Abstract: DeriNet is a lexical network of Czech lexemes that are interconnected according to derivational relations between them; the network has been developed at the Institute of Formal and Applied Linguistics.
The network was initialized with a set of lexemes whose existence was supported by corpus evidence, and it is currently limited to derivational relations. This limitation is reflected in the architecture of the network in that each lexeme is allowed to be linked with just a single base word. Derivational relations have been created semi-automatically in several steps, which will be addressed in detail in the talk.
The network has an applicational potential in several NLP tasks and is expected to be a reliable resource of data for theoretical research in word-formation.
The talk will be given in Czech.