TRAINING DIRECTORY
------------------

The file train.in contains 15,000 sentences for development and training.
Each line of the file consists of an untokenized sentence and is preceded by
a sentence identifier.

The "gold standard" of gene mentions for train.in is in the file GENE.eval,
acceptable variants are in the file ALTGENE.eval.

ALT_EVAL
--------

The evaluation program requires a file consisting of gene mentions for all of the
sentences in train.in (or for a subset of sentences, see CROSS VALIDATION).
The evaluation program is a perl script and should be invoked from this directory
with the command line:

% perl alt_eval.perl mention-file

The mention-file must be formatted according to the description in README.GM
(the files GENE.eval and ALTGENE.eval are also in this format).

The output of the evaluation program lists all true positives (with an asterisk
for acceptable alternatives), false positives, and false negatives. The final
lines of the evaluation output give the number of TP, FP, and FN, and the calculated
precision, recall, and F score.

CROSS VALIDATION
----------------

If you wish to evaluate on a subset of the sentences in train.in, make a file containing
a list of the sentence identifiers you wish to evaluate. Then invoke the evaluation
program with the "-ids" option:

% perl alt_eval.perl -ids id-file mention-file

The mention-file contains the mentions from the sentences listed in the id-file that
are to be evaluated.

WORKING DIRECORY
----------------

The files GENE.eval and ALTGENE.eval should be in the current working directory
when the program is run.  Alternatively, the file names can be specified on the
command line:

% perl alt_eval.perl -gene gene-eval-file -altgene altgene-eval-file mention-file

