VQuAnDa - Verbalization Question Answering Dataset

Background

VQuAnDa is knowledge base QA dataset based on LC-QuAD which uses DBpedia v04.16 as the target KB.

This QA task consists of two components:

1. A Text2Sparql task where a natural language query is translated to a SPARQL query.
2. An RDF triple to verbalized answer task where the knowledge base query result must be translated back into natural language.

The dataset is in JSON format, and it contains 5000 examples (4000 train/1000 test). An example is shown below. The query result is surrounded by brackets [] in the verbalized answer.

{
  "uid"               :                                              "3508"
  "question"          :       "How many shows are aired on Comedy Central?"
  "verbalized_answer" : "There are [73] television shows broadcasted by..."
  "query"             : "SELECT DISTINCT COUNT(?uri) WHERE {?uri <http:..."
}

Baseline models

Alongside the dataset, the authors provide some baseline models. Here you can find the baseline implementations and instructions for how to run them.

License

The dataset is under Attribution 4.0 International (CC BY 4.0)

Cite

@InProceedings{kacupaj2020vquanda,
    title={VQuAnDa: Verbalization QUestion ANswering DAtaset},
    author={Kacupaj, Endri and Zafar, Hamid and Lehmann, Jens and Maleshkova, Maria},
    booktitle={The Semantic Web},
    pages={531--547},
    year={2020},
    publisher={Springer International Publishing},
}