Term Extraction and Mining of Term Relations from Unrestricted Texts in the Financial Domain

In this paper, we present an unsupervised hybrid textmining approach to automatic acquisition of domain relevant terms and their relations. We deploy the TFIDFbased term classification method to acquire domain relevant terms. Further, we apply two strategies in order to learn lexico-syntatic patterns which indicate paradigmatic and domain relevant syntagmatic relations between the extracted terms. The first one uses GermaNet, while the second is based on different collocation acquisition methods to deal with free-word order languages like German. This domain-adaptive method yields good results even when trained on relative small training corpora. Therefore, it can be applied for solving information extraction and retrieval tasks within a realworld business information system.