A Beijing lab's living survey of every QA flavor under the sun

What it does

This repository is a curated, continuously updated survey of question-answering research and practice. The BDBC-KG-NLP team at Beihang University splits the field into eight segments — CQA, MRC, KBQA, TQA, and VQA — each examined through separate academic and industrial lenses (where applicable). The output is a stack of linked Markdown documents rather than code.

The interesting bit

The split between “academia” and “industry” per subfield is the useful structural choice. Most surveys conflate the two; this one forces the reader to confront how KBQA or reading comprehension actually deploys versus how it publishes. The team also dogfoods its own research through live demo platforms and a forthcoming book.

Key highlights

• Eight topical segments: CQA, MRC, KBQA, TQA, VQA, each with academic and (where noted) industrial write-ups
• Tracks recent conference progress: ACL, AAAI, SIGIR, CVPR, ICCV, EMNLP, NAACL through 2022
• Active maintenance with dated update logs and a call for open-source contributors
• Spin-off projects: Chinese pre-trained MRC model, a cross-domain KG Q&A platform, and a CCL 2022 shared task
• Backed by papers in TOIS and COLING, plus a stack of Chinese patents and deployed products

Caveats

• Entirely in Chinese; an English version is “in preparation” with no stated timeline
• Last major update batch was mid-2022; some promised VQA updates were slated for July 2022 with no confirmation in the README
• The “industry” side is thinner — only CQA, MRC, and KBQA get industrial sections, while TQA and VQA are academic-only

Verdict

Worth bookmarking if you read Chinese and need a structured entry point into QA research, or if you want to see how a Chinese academic group maps the gap between conference papers and production systems. Skip it if you need runnable code or English prose today.