One of my goals when I started the Inscriptions of Israel/Palestine project was to reach a point where we could not only make inscriptions relevant and exciting, but also that we could use them for digital analyses. I am happy to report a first attempt to do this. I co-authored, with Daiki Tagami (who did the lion’s share of the work), a paper on “Machine Learning Techniques for Analyzing Inscriptions from Israel.”
Abstract:
The date of artifacts is an important factor for scholars to get a further understanding of culture and society of the past. However, many artifacts are damaged over time, and we can often only get fragments of information regarding the original artifact. Here, we use the inscription data from Israel as a model dataset and compare the performances of eleven commonly used regression models. We find that the random forest model would be the optimal machine learning model to predict the year of inscriptions from tabular data. We further show how we can make interpretations from the machine learning prediction model through a variance important plot. This research shows an overview of how machine learning techniques could be used to resolve digital humanities problems by using the Inscription of Israel/Palestine dataset as a model dataset.
The full paper can be found here.