Skip to content

A machine learning pipeline for detecting Slovenian political bias based on sentiment towards various topics and political parties.

License

Notifications You must be signed in to change notification settings

martinkorelic/slovenian-political-bias

Repository files navigation

Analysis of topical political stance of Slovene tweets

PDF available at repozitorij.uni-lj.si.

Družbena omrežja omogočajo prosto javno izražanje političnih mnenj uporabnikov, ki zagovarjajo različna stališča glede aktualnih političnih vprašanj. V diplomski nalogi smo analizirali politično usmerjenost oziroma pristranskost slovenskih uporabnikov na podlagi njihovih objav na družbenem omrežju Twitter. Pri tem smo uporabili metode za obdelavo naravnega jezika. Z uporabo algoritma BERTopic smo poiskali in iz podatkovne množice izluščili različne politično družbene teme in jih uporabili pri analizi sentimenta za klasifikacijo politične usmerjenosti (levo, desno, nevtralno). Opazimo precejšen delež negativnega sentimenta do vseh tem in strank. Količina levo in desno usmerjenih tvitov v političnih temah obeh polov je približno enaka. Zaznamo, da v tvitih po priljubljenosti najbolj izstopata dve stranki, vsaka iz nasprotnega političnega pola.

Social networks allow free public expression of users' political opinions, advocating various views on the current political agenda. In the thesis, we analyzed the political orientation of Slovene users' posts on the Twitter social network. We used the BERTopic algorithm to find and extract political topics from the data and applied sentiment analysis to classify political orientation (left, right and neutral). The results show a significant proportion of negative sentiment towards all topics and parties. The amount of left- and right-leaning tweets on general political topics is approximately equal. We notice that two parties from opposite political poles stand out in tweet popularity.

Data

Data is available in /data sorted by the topics in each .json file:

  • Labeled topics: /data/labelled topics
  • Final postprocessed data: /data/final/

BibText citation:

@phdthesis{KORELIČ_2022,
  title={Analiza tematske politične usmerjenosti slovenskih tvitov},
  url={https://repozitorij.uni-lj.si/IzpisGradiva.php?lang=slv&id=142496},
  author={KORELIČ, MARTIN}, year={2022}
}

About

A machine learning pipeline for detecting Slovenian political bias based on sentiment towards various topics and political parties.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published