Laukemper, Anton (2020) Classifying Political YouTube Channels With Semi-Supervised Learning. Master's Thesis / Essay, Computing Science.
|
Text
Laukemper_Master_Thesis_Final.pdf Download (2MB) | Preview |
|
![[img]](https://fse.studenttheses.ub.rug.nl/style/images/fileicons/text.png) |
Text
toestemming.pdf Restricted to Registered users only Download (98kB) |
Abstract
Since many investigative journalists accused YouTube’s video recommendation system of recommending radical political and conspiratorial content, the inner workings of this algorithm have become the focus of public scientific research. Multiple studies were carried out to analyse the bias of the recommendation engine with conflicting results. An important drawback of all of these studies was that the core feature of the algorithm - personalized recommendations - was not investigated. The overall aim of this Master project was to create an experiment that would provide a dataset of personalized video recommendations with which a potential bias could be detected. The large time cost of manual classification of such a dataset necessitates algorithmic classification. Due to the easy availability of unlabeled data, semi-supervised learning seems like an attractive approach for this task. The computational research question to be addressed in this thesis was therefore whether unlabeled data improves performance of machine learning classifiers if used in a semi-supervised learning algorithm. To this end, multiple classifiers were trained in a supervised manner for a binary classification task, distinguishing political from nonpolitical YouTube channels, and a multi-label classification task, classifying channels according to a number of political categories. The classifiers were compared to classifiers trained using the semi-supervised Self-Training algorithm.
| Item Type: | Thesis (Master's Thesis / Essay) |
|---|---|
| Supervisor name: | Biehl, M. and Jaeger, H. |
| Degree programme: | Computing Science |
| Thesis type: | Master's Thesis / Essay |
| Language: | English |
| Date Deposited: | 28 Aug 2020 13:05 |
| Last Modified: | 28 Aug 2020 13:05 |
| URI: | https://fse.studenttheses.ub.rug.nl/id/eprint/23253 |
Actions (login required)
 |
View Item |