Three decades of MT papers, sorted so you don't have to

What it does

The Tsinghua NLP Group maintains a curated, hyperlinked bibliography of machine translation research spanning statistical methods to neural architectures. It organizes papers into nested categories—word-based models, phrase-based SMT, low-resource NMT, poetry translation, eco-friendly training, and more—typically pairing each entry with a direct PDF link and Google Scholar citation count. The list serves as a shared reference for researchers who need to trace ideas from 1990s alignment models to modern attention mechanisms without drowning in conference proceedings.

The interesting bit

What saves this from being a mere Zotero export is the explicit editorial humility: the maintainers admit the list is incomplete and that their categories may be “inappropriate,” yet they still carve out niche corners like endangered language revitalization and simultaneous speech translation. They also deliberately defer to the Edinburgh/JHU survey wiki for older SMT coverage rather than pretending to be exhaustive.

Key highlights

• “10 Must Reads” anchors the canon with papers from Brown et al. (1993) through Vaswani et al. (2017), each with citation counts.
• Deep nesting covers IBM alignment, hierarchical phrase-based models, non-autoregressive translation, fairness, and document-level MT.
• Includes practical meta-resources: WMT winning systems, tutorials by Koehn and Neubig, and recent survey papers.
• Cross-references the Edinburgh/JHU MT research survey wiki rather than duplicating historical SMT coverage.

Caveats

• The maintainers explicitly warn that the list is incomplete and that the categorization might be “inappropriate.”
• It is purely a reading list: no code, no automation, no BibTeX export, just manually curated Markdown.

Verdict

Worth bookmarking if you are entering MT research or need to reconstruct the field’s intellectual lineage; skip it if you are looking for runnable models or a managed reference manager.