Lionel Nicolas
Lavinia Nicoleta Aparaschivei
Verena Lyding
Christos Rodosthenous
Federico Sangati
Alexander König
Corina Forascu
Download article
Published in: Proceedings of the 10th Workshop on Natural Language Processing for Computer Assisted Language Learning (NLP4CALL 2021)
Linköping Electronic Conference Proceedings 177:1, p. 1-14
NEALT Proceedings Series 47:1, p. 1-14
Published: 2021-05-21
ISBN: 978-91-7929-625-4
ISSN: 1650-3686 (print), 1650-3740 (online)
In this paper, we present an experiment performed with the aim of evaluating if knowledge of expert quality about Romanian synonyms could be crowdsourced from language learners by collecting and aggregating their answers to two types of questions that are automatically generated from a dataset encoding semantic relations between words. Such an evaluation aimed at confirming the viability of a fully learner-fueled crowdsourcing workflow for improving such type of dataset. For this experiment, we relied on an existing open-source crowdsourcing vocabulary trainer designed for this very purpose, but for which the crowdsourcing potential could be further evaluated, especially with regards to lesser-resourced languages such as Romanian. Our results confirmed that producing expert knowledge regarding Romanian synonyms could indeed be achieved in such a fashion. Additionally, we took the occasion to further evaluate the learning impact of the trainer on the participants and gather their feedback about the different aspects of the trainer.
