Customisable semantic analysis of texts

Our customisable semantic analysis system implements a form of knowledge acquisition. It automatically extracts syntactic units from a text and semi-automatically assigns semantic information to pairs of units. The user can select the type of units of interest and the list of semantic relations to be assigned. The system examines parse trees to decide if there is interaction between concepts that underlie syntactic units. Memory-based learning proposes the most likely semantic relation for each new pair of syntactic units that may be semantically linked. We experiment with several configurations, varying the syntactic analyzer and the list of semantic relations.