A Dataset for Event-Centric Question Answering

Current Version 3.0
Questions 1000
Events 1005
Entities 1655
Predicates 309


Event-QA is a Dataset for answering Event-Centric questions over Knowledge Graphs (KGs). We target EventKG, a recently proposed Event-Centric KG. In our paper we describe details about the dataset creation process. Current datasets are mostly entity-centric, Event-QA is the first Question Answering dataset that focus on events.

The Natural Language questions are manually created through a Query Verbalisation Interface, where for each SPARQL query, the users had to write the question representation. The current version of Event-QA contains questions in English, Portuguese and German.

The Event-QA generation pipeline is done by random walk sub-graph traversal. Details of how to configure and execute the pipeline can be accessed in Event-QA GitHub project website.

Dataset VoID description


Cite as
   title={Event-QA: A Dataset for Event-Centric Question Answering over Knowledge Graphs},
   author={Souza, Tarcisio, Gottschalk, Simon and Demidova, Elena},
Examples of Event Queries covered by Event-QA
  • In what event hosted in the United States in 2004 was Lose My Breath recorded?
    • The 2004 NBA Playoffs
  • Which award organized by the Netherlands Film Festival in 1981 was received by Hannah Hoekstra?
    • The Golden Calf award
  • When did the Breakup of Yugoslavia started, which had Stjepan Mesić as one of the main characters and also triggered Yugoslav Wars?
    • 1991-06-25

Copyright ©2018-2019 L3S Research Center.

All Rights Reserved.