ComplexWebQuestions[1] is a famous extension of the WebQuestions dataset, introduced to solve the second issue of few complex multi-relational questions in the WebQuestions. It is created by adding more constraints and multi-hop relations into the SPARQL queries from WebQuestionsSP, and then automatically generating questions via templates and Amazon Mechanic Turk.
This dataset can be downloaded via the link.
| Year | Type | F1 | Acc | Reported by | Official Repo |
|---|---|---|---|---|---|
| 2021 | IR-based | - | 47.4 | Feng. et al | Repo |
| 2021 | IR-based | - | 48.8 | He. et al | Repo |
| 2020 | SP-based | 40.4 | - | Lan. et al | Repo |
| 2019 | SP-based | 33.9 | - | Bhutani. et al | - |
| 2019 | SP-based | 31.6 | - | Sun. et al | - |
| 2019 | IR-based | - | 45.9 | Sun. et al | - |
| 2018 | IR-based | - | 32.8 | Sun. et al | Repo |
| 2016 | IR-based | - | 21.1 | H. Miller. et al | - |
[1] Talmor, Alon, and Jonathan Berant. The web as a knowledge-base for answering complex questions. arXiv preprint arXiv:1803.06643 (2018).