Final Instructions for the Author Profiling Task

[Martin Potthast]

Dear participants of the author profiling task,

as many of you have already noticed, the training and test datasets
for author profiling have been updated recently. The update is due to
an error in the datasets that allowed for cheating. We have fixed
these errors now and hence re-release also the training datasets.

The new datasets are now available on TIRA, including test dataset 2.
The old datasets and all runs against them are now deprecated (TIRA
indicates that by striking out their names).

For the final evaluation, we would like you to do the following:

1) Train your software on the new training dataset.

2) Run your trained software against the following test datasets:
- pan15-author-profiling-test-dataset2-dutch-2015-04-23
- pan15-author-profiling-test-dataset2-english-2015-04-23
- pan15-author-profiling-test-dataset2-italian-2015-04-23
- pan15-author-profiling-test-dataset2-spanish-2015-04-23

3) Evaluate all runs on these datasets using the evaluator.

The deadline to finish these runs is May 24, 2015.

Please send an email to us once you are done.

Best,
Martin

PS: We thank the team Alberto Bartoli, Alex Dagri, Alessandra
Laderchi, Andrea De Lorenzo, Eric Medvet, and Fabiano Tarlao for
bringing these errors to our attention in time to fix them.