Extraction, Evaluation and Integration of Lexical-Semantic Relations for the Automated Construction of a Lexical Ontology

Several approaches for extracting semantic relations from various types of resources have been proposed during the last years. While already of great value when used separately, combining these techniques promises to lead to even broader and more reliable results. However, divergent information may occur when assembling such data. We present LexO, a framework for integrating semantic relations from different sources into an ontological structure. We provide different methods for assigning confidence values to the input data as well as mechanisms to detect and resolve inconsistencies. The present paper focuses on lexical-semantic relations, but the approach presented is extensible to include new kinds of data sources as well as further types of relations.