English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Zendy Plus

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Zendy Tools

Zendy Open

Large Language models (LLMs) have exhibited remarkable abilities inunderstanding complex texts, offering a promising path towards human-liketranslation performance. However, this study reveals the misalignment betweenthe translation-specific understanding and the general understanding insideLLMs. This understanding misalignment leads to LLMs mistakenly or literallytranslating some complicated concepts that they accurately comprehend in thegeneral scenarios (e.g., QA). To align the translation-specific understandingto the general one, we propose a novel translation process, DUAT (Difficultwords Understanding Aligned Translation), explicitly incorporating the generalunderstanding on the complicated content incurring inconsistent understandingto guide the translation. Specifically, DUAT performs cross-lingualinterpretation for the difficult-to-translate words and enhances thetranslation with the generated interpretations. Furthermore, we reframe theexternal tools to improve DUAT in detecting difficult words and generatinghelpful interpretations. We conduct experiments on the self-constructedbenchmark Challenge-WMT, consisting of samples that are prone tomistranslation. Human evaluation results on high-resource and low-resourcelanguage pairs indicate that DUAT significantly facilitates the understandingalignment, which improves the translation quality (up to +3.85 COMET) andreduces the literality of the translation by -25% to -51%.

Aligning Translation-Specific Understanding to General Understanding in Large Language Models