Abstract:
In this article we present a deep learning question answering (QA) setting that can work for any natural language. We recognize the problem of low-resource languages, i.e...Show MoreMetadata
Abstract:
In this article we present a deep learning question answering (QA) setting that can work for any natural language. We recognize the problem of low-resource languages, i.e. most languages other than English, which lack appropriately sized datasets or cutting-edge NLP tools. To address this problem, we have designed and implemented a QA dataset and system that are independent from language use; specifically, we test our solution on the Polish language which is both low-resource and grammatically complex. Both these features make the task of QA significantly harder. To the best of our knowledge, this is the first attempt to train a deep learning QA system in a language-agnostic setting.
Published in: 2018 Fifth International Conference on Social Networks Analysis, Management and Security (SNAMS)
Date of Conference: 15-18 October 2018
Date Added to IEEE Xplore: 02 December 2018
ISBN Information: