Multi-perspective Alignment for Increasing Naturalness in Neural Machine Translation
This addresses the issue of unnatural translations in NMT for applications requiring high-quality, human-like text, though it is incremental as it builds on reinforcement learning from human feedback.
The paper tackled the problem of neural machine translation systems producing artificially impoverished language due to lexical biases, which reduces their usefulness for tasks like creating evaluation datasets. The result was a method that increased lexical richness and naturalness in English-to-Dutch literary translation without loss in translation accuracy.
Neural machine translation (NMT) systems amplify lexical biases present in their training data, leading to artificially impoverished language in output translations. These language-level characteristics render automatic translations different from text originally written in a language and human translations, which hinders their usefulness in for example creating evaluation datasets. Attempts to increase naturalness in NMT can fall short in terms of content preservation, where increased lexical diversity comes at the cost of translation accuracy. Inspired by the reinforcement learning from human feedback framework, we introduce a novel method that rewards both naturalness and content preservation. We experiment with multiple perspectives to produce more natural translations, aiming at reducing machine and human translationese. We evaluate our method on English-to-Dutch literary translation, and find that our best model produces translations that are lexically richer and exhibit more properties of human-written language, without loss in translation accuracy.