Understanding what a person is asking via a question is one of the first steps that humans use to find the corresponding answer. The same is true for Knowledge Graph Question Answering (KGQA) systems. In this regard, expected answer type (EAT) classification is an important part of the Question-Answering pipeline within such systems. Our previous research on EAT classification [1, 2] was mainly focused on data augmentation techniques including multilingual data. In this demo, we extend our approach presented within SMART Task @ ISWC 2020 by predicting the most specific EAT and fetching its hierarchy from DBpedia. The provided functionality enables end-users to get the EAT predictions for 104 languages, see confidence of the prediction, and leave feedback. In addition, the API enables researchers and developers to integrate the EAT classification into their systems.
The ability to know the EAT may significantly narrow down the answer search space. In the case of the question: "In what city was Angela Merkel born?" is being asked to a KGQA system over DBpedia, the EAT classifier can lower down the search space from 861 to 6 entities (see the illustration below).
@inproceedings{DBLP:conf/semweb/PerevalovB21,
author = {Aleksandr Perevalov and
Andreas Both},
editor = {Oshani Seneviratne and
Catia Pesquita and
Juan Sequeda and
Lorena Etcheverry},
title = {Hierarchical Expected Answer Type Classification for Question Answering},
booktitle = {Proceedings of the {ISWC} 2021 Posters, Demos and Industry Tracks:
From Novel Ideas to Industrial Practice co-located with 20th International
Semantic Web Conference {(ISWC} 2021), Virtual Conference, October
24-28, 2021},
series = {{CEUR} Workshop Proceedings},
volume = {2980},
publisher = {CEUR-WS.org},
year = {2021},
url = {http://ceur-ws.org/Vol-2980/paper349.pdf},
timestamp = {Mon, 25 Oct 2021 15:03:56 +0200},
biburl = {https://dblp.org/rec/conf/semweb/PerevalovB21.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}