QALD评测数据集的全称为Question Answering over Linked Data ,是多语言的链接数据问答系统的评测竞赛活动, 数据github:https://github.com/ag-sc/QALD 。
评测工具:http://gerbil-qa.aksw.org/gerbil/
QALD-9 中的问题相比之前更复杂,除了事实类问题,还包括:
- 计数问题, e.g., How many children does Eddie Murphy have?
- 最高级, e.g., Which museum in New York has the most visitors?
- 比较级,e.g., Is Lake Baikal bigger than the Great Bear Lake?
- 时间聚合, e.g., How many companies were founded in the same year as Google?
multilingual 数据集大小如下表
训练集 | 测试集 | |
---|---|---|
QALD-9 | 408 | 150 |
QALD-8 | 250 | 100 |
QALD-7 | 215 | 50 |
QALD-6 | 350 | 100 |
QALD-5 | 340 | 50 |
QALD-4 | 200 | 50 |
QALD-3 | 100 | 100 |
QALD-2 | 100 | 100 |
QALD-1 | 50 | 50 |