Skip to main content

Computational Text Analysis

10 Week 6: Unsupervised learning (topic models)

This week builds upon past the scaling techniques we explored in Week 5 and instead turns to another form of unsupervised approach—topic modelling.

The substantive articles by Nelson (2020) and Alrababa’h and Blaydes (2020) provide, in turn, illuminating insights using topic models to categorize the thematic content of text information.

The article by Ying, Montgomery, and Stewart (2021) provides a valuable overview and accompaniment to the earlier work of Denny and Spirling (2018) when thinking about how we validate our findings and test the robustness of any inferences we make from these models.

Questions:

  1. What assumptions underlie topic modelling approaches?
  2. Can we develop structural models of text?
  3. Is topic modelling a discovery or measurement strategy?
  4. How do we validate any model?

Required reading:

  • Nelson (2020)
  • PARTHASARATHY, RAO, and PALANISWAMY (2019)
  • Ying, Montgomery, and Stewart (2021)

Further reading:

  • Chang et al. (2009)
  • Alrababa’h and Blaydes (2020)
  • J. Grimmer and King (2011)
  • Denny and Spirling (2018)
  • Smith et al. (2021)
  • Boyd et al. (2018)

Slides: