Twitterreko Euskal Komunitatearen Eduki Azterketa Pandemia Garaian

Abstract

Hizkuntzaren Prozesamenduak eskaintzen dituen teknika ez-gainbegiratuak erabiliz Twittereko euskal komu-nitatean COVID-19aren pandemiak izan duen eragina aztertzea da lan honen asmoa.Azterketa hau aurreraeramateko sare sozial horretako erabiltzaileen euskarazko txioak masiboki bildu eta denboraren arabera ordenatudira. Pandemiaren eragina neurtzeko, denboraren arabera edukia nola aldatu den aztertu da, horretarako testuetanazaltzen diren hitz zein emojien aldaketa kuantitatibo zein kualitatiboak baliatu dira. Azterketa kuantitatiboan,terminoek garai desberdinetan izan duten maiztasunaren aldaketari erreparatu zaio, maiztasunen erregresio linealaerabiliz. Azterketa kualitatiboan, hitzen bektore trinkoak baliatu dira, pandemiaren garai desberdinetan hitz etaemoji adierazgarrienek esanahian izan duten bilakaera aztertzeko.The aim of this work is to study the impact of the COVID-19 pandemic on the Basque Twitter community usingunsupervised techniques based on Natural Language Processing. In order to carry out this study, large quantitiesof tweets were gathered and sorted by time from Basque Twitter users. To analyze the impact of the pandemic,the variability of the content over time has been studied, through quantitative and qualitative changes in thewords and emojis that appear in the texts. In the quantitative analysis, the shift at the frequency of the termswas calculated using linear regression over frequencies. In the qualitative analysis, Word Embeddings were usedto study the changes in the meaning of the most significant words and emojis at different times during the pandemic.