Modeling social systems processes found in *text corpora through windowed latent semantic analysis and simulation of concept refreshment events

Corpora of texts were examined by means of Latent Semantic Analysis, and interval distributions of refreshment patterns of statistically coherent texts were extracted. These interval distributions were then used to model the textual refreshment in a search for the underlying dynamical behavior that generated the observed reinforcement patterns. The parameters of this search were calibrated with respect to two Latent Semantic Spaces derived from time windowed corpora.;These corpora were selected based on the observation of a temporal shift in the statistics of discourse surrounding the term "CLINTON" that indicated a shift in the meanings that defined the former president in contemporary news media. This work not only contributes a new dynamical approach to text processing and a calibration methodology to agent models that wish to use text as input, but also contributes a statistical technique by which shifts in media framing may be detected.