numpatterns | smallint          | support of distinct patterns e.g. 2
 source      | character varying | list of the sources e.g. ngram, wordnet
 score       | real              | normalized triple confidence e.g. 0.91
 higher_attr | character varying | coarse grained attribute e.g. hasAppearance

The details of the disambiguated synsets (e.g. plant#n#2) can be looked up at: noun.gloss.txt