A Discriminative Latent-Variable Model for Bilingual Lexicon Induction
This work addresses the problem of automatically inducing bilingual dictionaries for machine translation and NLP applications, but it is incremental as it builds on prior methods.
The paper tackles bilingual lexicon induction by introducing a discriminative latent-variable model that combines a bipartite matching dictionary prior with a representation-based approach, resulting in improved lexicons across six language pairs as shown empirically.
We introduce a novel discriminative latent variable model for bilingual lexicon induction. Our model combines the bipartite matching dictionary prior of Haghighi et al. (2008) with a representation-based approach (Artetxe et al., 2017). To train the model, we derive an efficient Viterbi EM algorithm. We provide empirical results on six language pairs under two metrics and show that the prior improves the induced bilingual lexicons. We also demonstrate how previous work may be viewed as a similarly fashioned latent-variable model, albeit with a different prior.