Eunho Yang, Aurelie C. Lozano, et al.
ICML 2014
This paper proposes a new type of transfer mechanism, called Rule Combination Transfer (or RCT), that produces an output structure by non-destructively combining target parts of translation rules, each of which consists of dependency structures in source and target languages and correspondences between them. This proposed mechanism employs more extended mapping of correspondences than one-to-one mapping used in conventional transfer systems, to allow expression of some peculiar or exceptional translation phenomena. Further, these translation rules can be used bi-directionally. Although, the proposed transfer mechanism is intended to be a foundation for an example-based transfer system by coupling it with a mechanism selecting translation rules based on the similarity with an input, it can be also used as a foundation for a conventional transfer system. © 1995 Kluwer Academic Publishers.
Eunho Yang, Aurelie C. Lozano, et al.
ICML 2014
Joseph Y. Halpern, Yoram Moses
Journal of the ACM
Yale Song, Zhen Wen, et al.
IJCAI 2013
Arthur Nádas
IEEE Transactions on Neural Networks