LEILA: Learning to Extract Information by Linguistic Analysis

Abstract

One of the challenging tasks in the context of the Semantic Web is to \nautomatically extract instances of binary relations from Web documents - for \nexample all pairs of a person and the corresponding birthdate. In this paper, \nwe present LEILA, a system that can extract instances of arbitrary given binary \nrelations from natural language Web documents - without human interaction. \nDifferent from previous approaches, LEILA uses a deep syntactic analysis. This \nresults in consistent improvements over comparable systems (such as e.g. \nSnowball or TextToOnto).