| dc.description.abstract | Online education has become a popular education form in recent years, with its use increasing massively during the COVID-19 pandemic. Neddy is a start-up company created at the start of the COVID-19 pandemic with the aim of making a learning tool for teachers facing distance learning for the first time. They created Addito, an online learning platform for learning content aimed at Norwegian primary and secondary education.
This thesis aims to tag educational content from Addito with subject and school year using neural networks based on LSTM, CNN and BERT architectures. The problem is solved as a multi-label classification problem as each learning resource can have multiple tags related to either subject or school year. The results show that the chosen models are able to predict subject labels with sufficient instances in the dataset quite well; however, it fails at predicting the two classes with the least instances in the dataset. None of the models are able to predict school year very well. The best results are obtained using pre-trained BERT models. | |
