Ranking Quality of Answers Drawn from Independent Evident Passages in Neural Machine Comprehension

Avi Bleiweiss

2018

Abstract

Machine comprehension has gained increased interest with the recent release of real-world and large-scale datasets. In this work, we developed a neural model built of multiple coattention encoders to address datasets that draw answers to a query from orthogonal context passages. The novelty of our model is in producing passage ranking based entirely on the answer quality obtained from coattention processing. We show that using instead the search-engine presentation order of indexed web pages, from which evidence articles have been extracted, may affect performance adversely. To evaluate our model, we chose the MSMARCO dataset that allows queries to have anywhere from no answer to multiple answers assembled of words both in and out of context. We report extensive quantitative results to show performance impact of various dataset components.

Download


Paper Citation


in Harvard Style

Bleiweiss A. (2018). Ranking Quality of Answers Drawn from Independent Evident Passages in Neural Machine Comprehension. In Proceedings of the 10th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management (IC3K 2018) - Volume 1: KDIR; ISBN 978-989-758-330-8, SciTePress, pages 61-71. DOI: 10.5220/0006923300610071


in Bibtex Style

@conference{kdir18,
author={Avi Bleiweiss},
title={Ranking Quality of Answers Drawn from Independent Evident Passages in Neural Machine Comprehension},
booktitle={Proceedings of the 10th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management (IC3K 2018) - Volume 1: KDIR},
year={2018},
pages={61-71},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0006923300610071},
isbn={978-989-758-330-8},
}


in EndNote Style

TY - CONF

JO - Proceedings of the 10th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management (IC3K 2018) - Volume 1: KDIR
TI - Ranking Quality of Answers Drawn from Independent Evident Passages in Neural Machine Comprehension
SN - 978-989-758-330-8
AU - Bleiweiss A.
PY - 2018
SP - 61
EP - 71
DO - 10.5220/0006923300610071
PB - SciTePress