In this paper we present a method for accurate and precise recognition of personal names im-plemented for Serbian. It is based on devel-opment of comprehensive e-dictionaries of Ser-bian personal names, as well as foreign personal names transcribed to Serbian. In order to ob-tain high precision, the set of finite state au-tomata (FSA) were developed to model various constraints. The same automata are also used to extract from a text personal names not yet covered by e-dictionaries.