The Perl Toolchain Summit needs more sponsors. If your company depends on Perl, please support this very important event.
Changes from NSP version 0.73 to 0.91
-------------------------------------

Saiyam Kohli, kohli003@d.umn.edu
University of Minnesota, Duluth

February 7, 2006

Measures
--------

Complete rewrite of Measures from scratch using Object Oriented Perl.

All the measures have now been moved from the Measures directory to the lib
directory and have been rearranged according to the following hierarchy:

Text::NSP
Text::NSP::Measures
Text::NSP::Measures::2D
Text::NSP::Measures::2D::MI
Text::NSP::Measures::2D::MI::ll
Text::NSP::Measures::2D::MI::pmi
Text::NSP::Measures::2D::MI::tmi
Text::NSP::Measures::2D::CHI::x2
Text::NSP::Measures::2D::CHI::phi
Text::NSP::Measures::2D::CHI::tscore
Text::NSP::Measures::2D::Fisher::left
Text::NSP::Measures::2D::Fisher::right
Text::NSP::Measures::2D::Fisher::twotailed
Text::NSP::Measures::2D::Fisher2::left
Text::NSP::Measures::2D::Fisher2::right
Text::NSP::Measures::2D::Fisher2::twotailed
Text::NSP::Measures::2D::odds
Text::NSP::Measures::2D::dice
Text::NSP::Measures::3D
Text::NSP::Measures::3D::MI
Text::NSP::Measures::3D::MI::ll
Text::NSP::Measures::3D::MI::pmi
Text::NSP::Measures::3D::MI::tmi

Similar measures have been grouped together, and the error checks and
computations common to these measures have been moved to the base class for
these measures. The error codes have also been reoragnised, with every mearure
familiy getting a particular range of error codes, the details are available in
the Text::NSP documentation.

We are now using hashes instead of arrays to store the data related to a
continhency table(observed/expected counts) internally.

The overflow errors in the rightFisher and leftFisher measures have been
resolved and a new measure for the twotailed Fishers exact test has been added.


Programs
--------

Changes to statistic.pl have been made to get it working with the new
implementation of measures.

Now statistic.pl searches for the Measure modules using PERL5LIB environment
variable instead of PATH. Some changes were made to make the commnadline
interface backwards compatibile, that is the systems that were using NSP-0.73
(specifically statistic.pl) will keep on working with the new NSP-0.91, without
any changes.

Other changes include printing all the error messaged to STDERR and also a check
to ensure that the sum of all the obseved bigram counts is less than the total
bigram count.