Revisiting Weimar Film Reviewers’ Sentiments Integrating Lexicon-Based Sentiment Analysis with Large Language Models

Film reviews are an obvious area for the application of sentiment analysis, but while this is common in the field of computer science, it has been mostly absent in film studies. Film scholars have quite rightly been skeptical of such techniques due to their inability to grasp nuanced critical texts. Recent technological developments have, however, given us cause to re-evaluate the usefulness of automated sentiment analysis for historical film reviews. The release of ever more sophisticated Large Language Models (LLMs) has shown that their capacity to handle nuanced language could overcome some of the shortcomings of lexicon-based sentiment analysis. Applying it to historical film reviews seemed logical and promising to us. Some of our early optimism was misplaced: while LLMs, and in particular ChatGPT, proved indeed to be much more adept at dealing with nuanced language, they are also difficult to control and implement in a consistent and reproducible way – two things that lexicon-based sentiment analysis excels at. Given these contrasting sets of strengths and weaknesses, we propose an innovative solution which combines the two, and has more accurate results. In a two-step process, we first harness ChatGPT’s more nuanced grasp of language to undertake a verbose sentiment analysis, in which the model is prompted to explain its judgment of the film reviews at length. We then apply a lexicon-based sentiment analysis (with Python’s NLTK library and its VADER lexicon) to the result of ChatGPT’s analysis, thus achieving systematic results. When applied to a corpus of 80 reviews of three canonical Weimar films (Das Cabinet des Dr. Caligari, Metropolis and Nosferatu), this approach successfully recognized the sentiments of 88.75% of reviews, a considerable improvement when compared to the accuracy rate of the direct application of VADER to the reviews (66.25%). These results are particularly impressive given that this corpus is especially challenging for automated sentiment analysis, with a prevalence of macabre themes, which can easily trigger falsely negative results, and a high number of mixed reviews. We believe this hybrid approach could prove useful for application in large corpora, for which close reading of all reviews would be humanly impossible.