In this study, we describe a method for parsing part-of-speech tagged unrestricted texts in Swedish using finite-state networks. We use the Xerox Finite-State Tool because of its expressiveness and power for writing and compiling regular expressions and relations. The parser is divided into four modules: i) contiguous phrase structure marker, ii) phrasal head marker, iii) syntactic function tagger, and iv) non-contiguous group boundary recognizer. The aim is to develop a parser that can be used as a light/shallow parser for marking phrase structure and, when needed, to label syntactic functions. We believe that modularity is necessary since different NLP tasks require various levels of analysis. The parser for Swedish is under development, but present-day results are promising.