Using Computational Methods to Explore Bias in Digitized Historical Collections

We introduce the “Environmental Scan” as a method for understanding biases in large digitized collections. This approach focuses on contextualizing big, historical data with descriptive metadata drawn from contemporaneous resources. As a concrete example, we demonstrate how enriching newspaper collections with information extracted from historical press directories enables us to investigate more critically, the overrepresentation and gaps in large, but nonetheless opaque data pools such as the nineteenth-century digitized British press. Firstly, we are especially interested in finding and capturing minority voices in large digitized collections, such as the radical press. To what extent these periodicals find their way to a digitized collection and how can we use enhanced metadata in combination with computational methods to better understand the social perspectives promoted by these newspapers. Secondly, we look more closely at the large but still understudied category of "independent" or "neutral" newspapers.