Queries

The original NPD SPARQL query set contains 25 queries obtained by interviewing users of the NPD dataset. Starting from the original NPD query set, we devised 14 queries having different degrees of complexity. The queries have been adapted in order to work in currently implemented obda systems, e.g. we removed aggregates since they are currently not supported by any obda system. The query set will be updated according to developments in the area.

In the obda context, it is recognized that queries involving classes (or object/data properties) with a rich hierarchy, or making use of existentially quantified variables in a rather sophisticated way (i.e., giving rise to tree witnesses) are harder to answer than other queries. In particular, observe that most complex queries involve both classes with a rich hierarchy and tree witnesses, which means that they are particularly suitable for testing the reasoner capabilities.

The following table reports statistics for each query in the benchmark.

Queries Statistics

query | #tot_op | #join | #depth | #tw | max(#subclasses) | # opt |
------|--------|-------|--------|-----|------------------|-------|
**q1**|   12   |   4   |   8    |  0  |        0         | 0     |
**q2**| 14 | 5 | 9 | 0 | 0 | 0 |
**q3** | 10 | 3 | 7 | 0 | 0 | 0 | 
**q4** | 14 | 5 | 9 | 0 | 0 | 0 |
**q5** | 14 | 5 | 8 | 0 | 0 | 0 |
**q6** | 18 | 6 | 12 | 2 | 23 | 0 |
**q7** | 17 | 7 | 10 | 0 | 0 | 0 |
**q8** | 10 | 3 | 7 | 0 | 0 | 0 |
**q9** | 10 | 3 | 7 | 0 | 38 | 0 |
 **q10** | 9 | 2 | 7 | 0 | 0 | 0 |
 **q11** | 20 | 7 | 12 | 2 | 23 | 0 |  
**q12** | 26 | 8 | 12 | 4 | 23 | 0 |
 **q13** | 8 | 2 | 7 | 0 | 0 | 2 |
**q14** | 12 | 2 | 7 | 0 | 0 | 2 |

Legenda

• #tot_op: Total number of SPARQL operators
• #join: Total number of JOIN operators
• #depth: Depth of the SPARQL algebra tree
• #tw: Number of tree-witnesses
• max(#subclasses): Maximum number of subclasses (resp., subrole) for a concept (resp., role) in the query
• opt: Number of optionals