Explaining Inconsistency-Tolerant Query Answering over Description Logic Knowledge Bases
Authors: Meghyn Bienvenu, Camille Bourgaux, François Goasdoué
AAAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Finally, we empirically study the efficiency of our explanation framework using the well-established LUBM benchmark. |
| Researcher Affiliation | Academia | Meghyn Bienvenu CNRS, Univ. Montpellier, Inria Montpellier, France Camille Bourgaux Univ. Paris-Sud, CNRS Orsay, France Franc ois Goasdou e Univ. Rennes 1, CNRS Lannion, France |
| Pseudocode | No | The paper defines formal concepts and presents complexity analysis but does not include structured pseudocode or algorithm blocks. |
| Open Source Code | No | We implemented our explanation framework in Java within our CQAPri system (www.lri.fr/ bourgaux/CQAPri), which supports querying of DL-Lite R KBs under several inconsistency-tolerant semantics, including the brave, AR and IAR semantics. |
| Open Datasets | Yes | We used the CQAPri benchmark available at www.lri.fr/ bourgaux/CQAPri, which builds on the DL-Lite R version (Lutz et al. 2013) of the Lehigh University Benchmark (swat.cse.lehigh.edu/projects/lubm). |
| Dataset Splits | No | The paper describes generating inconsistent ABoxes from a consistent database for experiments but does not provide specific details on training, validation, or test dataset splits. |
| Hardware Specification | Yes | Our hardware is an Intel Xeon X5647 at 2.93 GHz with 16 GB of RAM, running Cent OS 6.7. |
| Software Dependencies | Yes | We used the SAT4J v2.3.4 SAT solver (www.sat4j.org) to compute MUSes and cardinality-minimal models (Berre and Parrain 2010). |
| Experiment Setup | Yes | CQAPri classifies a query answer a into one of 3 classes: Possible: K |=brave q( a) and K |=AR q( a) Likely: K |=AR q( a) and K |=IAR q( a) (Almost) sure: K |=IAR q( a) [...] We used a consistent database with 100 universities (more than 10 million assertions) from which we generated seven inconsistent ABoxes with different ratios of assertions in conflicts by adding from 8005 to 351724 assertions. [...] Reported times are averaged over 5 runs. |