Predicting the Outbreak of Conflict in Online Discussions Using Emotion-Based Features
[ 1 ] Instytut Informatyki, Wydział Informatyki i Telekomunikacji, Politechnika Poznańska | [ D ] doktorant | [ P ] pracownik
2020
rozdział w monografii naukowej / referat
angielski
- online discussion
- emotions
- conflictual interactions
EN Anti-social online behaviour, such as harassment or vulgarity, leading to conflicts aimed at destroying any merit of the discussions, is a serious problem for the Internet community. Recognising the characteristics of conflict discussions and modelling their trajectory might help to predict and prevent derailing. My PhD thesis focuses on using emotion labels as such characteristics and building an explainable prediction model based on them. As a part of the thesis we have proposed a new dataset of discussions containing knowledge about their emotion-based features. It is a set of dialogues from Wikipedia Talk Pages annotated during a crowdsourcing experiment with labels from Plutchik’s model of emotions and described with EmoWordNet lexicon scores. With this explainable model we hope to introduce a new way of moderating Internet discussions and provide useful educational tool.
10.06.2020
505 - 511
20
70