I am really excited participating in the #DHgoesVIRAL with my presentation entitled: #κορονοιος: analyzing tweets containing the greek coronavirus hashtag in the first month of the COVID19 pandemics in Greece. (1/19)

Few words about me. I am currently a Professor of Computational and Quantitative Linguistics at @CHSS_HBKU in Qatar. My research focuses mainly on Computational Stylistics and Text Mining applied to a variety of fields ranging from #DH to #Forensic_Linguistics #DHgoesVIRAL (2/19)

Twitter data can provide a wealth of information related to a wide spectrum of human behavior and can be used among others to model social interaction and predict social trends, crime, election outcomes, new movie revenues, language change, and even pandemics. #DHgoesVIRAL (3/19)

Twitter can be used as a real-time content and public attention trend-tracking tool in emergency situations such as the current #COVIDー19 outbreak. Big data analytics and #NLP methods can help us study critical aspects of public health on social media #DHgoesVIRAL (4/19)

The present study analyzes the hashtag #κορονοιος (greek coronavirus) in all the tweets written in greek and appeared since 26/2/2020 (first confirmed #COVIDー19 case in Greece) until 23/3/2020 covering the first month of the outbreak in Greece. #DHgoesVIRAL (5/19)

During this period we collected 88,354 unique tweets (retweets excluded) written in Greek using the hashtag #κορονοιος. The corpus was cleaned using a standard #NLP preprocessing pipeline. The number of tweets per day since 26/2/2020 can be seen in this image. #DHgoesVIRAL (6/19)

A first hypothesis tested was whether the daily number of tweets correlates with the #COVIDー19 confirmed cases in #Greece. A strong correlation detected (Spearman rho = 0.84) confirming the relationship between frequency of tweeting and actual cases found. #DHgoesVIRAL (7/19)

We modeled further this relationship using linear regression. Since both variables (n of tweets and n of cases) were heavily skewed we applied ln transformation getting a good fit (R2=0.63) & confirming the interaction of physical and virtual world (fig below) #DHgoesVIRAL (8/19)

A vocabulary analysis followed producing a word-cloud of the most frequent words (filtered by a standard greek stopword list). The figure below is based on the 100 most frequent words of the corpus. #DHgoesVIRAL (9/19)

In addition, we plotted the frequency of terms related to the #COVIDー19 pandemic. The figure above depicts the relative frequency of three countries that were mentioned often by the Greek users, i.e. Greece, Italy and China #DHgoesVIRAL (10/19)

Interesting conclusions can also be extracted by looking at the collocates of the terms "Greece" and "Italy". It seems that Greek Twitter users are discussing these two countries with comparable, disaster-oriented vocabulary #DHgoesVIRAL (11/19)

Using #VoyantTools module Dreamscape we explored how tweets might be represented geo-spatially. The tool identifies locations (city names) mentioned in the tweets and plots a map of cities that are frequently mentioned in the corpus (fig. below) #DHgoesVIRAL (12/19)

The last analysis performed in this corpus was topic modeling. We extracted the 7 most prevalent topics and each topic was represented by 15 terms. The figure below displays topic distribution across each day of the analyzed time period. #DHgoesVIRAL (13/19)

Looking at the intensity of the topic modeling graph we can easily detect how specific topics emerged at specific dates. E.g. the topic related to staying at home directive appeared after 14/3/2020 when the lockdown in Greece was enforced #DHgoesVIRAL (14/19)

#NLP methods applied to Twitter data can reveal interesting interaction patterns between our physical and our online worlds The analysis of the Greek data confirms previous research findings that correlate social media content analysis and public health facts #DHgoesVIRAL (15/19)

The presented research contributes to "infodemiology", the application of text mining methods in social media for analyzing critical public health issues. Greece can also employ these advanced tools to improve decision-making process during the current crisis #DHgoesVIRAL (16/19)

I would like to thank everyone for reading this thread and of course the organizers of this amazing e-conference ( @athenaRICinfo, @agiati). You can send me any questions on Twitter ( @gmikros) or email me at [email protected]. #DHgoesVIRAL (17-19/19) Keep smiling !!! :-)