Extraction of Selling Events from Historical Documents

The extraction of information such as events and their participants is becoming increasingly important with the growing amount of digital text. This is also true for historical texts. This thesis aims to extract selling events from a collection of medieval manuscript summaries in German. The events were extracted with reference to FrameNet's Commerce_sell and Getting frames.

This work presents an approach that uses syntactic parses and a set of manually crafted rules to extract selling events and their participants. The presented system succeeds in extracting the most important parts of the selling events (the semantic heads). An additional named entity processing step could slightly improve the results.