README.md 1.14 KB
Newer Older
Nadine Steinmetz's avatar
Nadine Steinmetz committed
1
2
3
4
The files are structured as following:
- first column: id as in the respective benchmark dataset
- second column: the question string in English language
- third column: the detected answer type
Nadine Steinmetz's avatar
Nadine Steinmetz committed
5

Nadine Steinmetz's avatar
Nadine Steinmetz committed
6
7
8
9
10
11
12
13
14
15
16
17
18
19
The answer types are detected automatically according to the information provided in the dataset for each question. We identified 13 different answer types:
- date
- boolean
- string
- number count -> a number originating from a COUNT operator in the SPARQL query
- number prop -> a number originating from a property
- number unknown -> a number, but the origin is unknown (when no SPARQL query is provided)
- resource list typed -> a list of resources with a specific type set in the SPARQL query
- resource list untyped -> a list of resources without a specific type
- resource list unknown -> a list of resources, but a specific type is unknown (when no SPARQL query is provided)
- resource typed -> one resource with a specific type set in the SPARQL query
- resource untyped -> one resource without specific type
- resource unknown -> one resource, but a specific type is unknown (when no SPARQL query is provided)
- unknown -> the answer type could not be detected