TY - GEN
T1 - Multi-scale Sentiment Analysis of Location-Enriched COVID-19 Arabic Social Data
AU - Elsaka, Tarek
AU - Afyouni, Imad
AU - Hashem, Ibrahim Abaker Targio
AU - AL-Aghbari, Zaher
N1 - Publisher Copyright:
© 2021, Springer Nature Switzerland AG.
PY - 2021
Y1 - 2021
N2 - After the recent outbreak of COVID-19, researchers have risen working on several challenges related to the mining of social data to learn about people’s reactions to the epidemic. Recent studies have largely focused on extracting current themes and inferring broad attitudes, with a particular emphasis on the English language. This study presents various perspective for Arabic social data mining to provide in-depth insights related to the COVID-19 pandemic. We initially devised a method for inferring geographical whereabouts from Arabic tweets not initially geotagged. Secondly, a sentiment analysis mechanism based on Arabic word embeddings is introduced, with several levels of geographical granularity (regions/countries) considered. Sentiment-based classifications of topics and subtopics related to COVID-19 will also be presented. According to our findings, the overall percentage of location-enabled tweets has increased from 2% to 46% (about 2.5M tweets). During the pandemic, Arab Twitter users’ negative emotions about lockdown, restriction, and law enforcement were also widely expressed.
AB - After the recent outbreak of COVID-19, researchers have risen working on several challenges related to the mining of social data to learn about people’s reactions to the epidemic. Recent studies have largely focused on extracting current themes and inferring broad attitudes, with a particular emphasis on the English language. This study presents various perspective for Arabic social data mining to provide in-depth insights related to the COVID-19 pandemic. We initially devised a method for inferring geographical whereabouts from Arabic tweets not initially geotagged. Secondly, a sentiment analysis mechanism based on Arabic word embeddings is introduced, with several levels of geographical granularity (regions/countries) considered. Sentiment-based classifications of topics and subtopics related to COVID-19 will also be presented. According to our findings, the overall percentage of location-enabled tweets has increased from 2% to 46% (about 2.5M tweets). During the pandemic, Arab Twitter users’ negative emotions about lockdown, restriction, and law enforcement were also widely expressed.
KW - Arabic COVID-19
KW - Arabic tweets
KW - COVID-19 pandemic
KW - Sentiment Analysis
KW - Social data mining
UR - http://www.scopus.com/inward/record.url?scp=85118145166&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85118145166&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-88942-5_15
DO - 10.1007/978-3-030-88942-5_15
M3 - Conference contribution
AN - SCOPUS:85118145166
SN - 9783030889418
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 194
EP - 203
BT - Discovery Science - 24th International Conference, DS 2021, Proceedings
A2 - Soares, Carlos
A2 - Torgo, Luis
PB - Springer Science and Business Media Deutschland GmbH
T2 - 24th International Conference on Discovery Science, DS 2021
Y2 - 11 October 2021 through 13 October 2021
ER -