A field guide to the QA jungle, circa 2020
A curated index of papers, datasets, and models for when you need to build a machine that actually answers questions.

What it does
This is an awesome-list repository: a hand-maintained index of resources for Question Answering research. It catalogs papers (heavy on BERT variants and 2019–2020 conference proceedings), benchmark datasets like SQuAD and MS MARCO, competition leaderboards, code repositories, and a smattering of lecture slides and books. The scope runs from classical fact-extraction pipelines to modern dense retrieval and conversational QA.
The interesting bit
The curation has a point of view. The “Recent Trends” section front-loads efficiency-focused models—DistilBERT, TinyBERT, ALBERT—alongside task-specific architectures like DilBert and UnifiedQA, suggesting the maintainer was tracking the push to make transformers less computationally embarrassing. The competitions table with its “Over Human Performance” column is a nice touch of honest bookkeeping.
Key highlights
- Heavy coverage of ACL/EMNLP/AAAI 2019 and 2020 papers, with direct arXiv and anthology links
- Dataset leaderboard table tracks which benchmarks have been “beaten” by systems vs. humans
- Sections on QA system anatomy: fact extraction, question understanding, answer generation
- Includes non-English resources (Korean and Chinese descriptions, Baidu’s DuReader)
- Links to historical systems timeline: Wolfram Alpha, Watson’s Jeopardy win, Siri integration
Caveats
- Maintenance appears to have stalled around 2020; recent transformer architectures and post-2020 benchmarks are absent
- The “Codes” section is thin compared to the paper lists—more bibliography than cookbook
Verdict
Useful if you’re entering QA research and want a structured map of the 2018–2020 landscape, or if you’re teaching and need a quick syllabus of canonical papers. Skip it if you need working code or cutting-edge 2024 models; this is a time capsule, not a living benchmark.