244 0

Personal Information De-identification System using Sentence Classification and Named Entity Recognition

Title
Personal Information De-identification System using Sentence Classification and Named Entity Recognition
Author
서동국
Alternative Author(s)
서동국
Advisor(s)
이동호
Issue Date
2021. 2
Publisher
한양대학교
Degree
Master
Abstract
Recently, unstructured text documents containing personal information have been leaked or disclosed causing critical damage to not only people whose information was leaked but also companies which collected the information. To prevent data leakage and utilize data, the process of personal information detection and de-identification is essential. Data can be classified into two categories structured data and unstructured data. It is much difficult to detect and de-identify personal information in unstructured data in terms of cost and accuracy rather than structured data. To solve these difficulties, there have been recent studies to automate the de-identification process using deep learning models showing visible performance. However, these studies have a common problem that, since personal information was detected only by relying on named entity information without considering the ambiguity of words, even words that are not subject to non-identification can be de-identified. As a result, it hinders the usefulness of data. To overcome the limit of existing studies, this paper proposes a personal information detection model that uses sentence intent information as additional information in the course of named entity recognition learning process and a de-identification technique considering the usefulness of personal information data. Finally, we verified the excellence of the models and the systems proposed in this paper comparing the recent studies through experiments.
URI
https://repository.hanyang.ac.kr/handle/20.500.11754/158951http://hanyang.dcollection.net/common/orgView/200000485662
Appears in Collections:
GRADUATE SCHOOL[S](대학원) > COMPUTER SCIENCE & ENGINEERING(컴퓨터공학과) > 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