Recognition Targets
Recognition targets a set of gensyms that are processed by the
recognize operator to determine if particular criteria have
been satisfied. We expect that the productions that produce these
gensyms could be constructed by deliberate data chunking, but at
present we have not implemented this explicitly. Instead, we have a
number of hand-written productions which recognize the necessary
conditions and put out simulated gensyms.
The gensyms and criteria being tested are:
- recog1
- Matches to visually (foveally) perceived NTD in the CMD or
RESP column.
- recog2
- Never matches. Originally designed to match words appearing
inside attention but outside the fovea, but because of changes
to the I/O structure this is no longer used. Retained for
numbering purposes and because it constitutes an interesting
hypothesis about what we know about the extra-foveal visual
field.
- recog3
- Matches when the step number is present in the fovea but there
is no crossout in the same location.
Productions are in file fudge.recognition.soar6