LingLang Lunch (10/22/2014): Masako Fidler (Brown University)

Mining reader receptions of text with keyword analysis

“Keyness” is a property attributed to words extracted from statistical tests (e.g., chi-square and log-likelihood tests), which contrast word frequencies in the target text (Ttxt) against the background of the word frequencies in a larger corpus (the reference corpus, RefC) (Scott 1996, Baker and Ellece 2011). Words with keyness (keywords, KWs) are said to point to what the text is about (“aboutness”), and/or the structural characteristics of the text (Bondi 2010), although what exactly constitutes “aboutness” is still under debate. It is also noted in existing literature that KWs differ when different reference corpora are used as the background.

This presentation will show one application of such keyword analysis (KWA). It attempts to demonstrate that KWA can be sensitive to political shifts in a society/region to varying degrees when RefCs from two distinct historical periods are used to extract data. KWA, then, can point not only to genre-specific properties of a text, but also to what readers, whose usage patterns are reflected in the reference corpus, consider prominent or surprising in a text. KWA can help us motivate different reader receptions of a text.