No Detail Found!

The limits of context simplification on QA system results

Author(s)
RACHID KARRA LASTIMI Laboratory, Mohammadia School Of Engineers, Mohammed V University in Rabat, Morocco
ABDELATIF LASFAR LASTIMI Laboratory, Mohammadia School Of Engineers, Mohammed V University in Rabat, Morocco
Abstract
QA systems play a key role in human activities such as customer support, digital assistance in education, health, and public services. Our work is to use a QA system as a black box and see the effect of different simplification models on its results. The present study explores how far state-of-art simplification models can conserve text content. We measure text complexity with different linguistic metrics and meaning conservation with a BERT-based QA system score. Through evaluations, we measured text complexity and proved that context simplification as a multi-step simplification process gives better results in QA systems than 'direct' or 'whole' simplification. The proposed method has a better performance compared to automatic simplification. It is beneficial for a QA system with changeable contexts. As a task-oriented feature, choosing the convenient text simplification system should depend on its usefulness and the nature of the problem.
Publication Details
Page(s) 6789-6797
DOI DOI not available
Published Journal: Journal of Theoretical and Applied Information Technology, Volume: 100, Issue: 22, Year: 2022

Request the full text from PASTIC

Keywords
Linguistic analysis, QA system, Text simplification., BERT, Seq2Seq
0

Citations

0

Downloads

38

Views