Knowledge Graph Question Answering (KGQA) is a popular task among NLP enthusiasts, however, the KG used is generally open-domain and contains general knowledge. For example, Wikidata, DBpedia and Freebase are the most popular general purpose KGs. In this challenge, we focus on a special class of domain specific KGs, i.e., the scholarly domain. These KGs contain information about research papers, authors, research institutions etc., in a graph format. In particular, we use the DBLP and the SciQA KGs in this challenge. There were three tasks in this challenge: Task 2: For this task, participants had to perform question answering over the DBLP_QuAD dataset.
papers were accepted for publication, and are a part of this proceedings. System submissions were made on the codalab platform, and the evaluation was performed in an automated fashion. Based on the results of the evaluation, the following papers were selected as winners:
“When context matters: Entity Linking in the Scholarly Domain” - Nadine Steinmetz Task 2: DBLP_QuAD Question Answering “NLQxform: A Language Model-based Question to SPARQL Transformer” Ruijie Wang, Zhiruo Zhang, Luca Rossetto, Florian Ruosch, Abraham Bernstein
“A Structure and Content Prompt-based Method for Knowledge Graph Question Answering over Scholarly Data” - Longquan Jiang, Xi Yan, Ricardo Usbeck
In this section, we list the people who organized and contributed to the success of this event.
Challenge Chairs ● Debayan Banerjee ● Sushil Awale ● Ricardo Usbeck ● Nandana Mihindukulasooriya ● Mohamad Yaser Jaradeh ● Sören Auer Challenge Program Committee Members The challenge program committee helped review the submitted papers. The organizers would like to thank them for their valuable time.
■ Mohnish Dubey ■ Debanjan Chaudhuri ■ Xi Yan ■ Cedric Möller ■ Junbo Huang