Computational Text Analysis
Frederik Elwert · 2021
The increasing availability of large digital text collections, both historical and contemporary, poses challenges for traditional text analysis methods. At the same time, methodological advances in the area of computational text analysis make it possible to study large bodies of texts. Instead of restricting the analysis to a limited sample of the collection, computational methods make it possible to identify structural patterns across the complete corpus. Central for these methods is the concept of collocation, i.e., the observation that some words regularly co-occur and that studying these phenomena enables insights about contextual meaning. The chapter presents an overview of different approaches that work on the local level of individual collocations, the intermediate level of collocation networks and the global level of corpus-wide patterns. Emphasis is put on the example of topic modeling as an advanced method of computational text analysis that identifies recurring themes across a large text corpus and allows studying their development. The method&s;s theoretical basis, as well as practical requirements and procedures, are discussed.