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/

Global: Zendy Plus annual

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

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Existing large language models (LLMs) for machine translation are typicallyfine-tuned on sentence-level translation instructions and achieve satisfactoryperformance at the sentence level. However, when applied to document-leveltranslation, these models face a significant challenge, particularly whendealing with documents containing over 512 tokens. This challenge arises fromthe issue of sentence-level coverage, where subsequent sentences in thedocument remain untranslated. As a result, the document-level translationcapability of LLMs fine-tuned on sentence-level translation instructions issignificantly limited. We conjecture that the primary cause of LLMs' weakdocument-level translation performance is the absence of document-to-documentmapping ability. To address the issue, we propose an approach that combinessentence-level and document-level translation instructions of varying lengthsto fine-tune LLMs. Our proposed translation mixed-instructions enable LLMs(Llama-2~7B and 13B) to maintain consistent translation performance from thesentence level to documents containing as many as 2048 tokens. Extensiveexperimental results show that the proposed approach significantly enhances thedocument-level translation capabilities of LLMs on 10 language pairs,effectively mitigating the sentence-level coverage issue in document-leveltranslation. Experimentation on discourse phenomena has demonstrated that ourdocument-level translation approach significantly improves translation quality,both in terms of BLEU score and discourse coherence.

Enhancing Document-level Translation of Large Language Model via Translation Mixed-instructions