Péter Gelányi (E-mail)
Word embeddings offer a quantitative representation of words’ semantic relationships. In my thesis, I explore their potential use in studying media bias and slant. The theoretical background of my work is embedded in both the literature on media bias and word embeddings. I detail my analysis of a newly collected Hungarian online media corpus. I fit multiple word embedding models, compare their performance, and use the best one to explore the semantic relationships of specific keywords across mediums and with elements of a sentiment dictionary. My results highlight both the advantages and drawbacks of word embeddings.