Correct-and-Memorize: Learning to Translate from Interactive Revisions
This work addresses the challenge of making machine translation more efficient and accurate for users by reducing repetitive human corrections, though it is incremental as it builds on existing interactive translation approaches.
The paper tackles the problem of improving machine translation through human interaction by proposing CAMIT, a method that allows humans to correct critical errors and learns from these corrections to avoid repeating mistakes. Experiments show CAMIT significantly enhances translation results while requiring fewer human revisions compared to previous methods.
State-of-the-art machine translation models are still not on par with human translators. Previous work takes human interactions into the neural machine translation process to obtain improved results in target languages. However, not all model-translation errors are equal -- some are critical while others are minor. In the meanwhile, the same translation mistakes occur repeatedly in a similar context. To solve both issues, we propose CAMIT, a novel method for translating in an interactive environment. Our proposed method works with critical revision instructions, therefore allows human to correct arbitrary words in model-translated sentences. In addition, CAMIT learns from and softly memorizes revision actions based on the context, alleviating the issue of repeating mistakes. Experiments in both ideal and real interactive translation settings demonstrate that our proposed \method enhances machine translation results significantly while requires fewer revision instructions from human compared to previous methods.