SPARQLtoUser: Did the question answering system understand me?

Dennis Diefenbach (1) , Youssef Dridi , Kamal Singh (1) , Pierre Maret (1)

1 LHC - Laboratoire Hubert Curien

In Question Answering (QA) systems, user questions are automatically converted into SPARQL queries. We present SPARQLtoUser, a simple method to produce a user understandable version of a SPARQL query so that the users can check what the QA system understood. Compared to existing similar methods, SPARQLtoUser is multilingual and is not restricted to one single KB, but easily portable to any KBs. This paper presents our method to generate the representation of SPARQL queries, and compares this method to existing ones. Moreover, we present the results of our user study that show that users have difficulties to understand whether the responses of a QA system are correct or not, with or without SPARQL representations.

Mots clés: Question answering system; User feedback; User interaction; SPARQL- ToUser; SPARQL2NL; Spartiqulator

Domaines: Informatique [cs]

Dennis Diefenbach, Youssef Dridi, Kamal Singh, Pierre Maret. SPARQLtoUser: Did the question answering system understand me?. ISWC 2017, Oct 2017, Vienne, Austria. ⟨hal-01637139⟩