On the Limitations of Unsupervised Bilingual Dictionary Induction
Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review
Standard
On the Limitations of Unsupervised Bilingual Dictionary Induction. / Søgaard, Anders; Ruder, Sebastian ; Vulic, Ivan.
Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics: (Long papers). Association for Computational Linguistics, 2018. p. 778–788.Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review
Harvard
APA
Vancouver
Author
Bibtex
}
RIS
TY - GEN
T1 - On the Limitations of Unsupervised Bilingual Dictionary Induction
AU - Søgaard, Anders
AU - Ruder, Sebastian
AU - Vulic, Ivan
PY - 2018
Y1 - 2018
N2 - Unsupervised machine translation—i.e.,not assuming any cross-lingual supervisionsignal, whether a dictionary, translations,or comparable corpora—seems impossible,but nevertheless, Lample et al.(2018a) recently proposed a fully unsupervisedmachine translation (MT) model.The model relies heavily on an adversarial,unsupervised alignment of word embeddingspaces for bilingual dictionary induction(Conneau et al., 2018), which weexamine here. Our results identify the limitationsof current unsupervised MT: unsupervisedbilingual dictionary inductionperforms much worse on morphologicallyrich languages that are not dependent marking,when monolingual corpora from differentdomains or different embedding algorithmsare used. We show that a simpletrick, exploiting a weak supervision signalfrom identical words, enables more robustinduction, and establish a near-perfectcorrelation between unsupervised bilingualdictionary induction performance and a previouslyunexplored graph similarity metric
AB - Unsupervised machine translation—i.e.,not assuming any cross-lingual supervisionsignal, whether a dictionary, translations,or comparable corpora—seems impossible,but nevertheless, Lample et al.(2018a) recently proposed a fully unsupervisedmachine translation (MT) model.The model relies heavily on an adversarial,unsupervised alignment of word embeddingspaces for bilingual dictionary induction(Conneau et al., 2018), which weexamine here. Our results identify the limitationsof current unsupervised MT: unsupervisedbilingual dictionary inductionperforms much worse on morphologicallyrich languages that are not dependent marking,when monolingual corpora from differentdomains or different embedding algorithmsare used. We show that a simpletrick, exploiting a weak supervision signalfrom identical words, enables more robustinduction, and establish a near-perfectcorrelation between unsupervised bilingualdictionary induction performance and a previouslyunexplored graph similarity metric
M3 - Article in proceedings
SP - 778
EP - 788
BT - Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics
PB - Association for Computational Linguistics
Y2 - 15 July 2018 through 20 July 2018
ER -
ID: 214756841