Building and Using Parallel Texts:
Data Driven Machine Translation and Beyond

Shared Task on Word Alignment


This directory includes two small scripts provided for the
participants in the shared task on word alignment.

The code is provided free of charge, AS IS. Feel free to use
it and/or modify it.  

wa_eval_align.pl  - evaluates word alignments; the output consists
                    in standard precision, recall, and F-measure,
                    and the AER error rate. 

wa_check_align.pl - checks if the format of the lines in a word
                    alignment file is valid, that is if the lines
                    follow the format required by the guidelines.

The header of each script provides additional details.

These scripts are only meant to help the participants in the
shared task during the tests they run with their word alignment 
systems. The use of these tools is by no means required.

For comments, suggestions, contact Rada Mihalcea (rada at cs unt edu)