nature-reader

You can build a bilingual paper reader from PDFs, DOIs, arXiv links, HTML pages, or pasted text. It generates paragraph-level Original and Chinese aligned pairs while preserving document structure and source anchors for citation.

Bilingual translation preserves figures and tables by extracting visuals and captions, then placing them near the first substantive mention in the text. This maintains the original reading flow for journal and conference papers.

Figure extraction with captions works on selectable-text PDFs, scanned PDFs, and HTML pages. For scanned PDFs, it performs OCR and clearly labels low-confidence recognition or missing pages instead of silently skipping content.

Source grounding in academic translation keeps stable page and block anchors so every translated claim traces back to the original document. It ensures traceability for follow-up citation and verifies terminology consistency across the full paper.

When handling missing pages or low-confidence OCR during paper translation, the tool applies uncertainty labeling. It clearly marks incomplete retrieval or low-confidence optical character recognition results instead of silently skipping the affected content.

The best way to translate full academic papers while keeping structure is using a source-grounded bilingual reader. It ensures paragraph-level Original and Chinese alignment, terminology consistency, and complete output bundles containing extracted figures and captions.