In this paper we describe an extended version of the QALD dataset, a well-known benchmark resource used for the task of Question Answering over knowledge graphs (KGQA). Along the lines of similar projects, the purpose of this work is to make available a) high-quality data even for languages other than English, and b) multiple reformulations of the same question, to test systems' robustness. The QALD version we used is the one released for the 9th edition of the challenge of Question Answering over Linked Data, and the languages involved are English and Italian. Besides a revised and improved quality of Italian translations of questions, the resource presented here features a number of alternative rewordings of both English and Italian questions. The usability of the resource has been tested on the QAnswer multilingual Question Answering system and through the GERBIL platform. The resource has been publicly released for research purposes.
rewordQALD9: A Bilingual Benchmark with Alternative Rewordings of QALD Questions
Sanguinetti M.
Primo
;Atzori MaurizioSecondo
;Puddu N.Ultimo
2022-01-01
Abstract
In this paper we describe an extended version of the QALD dataset, a well-known benchmark resource used for the task of Question Answering over knowledge graphs (KGQA). Along the lines of similar projects, the purpose of this work is to make available a) high-quality data even for languages other than English, and b) multiple reformulations of the same question, to test systems' robustness. The QALD version we used is the one released for the 9th edition of the challenge of Question Answering over Linked Data, and the languages involved are English and Italian. Besides a revised and improved quality of Italian translations of questions, the resource presented here features a number of alternative rewordings of both English and Italian questions. The usability of the resource has been tested on the QAnswer multilingual Question Answering system and through the GERBIL platform. The resource has been publicly released for research purposes.File | Dimensione | Formato | |
---|---|---|---|
semantics 2022.pdf
accesso aperto
Tipologia:
versione editoriale (VoR)
Dimensione
211.73 kB
Formato
Adobe PDF
|
211.73 kB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.