Mining the Cultural Memory of Irish Industrial Schools Using Word Embedding and Text Classification

The Industrial Memories project aims for new distant (i.e., text analytic) and close readings (i.e., witnessing) of the 2009 Ryan Report, the report of the Irish Government’s investigation into abuse at Irish Industrial Schools. The project has digitised the Report and used techniques such as word embedding and automated text classification using machine learning to re-present the Report’s key findings in novel ways that better convey its contents. The Ryan Report exposes the horrific details of systematic abuse of children in Irish industrial schools between 1920 and 1990. It contains 2,600 pages with over 500,000 words detailing evidence from the 9- year-long investigation. However, the Report’s narrative form and its sheer length effectively make many of it findings quite opaque. The Industrial Memories project uses text analytics to examine the language of the Report, to identify recurring patterns and extract key findings. The project represents the Report via an exploratory web-based interface that supports further analysis of the text. The methodology outlined is scalable and suggests new approaches to such voluminous state documents.