[
CogSci Summaries home |
UP |
email
]
http://www.jimdavies.org/summaries/
Forbus, K.D, Ferguson, W. R. Gentner, D. (1994). Incremental
Structure Mapping, Proceedings of the 14th Annual Conference of the
Cognitive Science Society, 313--318.
@InProceedings{,
author = {Kenneth D. Forbus, Ronald W. Ferguson, and Dedre Gentner},
title = {Incremental Structure-Mapping},
booktitle = {Proceedings of the Sixteenth Annual Conference of
the Cognitive Science Society},
OPTcrossref = {},
OPTkey = {},
OPTpages = {313--318},
OPTyear = {1994},
OPTeditor = {Ashwin Ram and Kurt Eiselt},
OPTvolume = {},
OPTnumber = {},
OPTseries = {},
OPTaddress = {},
OPTmonth = {},
OPTorganization = {},
OPTpublisher = {Lawrence Erlbaum Associates},
OPTnote = {},
OPTannote = {}
}
Author of the summary: Jim Davies, 2003, jim@jimdavies.org
Cite this paper for:
Problem: how to extend mappings when new information is found.
Inspired by Keane and Bradshaw's (1988) IAM model and incremental
mapping in PHINEAS.
There's some discussion of what about incremental mapping should be
seriel and what should be parallel, then about the role of pragmatic
constraints. [314]
"When I-SME is given every item in the base and target at the same
time, its results are identical to those of SME."
Major features: [315]
- can determine which previous results are still valid
- default operation: extend mapping
- the remap operation backtracks. (the system does not
automatically do this because remapping is task specific.)
Two filters
- required: exclude a mapping that is not one to one.
- identical-functions: only identical functions.
Example 1: giving the agent a few facts at a time, and watch how the
mapping changes. Models evidence gathered by Keane (in press)
Example 2: MARS. comparing a worked example and a problem to be
solved. [317]
Summary author's notes:
- In Covlan, sequential simages are made up of different
symbols. I-SME extends a given mapping with new
information. So to handle Covlan's representation, it would have
to use the horizontal mapping given find what changed, turn the
next simage into the symbols of the first, plus whatever new is
added, then add the new stuff to the incremental mapper. Then it
would have to turn it all back into the symbols used in the next
simage again.
-
Back to the Cognitive Science Summaries homepage
Cognitive Science Summaries Webmaster:
JimDavies
(jim@jimdavies.org)
Last modified: Fri Nov 7 10:41:03 EST 2003