Abigail Swenor, University of Notre Dame, Neil Coffee, University at Buffalo, Walter Scheirer, University of Notre Dame
The use of artificial intelligence and machine learning has become increasingly prevalent across a multitude of disciplines, including humanities-based research. The study of semantics in computational linguistics and language has been reorganized by the introduction of tools such as neural networks and computational representations of language. In this paper, we explore the use of AI to study semantics specifically in classical literature and how we see this pedagogical relationship developing as AI continues to evolve.
We highlight the influence of semantic matching capabilities of AI models in various applications (Salloum et al. 2020) and how they can be applied to specific classical texts to discover different types of patterns, such as quantitative intertextuality (Scheirer et al. 2016) and enriched topic models for semantic analysis (Osmani et al. 2020). This combination of methods from AI and research questions from Classics allows researchers to produce scholarship more efficiently, while allowing the technology to perform distant readings with reliable results; these in turn can then be used to support further research. We also recognize the limitations of current AI technology when it comes to close readings and human-interpretable results for humanities scholars wishing to use these technologies to increase their research efficiency and find new patterns in literature, especially in areas such as topic modeling (Baumer et al. 2020; Pääkkönen and Ylikoski 2021). We further explore the need for humanistic interpretation of texts, as large language models are often trained on massive amounts of data with completely different text domains than Classics. Lastly, we also examine “human-in-the-loop” AI approaches (Mosqueira-Rey et al. 2022; Wu et al. 2022) and evolving semantic matching algorithms to provide a more humanistic analysis of classical literature.