KORPUSBASIERTE EXTRAKTION VON (MEHRWORT-) TERMINOLOGIEKANDIDATEN: EINE STUDIE ANHAND DER JURISTISCHEN FACHSPRACHE DES DEUTSCHEN

Abstract

The paper represents a pilot study on terminology extraction. The main objective of the study is to clarify to what extent the effectiveness of automated terminology extraction can be improved based on the keyword identification method as opposed to mere structure-based term extraction. In the project, single- and multi-word terms are extracted from a specialized legal corpus of decisions of the German Federal Court of Justice (BGH). Thereby, two relatively simple procedures for automated terminology extraction are tested. On the one hand, multi-word expressions are identified according to a pure syntactic pattern, and on the other hand, the keywords of the BGH corpus are evaluated as terminological candidates and as a starting point for the identification of multi-word terms. The two methods are compared using precision and recall measures. These are calculated based on matching automatically extracted terminological candidates with manually annotated terms in the BGH corpus that are regarded in the study as the gold standard annotation.