It may be open domain question answers or restricted to any particular domain.
Related (but not restricted) to the Linked Data domain, QALD provides a benchmark for multilingual question answering, as well as a yearly evaluation campaign.
http://greententacle.techfak.uni-bielefeld.de/~cunger/qald/
The StackExchange's dataset is a very rich one: https://archive.org/details/stackexchange
This is composed by all the public data from all platform's sites.
https://archive.org/details/stackexchange
Actually QALD also provides hybrid questions as well as questions from the biomedical domain. In the BioASQ project (http://bioasq.org) we also created benchmarks (domain: bio-medicine) with more than 800 questions.
I registered as a participant in bioasq.org.. How can i download the benchmark dataset
The SQuAD is one of the popular datasets in QA which is consist of some passages. Each question can be answered by finding the span of the text in the passage. So it is useful when you are building a QA model based on information retrieval (IR) and Reading Comprehension (RC):
https://rajpurkar.github.io/SQuAD-explorer/
HotpotQA is also a QA dataset and it is useful for multi-hop question answering when you need reasoning over paragraphs to find the right answer.
https://hotpotqa.github.io/
The other datasets:
https://stanfordnlp.github.io/coqa/
https://research.fb.com/downloads/babi/
https://github.com/brmson/dataset-factoid-webquestions
https://wiki.dbpedia.org/datasets
We published a list of large QA datasets at https://github.com/ad-freiburg/large-qa-datasets including links to the corresponding publication and dataset.
- Badges
- Science topic