Lost in Translationese? Reducing Translation Effect Using Abstract Meaning Representation
This work addresses the issue of translationese affecting model performance in natural language processing, representing a novel task with domain-specific implications.
The paper tackles the problem of translationese in translated texts by using Abstract Meaning Representation (AMR) as an interlingua to reduce its effects, resulting in text that more closely resembles originally English text across three quantitative macro-level measures without severely compromising fluency or adequacy.
Translated texts bear several hallmarks distinct from texts originating in the language. Though individual translated texts are often fluent and preserve meaning, at a large scale, translated texts have statistical tendencies which distinguish them from text originally written in the language ("translationese") and can affect model performance. We frame the novel task of translationese reduction and hypothesize that Abstract Meaning Representation (AMR), a graph-based semantic representation which abstracts away from the surface form, can be used as an interlingua to reduce the amount of translationese in translated texts. By parsing English translations into an AMR and then generating text from that AMR, the result more closely resembles originally English text across three quantitative macro-level measures, without severely compromising fluency or adequacy. We compare our AMR-based approach against three other techniques based on machine translation or paraphrase generation. This work makes strides towards reducing translationese in text and highlights the utility of AMR as an interlingua.