*********************Readme File for the English Input Files********************

This archive contains the input files for the English past tense 
simulations using the Albright and Hayes Rule-based Learner, as described 
in "Rules vs. Analogy in English Past Tenses: A Computational/Experimental 
Study" (http://www.linguistics.ucla.edu/people/hayes/RulesVsAnalogy/index.htm)

These files were created on a mac; it is possible that you may need to 
convert the line endings in order to edit them on a PC or in unix!
(The learner program itself can use text files of any format, since java 
is designed to handle all three text file formats; Microsoft Excel can 
also open both Mac and DOS text files.  Command-line perl is pickier, 
though, so if you wanted to use the perl script in windows or unix, you may 
need to convert the file to the relevant text format)

It contains two input files: one with the entire set of verbs with a lemma 
frequency of 10 or greater in CELEX, and one in which "prefixed verbs" have 
been removed.
The definition of "prefixed verb" was: the root must be a free verb stem 
(i.e., occurs in isolation without the prefix), and prefix must be acting 
identifiably as the prefix (i.e., its meaning is somehow contained in the 
complex verb, even if the entire verb is not transparent).  For example, 
"react" contains the free stem "act", and the meaning of "react" was 
judged to contain the meaning of "re-", so this was treated as a prefixed 
form and removed.

The .in files are the input files, the .fea files contain the feature 
matrix which we employed for English.

Thus, there are four files:

CELEXFull.fea
CELEXFull.in
CELEXPrefixStrip.in
CELEXPrefixStrip.fea

There is also a batch file (CELEXFitAlphasAll.bat), which can be run using 
the "Open Batch File" option in the File menu of the Learner program; 
this is designed to run, in serial, a variety of parameter settings for 
confidence values, for purposes for fitting to experimental data.

Finally, there is a perl script for collecting the predictions of the 
rule-based learner for the specific novel verbs which were considered in 
the "Rules vs. Analogy" paper; this is probably of limited usefulness by 
itself, but could be modified to collect ratings for other verbs and 
outputs for other studies:  GetLearnerPreds.pl

********************************************************************************

Created 10/17/01 by Adam Albright
For additional questions and information, contact me: aalbrigh@ucla.edu
