CDH webinar: Analyzing 23.361 Literary Reviews Using Machine Learning
Does the way books are reviewed depend on the nationality of the reviewer? This seminar focuses on the study Joris Veerbeek (Utrecht Data School) carried out as part of his research master’s thesis, where he compared a collection of 23.361 Flemish and Dutch newspapers. Building upon established classification schemes for literary reviews, he fine-tuned and employed three BERT-models in order to automatically classify: 1.) whether a given sentence is evaluative or descriptive, 2.) what ‘aspects’ are discussed (e.g. plot, style, author), and 3.) what ‘characteristics’ are assigned to those aspects (e.g. humoristic, emotional, political). Comparing the aspects and characteristics discussed in Flemish and Dutch reviews revealed that Flemish reviewers are more concerned with the ideological (e.g. the religious, political, moral, and didactic) characteristics of literary works than Dutch reviewers. The seminar will mainly highlight the methodological part of this research, and discuss the challenges and pitfalls of applying recent NLP techniques to humanities data.
Please register before March 18th by sending an email to email@example.com, stating your name, email address and affiliation.
You will receive the MS Teams meeting link through e-mail on the day of the webinar.