Developing a digital archaeology classification system using Natural Language Processing and Machine Learning techniques

The Authors propose a knowledge map to analyse and access scientific contents related to Digital Archeology by leveraging various Machine Learning (ML) techniques. The case study concerns the articles published in our international journal «Archeologia e Calcolatori» in the decade from 2011 to 2020 and, as a benchmark, the publications in the 'Computer Applications and Quantitative Methods in Archaeology' (CAA) conference proceedings and journal. The titles and abstracts of the publications featured in these two data sets were analysed using a supervised classification approach into the subfields of computer science, based on the ACM's taxonomy, and by applying topic modelling techniques to discover emergent topics, Named Entity Recognition to identify specific archaeologically relevant entities, and geotagging techniques to link articles with the geographical locations they discuss. The results achieved, although preliminary, provide some methodological suggestions: i) the opportunity to build custom analyses by taking advantage of the increasing availability of open data and metadata; ii) the scope of the contribution of archaeology, and in particular of computational archaeology, to the Heritage Science interdisciplinary domain; the heuristic and predictive role of different ML techniques to gain a multi-faceted access to data analysis and interpretation.