Incremental morphosyntactic disambiguation of nouns in German-language law texts

Morphosyntactic disambiguation is a crucial pre-processing step for the recognition of grammatical functions in morphologically rich languages like German and heavily nominalized domains like law texts. This paper explores how far linguistically motivated hard rules can contribute to morphosyntactic disambiguation. It introduces an incremental system that is capable of reducing the rate of morphosyntactically ambiguous nouns in sentences from German-language law texts from 91.12% to 32.31%. The evaluation indicates that disambiguation rules based on feature unification within complex sub-clausal structures such as noun phrase coordinations and participle phrases have the most impact on the reduction of morphosyntactic ambiguity