Evaluating computational techniques from image signal processing for finding formulaic language patterns in historical sources

This paper proposes to use methods from image and audio signal processing to evaluate whether the approach behind them can be helpful in the domain of computational historical research. More specifically image compression, noise reduction and image segmentation. At first glance it might not be obvious how these techniques are related to the problem of formulaic language. The image techniques listed have in common that they try to represent the essentials of the image in a more compact form. They keep the structure of the image and diminish the variation. Formulaic language patterns lend structure to a text and function as anchor points in the text. In that sense they fulfill a similar role.