372 0

Misinformation Detection and Rectification using QA system with COVID-19

Title
Misinformation Detection and Rectification using QA system with COVID-19
Other Titles
QA 시스템을 활용한 코로나 19에 대한 가짜정보의 탐지 및 정정
Author
임인섭
Alternative Author(s)
임인섭
Advisor(s)
조남재
Issue Date
2021. 2
Publisher
한양대학교
Degree
Master
Abstract
As the number of COVID-19 patients increases, the number of misinformation on COVID-19 also increases rapidly, misleading people, and causing infection. WHO defined this phenomenon as “Infodemic” and has been trying to prevent it. The proposed model is to detect misinformation and to rectify it by only using true data with an Open-Domain Question-Answering system and cosine similarity. The testing environment is divided into a total of nine. First is the base model, a Closed-domain question-answering system next is an Open-domain question-answering system, choosing the top related answer, and the last is also an Open-domain question-answering system with voting top three related answers choosing the majority. Within these testing environments, three types of datasets are used, scientific datasets, social media and news datasets, total datasets combining both datasets. The top accuracy among all testing environments was 76% using total datasets with voting top three related answers based on Open-Domain model, outperforming the highest accuracy from Closed-Domain model of 59.2%, by over 16%. This unique model has shown the possibility of using the blend of QA system and text similarity on detection and rectification of misinformation.
URI
https://repository.hanyang.ac.kr/handle/20.500.11754/159562http://hanyang.dcollection.net/common/orgView/200000485616
Appears in Collections:
GRADUATE SCHOOL[S](대학원) > BUSINESS INFORMATICS(비즈니스인포매틱스학과) > Theses (Master)
Files in This Item:
There are no files associated with this item.
Export
RIS (EndNote)
XLS (Excel)
XML


qrcode

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

BROWSE