This talk aims to envision how the use of machine learning, High-Performance Computing (HPC) and data-driven approaches can become an everyday practice for historians in the not-too-distant future and change the way intellectual history is practiced. Everyone is talking about large language models such as BERT and how transformers will change the way we approach big data, but what does this mean in practice for historical research? How should historians conceptualize the use of structured and unstructured data and the interplay between them to study such phenomena as the Scottish Enlightenment?
Drawing from the lessons learned from a decade of collaborative work at an interdisciplinary Helsinki Computational History Group (COMHIS), the talk will discuss our group’s research strategy and some of the recent studies of the Enlightenment. Focus will be placed on our Academy of Finland funded, ongoing consortium on the use of HPC in historical research, which is a collaboration between COMHIS and 4 different computer and data science groups. Use of AI and HPC in history is scarce due to the complexities in the data. The aim of the consortium is to use HPC to detect discourses from books, pamphlets and newspapers, and study the interconnections and evolution of the detected patterns. Previously, COMHIS has worked to harmonize and integrate metadata and full text sources, including the English Short Title Catalogue (ESTC) and Eighteenth Century Collections Online (ECCO). The talk will introduce the concept of bibliographic data science and use it to examine the representativeness and biases in ECCO. Of the many down-stream use cases, the talk will discuss eighteenth-century reception studies and networks of publishing in the Scottish Enlightenment. By looking at the possibilities of use of text reuse detection, the talk will demonstrate how this simple, yet effective way of using BLAST for text reuse detection developed in our consortium can help to explore the role that book publishers played in the development of the Scottish Enlightenment.
The talk will also demonstrate the possibilities of using unstructured ECCO data and embeddings created with a transformer-based model for clustering eighteenth-century subject topics that feeds back to the ESTC metadata creating a virtuous circle in the research use of the available data. Our aim is to make novel advancements in the detection of sequential large-scale genres for eighteenth-century documents. Readily available genre information is often sporadic, but the opportunities to use it can open a whole new window to the development of public discourse. With better structured data, we will be able to study the systematization of genres/subject topics in a new manner which is crucial for historical discourse detection and for understanding phenomenon such as the Scottish Enlightenment.
Future vision of COMHIS is to use the latest advancements in deep learning and natural language processing to computationally study such central features for intellectual history as intention, reception, and multiple meanings of an idea, which previously was impossible. We will develop methods by which historical text passages of similar meaning will be recognized, irrespective of their exact wording or language, and despite the errors inherent in digitized historical texts, which has never been done before. Novel computational models of meaning allow us to focus on the development and spread of ideas in their context rather than concepts and keywords, as is typical in current work. At the same time, we will also focus on enhancing the study of books as material objects.
Diesen Blogbeitrag zitieren
digitalhistory (2022, 28. November). Mikko Tolonen: Helsinki Computational History Group, High Performance Computing, and the Study of the Scottish Enlightenment. Digital History Berlin. Abgerufen am 4. März 2024, von https://doi.org/10.58079/nl45