Predicting the Outcome of the Complaints Submitted to the Press Council Using Machine Learning
Name
Anne-Liis Rämson
Abstract
This thesis discusses complaints submitted to the Press Council between 2001 and 2021 and media texts in accordance with the complaints. The aim of this thesis is to provide statistical overview about the complaints lodged with the Press Council and mentioning the elements of the Code of Ethics in the decisions appointed by the Press Council, apply the methods of classification to media texts in line with the complaints and to identify a classification model that would distinguish acquittal and reprehensible media texts. The theoretical part of this thesis gives an overview about text mining, classification models (logistic regression, Support Vector Machine, fastText) and evaluation methods for classification. The analysis of complaints revealed that there were two main groups of the largest media outlets in Estonia with regard to mentioning the key elements of the Code of Ethics. It was found that three media outlets received more mentions about section 4.2 of the Code of Ethics and three media outlets received most mentions about section 1.4. Reprehensible decisions were best predicted by the fastText classifier on lemmatized texts.
Graduation Thesis language
Estonian
Graduation Thesis type
Master - Conversion Master in IT
Supervisor(s)
Kairit Sirts
Defence year
2022