Improving the Performance of Standard Part-of-Speech Taggers for Computer-Mediated Communication

We assess the performance of off-the-shelve POS taggers when applied to two types of Internet texts in German, and investigate easy-to-implement methods to improve tagger performance. Our main findings are that extending a standard training set with small amounts of manually annotated data for Internet texts leads to a substantial improvement of tagger performance, which can be further improved by using a previously proposed method to automatically acquire training data. As a prerequisite for the evaluation, we create a manually annotated corpus of Internet forum and chat texts.