Grammatical Error Correction and Style Transfer via Zero-shot Monolingual Translation
This approach addresses data scarcity for grammatical error correction and style transfer in multiple languages, though it is incremental as it builds on existing monolingual translation methods.
The paper tackled grammatical error correction and text style transfer by framing them as monolingual translation tasks, using only regular language parallel data without direct annotations, and demonstrated reliability across three languages with evaluations on multiple error types and style aspects.
Both grammatical error correction and text style transfer can be viewed as monolingual sequence-to-sequence transformation tasks, but the scarcity of directly annotated data for either task makes them unfeasible for most languages. We present an approach that does both tasks within the same trained model, and only uses regular language parallel data, without requiring error-corrected or style-adapted texts. We apply our model to three languages and present a thorough evaluation on both tasks, showing that the model is reliable for a number of error types and style transfer aspects.