What it does This is a hand-picked bibliography of NLP papers, grouped by task—POS tagging, parsing, NER, coreference, sentiment, MT, semantic parsing, QA, and more. Each entry gets a one-sentence “TLDR” explaining the actual contribution, not just the title. Special emoji flags mark “LEGEND” papers (field-shaping work) and “RESOURCE” papers (datasets or tools you might actually use).

The interesting bit The curation has an opinion. Papers span 1993 to 2019, and the author clearly values knowing your history: the 2003 unlexicalized parsing paper sits next to 2014’s neural dependency parser, the IBM MT models share space with seq2seq. There’s no forced narrative of linear progress, just a practitioner’s sense of what ideas keep resurfacing.

Key highlights

• Covers ~10 NLP subfields with 3–9 papers each, from classical methods through deep learning
• Explicit “LEGEND” and “RESOURCE” tags help prioritize when you’re drowning in citations
• TLDRs are genuinely informative, not resume-padding—e.g., “Don’t forget non-deep learning methods!” for a 2016 reading-comprehension paper
• Includes dataset papers (SQuAD, SNLI, Stanford Sentiment Treebank) alongside model papers
• Open to PRs; maintained by a single curator with a Twitter handle and a point of view

Caveats

• Stops around 2019; the transformer era and LLM explosion are essentially absent
• No search, no tagging beyond the two emoji categories, no PDF hosting—just Markdown links
• Coverage depth varies: MT gets 9 papers, some sections get 3

Verdict Worth bookmarking if you’re building NLP systems and want to stop pretending you’ve read the foundational papers. Skip it if you need a living, automatically updated survey—the pace of the field has outrun this snapshot.