[Heartlogic-dev] new idea

[William L. Jarrold]

OHL-ers!

Two more ideas:

Hrm, there is a risk that I am adding yet another tangent diluting our 
energies.  I wil not work on this in the near future.  But maybe someone
out there will take one ofthese up at this website.  They are not hard
and they are welll defined problems with lotsa pay dirt.

Idea One:

Run a rumination process on the latest version of Open Cyc.  This will
generate deductions.  Apply cyc's natural language generation to
the deductions.  Ask verteran research participatns to rate the 
believeability of the times.  (Maybe some of the items will be
manipulated (or the process that generated the deductions)
in the spirit of computational ablation).  (We really need data
on the average believability of the contents of KBs...Especially
the "vienna is wet" example.  Or the "Hilary feels zero contempt for
Bill Clinton exmaple." You gotta check out those papers.  Jarrold
2001 or so.

Idea Two:

We could have little excerpts from chat room sort of conversations.  do 
the turning challenge to that.

the research participant will see an item like this..

Eliza: How can I help you?

Fred: Well, I am having a good day so I don't really need any help.

Eliza: You don't need any help.

Fred: No, I dont, but I would like to chat.

The above is a conversation between Fred a real human and Eliza.
Eliza may be a human or it may be a computer.  Your job is to
tell us how believable Eliza's reply is as the kind of thing a human
would say.

Please rate the believability of Eliza's statements.

1 Highly Unbelievable

2 Moderately Unbelievable

3 Neutral

4 Moderately Believable

5 Highly Believable