view release on metacpan or  search on metacpan

lib/Algorithm/AM/algorithm.pod  view on Meta::CPAN

associated a number which is one of the following:

=over 4

=item Number of I<occurrences>

This is merely the number of supracontexts containing the subcontext
containing the data item.

=item Number of I<pointers>

Assign to each supracontext a number representing the total number of
data items in the subcontexts it contains.  The number of pointers of
a particular data item is the sum of the numbers assigned to the
supracontexts containing the subcontext containing the data item.
(Number of occurrences can be thought of as assigning 1 to each
supracontext.)

=back

Using pointers gives rise to I<gang effects>: the ability of many data
items less similar to the test item but appearing in the same
subcontext to have more influence on the outcome than a few data items
more similar to the test item.

With the analogical set in hand, one can then describe the likelihood
of the various outcomes actually occurring by looking at the numbers
assigned to its data items.

=head1 OVERVIEW

I<Exemplar-based modeling> works as follows: there is a set of data
items, each assigned an outcome, and there is a test item.  The test
item is compared with the data items; the result of this comparison
tells what the possible outcomes of the test item are, along with
their likelihoods.

Exemplar-based modeling is often contrasted with I<rule-based
modeling>.  Note that in rule-based modeling, there can be only one
possible outcome, unless the model is fudged by introducing
probability.  (Some types of exemplar-based modeling also give only
one possible outcome.)

I<Analogical Modeling> (AM) is one way to do the comparison and
determine the outcome.  Some of its salient features are as follows:

=over 4

=item *

Exemplars that seem less similar to the test item than those that seem
more similar can still have a magnified effect if there are many of
them.  This is known as the I<gang effect>.

=item *

AM accounts for I<leakage>.

For instance, it is possible for someone to accidentally say "snew"
instead of "snowed", in analogy with "know/knew", "grow/grew",
"throw/threw", "blow/blew", etc.  (I've never done this myself, though
I know someone who has.)  In rule-based modeling, this could never
occur; in AM, this is predicted to occur, though with very low
frequency.

=back

=head1 AUTHOR

Theron Stanford <shixilun@yahoo.com>, Nathan Glenn <garfieldnate@gmail.com>

=head1 COPYRIGHT AND LICENSE

This software is copyright (c) 2021 by Royal Skousen.

This is free software; you can redistribute it and/or modify it under
the same terms as the Perl 5 programming language system itself.

=cut