Typos in Czech Corpora

This publication doesn't include Faculty of Arts. It includes Faculty of Informatics. Official publication website can be found on muni.cz.

Authors

GRÁC Marek

Type Article in Proceedings
Conference Seventh Workshop on Recent Advances in Slavonic Natural Language Processing, RASLAN 2013
MU Faculty or unit

Faculty of Informatics

Citation
Field Informatics
Keywords text corpus; errors in text;
Description The extended usage of written corpora not only for manual querying but also for machine learning led to the creation of massive corpora. These corpora are almost solely crawled from the internet and contain texts of various quality. Corpora that contain more typos or ungrammatical texts are more difficult to use for computational linguists and are thus a major obstacle in automatic development. In this paper we attempt to qualify some of existing Czech corpora using manually created wordlist. We will show that building such a list of frequent typos can be done without major investing when agile techniques are used.
Related projects: