Using Artificial Intelligence to Analyze Nature of Science Themes: Einstein and Eddington Documentary Film Example

Abstract

Understanding the nature of science is an essential component of scientific literacy. In a technology and media-oriented environment, text-processing algorithms and various artificial learning approaches are crucial and continue to develop. Latent Dirichlet Allocation is a topic modeling algorithm that has been used frequently for many years to extract the main themes in many documents. This study examined the nature of science themes in the documentary film "Einstein and Eddington," which lasted 1 hour and 28 minutes, with the Latent Dirichlet Allocation topic extraction algorithm. First, the texts in the documentary film were fragmented into 30-second, 1-minute, and 3-minute periods to obtain three different datasets. Considering the literature on these datasets, experiments were carried out using the Latent Dirichlet Allocation algorithm to extract 5, 7, and 9 topics. The Latent Dirichlet Allocation algorithm developed with R programming language was used to analyze the data. In the analysis made by the computer, it is seen that science-related words such as science, scientist, theory, the universe, Eddington, think, speed, and answer stand out. While it was observed that it was difficult to distinguish the detected topics from each other on limited data, it was concluded that the dataset created with 30-second periods made more sensitive topic inferences. Despite the challenges posed by subjectivity regarding the nature of science, it is thought that computer-aided models can provide much more than information retrieval and advanced search. In this context, although it seems pretty difficult to extract the nature of science themes on limited data with the Latent Dirichlet Allocation algorithm, it is possible that the artificial learning models to be used in addition to the Latent Dirichlet Allocation algorithm can detect the nature of science themes.