Richter, Michael Paul RobertWerzner Regalado, JordanyJordanyWerzner Regalado2023-07-182023-07-182021https://jlupub.ub.uni-giessen.de/handle/jlupub/17802http://dx.doi.org/10.22029/jlupub-17177In 2020 the effects of the COVID-19 pandemic and the related government-applied nationwide measures deeply influenced the Italian and Spanish population, not only financially and socially, but also in terms of the emotional state of all individuals. Due to social isolation, social media and other online communication platforms served as an outlet to express opinions and feelings about the situation and to interact with other peers. This resulted in the availability of a great amount of online data that can be utilised to analyse the sentiment and emotions expressed during this time. The aim of this study was to use this data to perform a context-based sentiment and emotion analysis in order to determine a relationship between shifting changes of sentiment and emotion and the application restriction measures in Italy and Spain. Most current studies focus on English data extracted from Twitter and show little interest towards other data sources and other languages. This thesis therefore based its analysis on over 700 000 Spanish and Italian comments extracted from the newspapers La Repubblica and El Mundo using a Pythonbased web crawler. The created press corpora represent another user demographic and can be used for the analysis of the Italian and Spanish language in other research areas. A sentiment classification algorithm based on one-layer convolutional network was used to determine the polarity in the comments. The classification method achieved an F-Score of 0.87 for the Spanish language model and an F-Score of 0.81 for the Italian model. The emotion detection was performed using the Syuzhet R Package and NRC Emotion Lexicon to create emotion scores during different time frames. Using a graphical analysis the study determined an existing emotional reaction, that could be put in relation to the measures applied by the governments. Furthermore, utilising Pearson’s correlation coefficient, it was determined that Robert Plutchik’s theory of emotional opposites applies to the analysed context. Moreover, a negative correlation was detected between the level of trust and the level of fear expressed in comments mentioning the government. All results were shown in a contrastive manner to compare the emotional reactions in the comment sections. The insights gained from this study can be used for cultural, linguistic and political analysis, political decision-making and for the development of strategies to manage the pandemic and other national catastrophes, while considering the emotional state of the population.enAttribution-NonCommercial-NoDerivatives 4.0 InternationalSentiment AnalysisEmotion DetectionCoronavirusCovid-19EmotionsanalyseStimmungsanalyseCOVID-19 pandemicCOVID-19 Pandemiepandemic managementpress corporadata analysisContextual emotion analysisemotional statesocial isolationsentiment classification algorithmconvolutional networkemotional reactionpolitical analysiscultural analysisweb crawlingCorpuscorporaCorpus analysisSentiment miningcomputational linguisticssentiment lexiconscross-cultural analysissentiment variationsweb scrapingpublic opiniononline discoursesocial distancingpandemic responseddc:004ddc:070ddc:300ddc:310ddc:360ddc:400ddc:450ddc:460