Evaluating Long Contexts in the Czech Answer Selection Task


This publication doesn't include Faculty of Arts. It includes Faculty of Informatics. Official publication website can be found on muni.cz.

MEDVEĎ Marek SABOL Radoslav HORÁK Aleš

Year of publication 2021
Type Article in Proceedings
Conference Recent Advances in Slavonic Natural Language Processing (RASLAN 2021)
MU Faculty or unit

Faculty of Informatics

Keywords Question answering; Answer selection; Answer context; Evaluation
Description In the search for the answer to an open-domain question, the size of the search window, or the answer context, can greatly influence the resulting determination of the answer. The presented paper offers a detailed evaluation of different sizes of the answer context in case of Czech question answering. We compare six different context types in four different lengths. The conclusion of the experiments is that prolonging the context can improve the precision for specific types but in general the best results are obtained with one-sentence contexts.
Related projects:

You are running an old browser version. We recommend updating your browser to its latest version.