Jacob Reidhead (Sociology)
Lunch will be provided.
Abstract
For first-time practitioners, computational methods for text analysis can sound a little foreign and intimidating. This presentation is intended to introduce graduate students in the humanities and social sciences to two new methods for surveying large bodies of text. These tools are useful for 1) identifying underlying themes, 2) periodicizing the text based on continuity and discontinuity in the structure of discourse, and 3) detecting the structure of multiple sub-discourses over time. The following descriptions of the methods may seem too technical for first-timers, but they will be explained with much greater simplicity and with clear examples in the presentation.
Multi-layer networks have most often been used to model social and physical structure over time. However, when applied to texts or other semantic fields, they can be a powerful tool for surveying the general topology of a corpus. For example, community detection on these networks can capture threads of meaning and pockets of discourse as they weave through a body of text. Layer reducibility techniques provide a way to objectively periodicize a discourse based on its structure.
Latent topic analysis (LDA) is a class of methods used to reveal underlying topics in a body of text.
This week's presentation demonstrates how multi-layer network and latent topic analysis methods may be combined in order to analyze textual discourse over time. These methods offer a much lower resolution of analyses compared to a deep reading of texts. Thus they are not meant to be a substitute for conventional methods in the humanities and social sciences, but rather a complementary approach to help researchers characterize bodies of text which may be too large to read in their entirety. These methods can also help researchers overcome subjective biases when identifying the themes and structure of a corpus. Lastly, these methods are longitudinal, meaning that they can evaluate the dynamics of a discourse over time, whereas most of the current techniques in text analysis are cross-sectional.
Biography
Jacob Reidhead is a PhD candidate in sociology. He is currently working on four projects involving factional dynamics and/or political discourse among 1) South Korean political parties, 2) North Korean elites, 3) North Korean human rights organizations, and 4) South Korea's media discourse on ethnic minorities. He is interested in developing mixed method research designs for the study and interpretation of informal organization and semantic fields. Current methodological interests include network-informed ethnography, multi-layer network analysis, Markov chains, and game theory.
Resource Links
3) muxviz: Multilayer Analysis and Visualization Platform ---- http://muxviz.net/
4) muxviz Youtube Tutorial ---- https://www.youtube.com/watch?v=8JU_d93VGzE
