About the Available Data
Some of the data generated by Knext is available
for download and use. Note that released knowledge bases should be considered
historical and be identified by the date they were last updated. More recent
results can be seen in the browser.
Factoids with Expected Frequencies: 10,000 verbalized factoids annotated with how often we might expect them to occur, based on textual patterns – as described in Gordon & Schubert 2012. Download (408K).
82,247 unique unsharpened factoids learned from the Brown Corpus.
Updated 2011-08-18. Download (1.5M).
5,358,996 unique unsharpened factoids learned from the British National
Corpus. Updated 2011-08-18. Download
An initial release of rated factoids, evaluated
using Mechanical Turk as described in Gordon
et al. 2010.
Available by Request
Other output, such as the results of sharpening or the results from larger
corpora such as Wikipedia or weblogs, may be made available by request.