Classification-Based Self-Learning for Weakly Supervised Bilingual Lexicon Induction

• DOI: 10.18653/v1/2020.acl-main.618
• Publication Date: 2020-07-29T14:14:43Z
• Authors: Mladen Karan; Ivan Vulić; Anna Korhonen; Goran Glavaš
• Keywords: Benchmark (surveying); Iterative Learning Control; Word embedding

Effective projection-based cross-lingual word embedding (CLWE) induction critically relies on the iterative self-learning procedure. It gradually expands initial small seed dictionary to learn improved mappings. In this work, we present ClassyMap, a classification-based approach self-learning, yielding more robust and effective of CLWEs. Unlike prior methods, our allows for integration diverse features into process. We show benefits ClassyMap bilingual lexicon induction: report consistent improvements in weakly supervised setup (500 translation pairs) benchmark with 28 language pairs.