In this paper we give an overview of our approach to the implementation and use of innovative text processing technologies. Our goal is to develop, on the basis of predicate-argument lists, a fact extraction application that will allow us to automatically recognize semantic relations between the most significant facts of predefined domains represented in our corpus collection (transcripts and web texts). Taking into account the latest developments in this field, as well as a text processing infrastructure already developed at ILSP, we report on the methodology followed in the implementation of the fact extraction module.