A Linguistically Grounded Graph Model for Bilingual Lexicon Extraction

Abstract

(such as adjectival modification) as edges. Each relation acts as a similarity source for the node We present a new method, based on graph theory, for bilingual lexicon extraction without relying on resources with limited availability like parallel corpora. The graphs we use represent linguistic relations between words such as adjectival modification. We experiment with a number of ways of combining different linguistic relations and present a novel method, multi-edge extraction (MEE), that is both modular and scalable. We evaluate MEE on adjectives, verbs and nouns and show that it is superior to cooccurrence-based extraction (which does not use linguistic analysis). Finally, we publish a reproducible baseline to establish an evaluation benchmark for bilingual lexicon extraction. 1