Abstract
DeriNet is a large linguistic resource containing over 1 million lexemes of Czech connected by almost 810 thousand links that correspond to derivational relations. In the previous version, DeriNet 1.7, it only contained very sparse annotations of features other than derivations - it listed the lemma and part-of-speech category of each lexeme and since version 1.5, a true/false flag with lexemes created by compounding.
The paper presents an extended version of this network, labelled DeriNet 2.0, which adds a number of features, namely annotation of morphological categories (aspect, gender and animacy) with all lexemes in the database, identification of root morphemes in 250 thousand lexemes, annotation of five semantic labels (diminutive, possessive, female, iterative, and aspect) with 150 thousand derivational relations, a pilot annotation of parents of compounds, and another pilot annotation of so-called fictitious lexemes, which connect related derivational families without a common synchronous pare