Building Concept Frames based on Text Corpora

Linguists have been using different kinds of frame representation since the emergence of the notion “frame”. The main goal of the annotation system described in this paper is to provide an interactive and easy-to-use tool for structuring concept-specific information in linguistic frames for discourse analysis or cultural studies. These frames take into account background or “world” knowledge associated with the concepts, which is not necessarily present in lexicographic frames. A frame hierarchy providing default information, example texts containing specific information on a concept, and the annotations made by a user are combined together in one database. All frames have a predefined structure, and the information they contain is represented in natural language. The collected information can also be used as input to knowledge bases, or for defining patterns for Information Extraction.