Automatically Linking GermaNet to Wikipedia for Harvesting Corpus Examples for GermaNet Senses

The comprehension of a word sense is much easier when its usages are illustrated by example sentences in linguistic contexts. Hence, examples are crucially important to better understand the sense of a word in a dictionary. The goal of this research is the semi-automatic enrichment of senses from the German wordnet GermaNet with corpus examples from the online encyclopedia Wikipedia. The paper describes the automatic mapping of GermaNet senses to Wikipedia articles, using proven, state-ofthe-art word sense disambiguation methods, in particular different versions of word overlap algorithms and PageRank as well as classifiers that combine these methods. This mapping is optimized for precision and then used to automatically harvest corpus examples from Wikipedia for GermaNet senses. The paper presents details about the optimization of the model for the GermaNet-Wikipedia mapping and concludes with a detailed evaluation of the quantity and quality of the harvested examples. Apart from enriching the GermaNet resource, the harvested corpus examples can also be used to construct a corpus of German nouns that are annotated with GermaNet senses. This sense-annotated corpus can be used for a wide range of NLP applications.