Transducers for Annotating Weather Information in Meteorological Texts in Serbia

Abstract

We present a process of extracting information on meteorological phenomena from texts in Serbian. We used finite state automata and transducers for both text processing and information extraction, through software specialized for linguistic text processing. Information extraction was done by annotating text segments. The extraction rules were described with transducers (finite state transducers and recursive transition networks). Some details of used transducers are presented in this paper, aiming to demonstrate the application of different electronic resources for Serbian, especially the electronic morphological dictionary. Transducers are very efficient tools for language processing. In the case of processing Serbian, it is very important to create different resources and corpora which could allow linguistic research. Therefore, we plan to form a collection of transducers and make it publicly available for different kinds of research in the computational linguistics domain.