Interactive-Predictive Neural Machine Translation through Reinforcement and Imitation
This work addresses the challenge of model personalization for users in machine translation, though it is incremental as it builds on existing interactive and neural methods.
The paper tackles the problem of personalizing neural machine translation with less human effort by introducing an interactive-predictive framework that uses reinforcement and imitation learning to incorporate user feedback. In simulation experiments on two language pairs, the system achieved performance close to supervised training with reduced human effort.
We propose an interactive-predictive neural machine translation framework for easier model personalization using reinforcement and imitation learning. During the interactive translation process, the user is asked for feedback on uncertain locations identified by the system. Responses are weak feedback in the form of "keep" and "delete" edits, and expert demonstrations in the form of "substitute" edits. Conditioning on the collected feedback, the system creates alternative translations via constrained beam search. In simulation experiments on two language pairs our systems get close to the performance of supervised training with much less human effort.