This work presents a system for the categorization of noisy texts. By noisy it is meant any text obtained through an extraction process (affected by errors) from media different than digital texts. We show that, even with an average Word Error Rate of around 50%, the categorization performance loss with respect to the clean version of the same documents is negligible.
Lesly Sadiht Miculicich Werlen
Martin Jaggi, Vinitra Swamy, Angeliki Romanou
Frédéric Kaplan, Maud Ehrmann, Matteo Romanello, Sven-Nicolas Yoann Najem, Emanuela Boros