|
Title
|
|
|
|
Exploring the classification of security events using sparse and dense representation of text
| |
|
Author
|
|
|
|
| |
|
Abstract
|
|
|
|
| With the availability of open source intelligence systems, such as the Integrated Crisis Early Warning System (ICEWS), news stories are constantly being monitored and analyzed However, with the increasing amount of data available on the Internet, automated methods such as machine learning can be used to quantify and speedup analysis. Although projects such as ICEWS have a globalfocus, it is important that in South Africa, we focus on the local situation. Using WhatsApp as our new source, we evaluated a machine learning approach to automatically classify violent events, such as protests of mass violence, by using dense vector representations. Our experimental results indicated that dense vector representation did not necessarily improve the performance of the machine learning classifier. |
| |
|
Language
|
|
|
|
English
| |
|
Source (journal)
|
|
|
|
2020 INTERNATIONAL SAUPEC/ROBMECH/PRASA CONFERENCE
| |
|
Source (book)
|
|
|
|
International SAUPEC/RobMech/PRASA Conference, JAN 29-31, 2020, Cape Town, SOUTH AFRICA
| |
|
Publication
|
|
|
|
New york
:
Ieee
,
2020
| |
|
ISBN
|
|
|
|
978-1-72814-162-6
| |
|
DOI
|
|
|
|
10.1109/SAUPEC/ROBMECH/PRASA48453.2020.9041092
| |
|
Volume/pages
|
|
|
|
(2020)
, p. 143-148
| |
|
ISI
|
|
|
|
000583042800026
| |
|
Full text (Publisher's DOI)
|
|
|
|
| |
|
Full text (publisher's version - intranet only)
|
|
|
|
| |
|