Many evaluation campaigns have shown that knowledge-based and data-driven approaches remain equally competitive for Named Entity Recognition. Our research team has developed a symbolic system based on finite state tranducers, which achieved promising results during the Ester2 French-speaking evaluation campaign. Despite these encouraging results, manually extending the coverage of such a hand-crafted system is a difficult task. In this paper, we present results about the use of text mining techniques to automatically enrich our system's knowledge base. We exhaustively search for lexico-syntactic patterns, that recognize named entitites boundaries. We assess their efficiency by using such patterns in a standalone mode and in combination with the existing system