In the past, similar systems for encoding heuristics about
common-sense knowledge (such as Cyc and OpenCog) have been used in
conjunction with domain-specific expert systems to avoid the strange
glitches that sometimes occur in reasoning when an expert system lacks
any general knowledge. (The archetypal example is the expert system
written by the US military during the cold war to predict soviet
responses to news articles, which at one point suggested that the
Russians would throw eggs at the United States).
That said, Nell is a bit different from Cyc and Cog in that Nell's
heuristics are learned rather than written by hand. So, a system like
Nell's could be plugged into some domain-specific search system for
the purpose of disambiguation and entity tagging (a search for "bank"
might bring up a suggestion to choose between "river bank" and
"financial institution", and segregate the results based on some
analysis of which context the search term was used in)
--
--
John Ohno
http://firstchurchofspacejesus.blogspot.com/