Applying Computational Analysis to Textual Data from the Wild: a Feminist Perspective
summary
Unsupervised topic modeling combined with hand coding and collaborative sense‑making maps 1,647 FeedFinder reviews onto socioeconomic data (using the Index of Multiple Deprivation), showing privacy appears across five distinct topic clusters and that 23% in most deprived quintile vs 16% in least deprived locations drive different concerns: deprived-area reviews emphasize designated feeding spaces, locks and signposting, affluent-area reviews emphasize comfort and atmosphere. The pipeline uses 30 topics, is statistically validated against manual codes, and surfaces actionable, nonreductive distinctions.
AI-written summary · about
chapters (10)
- 0:00Introduction and overview
- 1:11Hybrid approach for qualitative data
- 3:19Feed Finder app and data
- 5:06Feminist GIS perspectives
- 8:21Combining datasets for analysis
- 9:03Topic modeling process
- 13:04Socioeconomic analysis results
- 15:30Key findings and implications
- 16:55Conclusion and call for hybrid work
- 17:42Q&A session
related talks
57mDiscover semantic topics in a collection of documents
47mData for good: Lessons from the frontline
45mDecoding Topics: A Comparative Analysis of Python’s Leading Topic Modeling Libraries Using Climate C
35mAnalyzing Public Conversation using LDA and Topic Modeling, in Python
31mPanel Discussion : Ethics in Data Science
32mData Science for Feminist Comebacks
31mWhen humans teach machines
15mHow Big Data Makes Content More Relevant in Search