The data:
=========
The data directory contains files derived from the Canadian Hansards,
originally aligned by Ulrich Germann:
-hansards.e English
-hansards.f French
-hansards.a alignments of the first 37 sentences. The notation
  i-j means the word as position i of the French is aligned
  to the word at position j of the English (zero-indexed);
  while notation i?j means they are probably aligned.

Expected output:
================
Your code should produce a file representing an alignemnt for the data in data.
The format of the output file should be:

* Each line corresponds to an alignment for one sentence pair.
* Each line consists of space-separated items of the form i-j where i is the index of a word
  on the f sentence, and j is the index of a word on the e sentence (both indices are 0-based).

The file example_alignment is an example of an alignment file for the provided data (not random, but not a very good one..).

I provide you with a python script called validate.py, which verifies that your data format is correct.
Run is as:
   python validate.py -e data/hansards.e -f data/hansards.f < your_alignemnt_file

For example:
   python validate.py -e data/hansards.e -f data/hansards.f < example_alignment
   python validate.py -e data/hansards.e -f data/hansards.f < README

Evalutation Script
===================
I provide you with an evaluation script called eval.py.

eval.py computes the aignment error rate (AER) of a given alignment based on the gold alignment,
as well as provide a semi-graphical display of the alignments compared to the gold alignments.

to invoke it, use the following syntax from the commandline:

    python eval.py -e e_file -f f_file -a gold_a < predicted_a

for example:

    python eval.py -f data/hansards.f -e data/hansards.e -a data/hansards.a < example_alignment


