Philly Tweet Dashboard

Overview

The Philly Tweet Dashboard is an ongoing project to provide demographic and topic information of tweets referencing Philadelphia. All tweets used in the analysis contain the term Philly or Philadelphia.

Demographics

The charts above provide hourly tallies of tweets referencing Philadelphia sorted by age group and gender over the last 24 hours. The gender and age of the author is estimated using lexica provided in Sap, et al.(2014) available courtesy of the World-Well Being Project.

Tweet Topics on

The charts above provide an insight into the topics that are most prevalent in tweets referencing Philadelphia on the given date. The chart on the left provides a visualization of the distance between topics. (Chuang, et al.,2012) Essentially this is a measure of how different one topic is from another. The plot on the right shows the overall frequency of the term (red) and the estimated frequency in the given topic (blue). The relevance metric, λ, determines the terms which will be displayed. In general:

  • λ values closer to 1: Shows terms with the highest probability within the topic
    • favors terms used very frequently, often across many topics
  • λ values closer to 0: Shows terms with the highest ratio of a term's probability within a topic to its marginal probability across the corpus. (Seiver, Shirley, 2014)
    • favors terms which are very topic specific, but may occur rarely

Topics are estimated using Latent Direchlet Allocation (Blei, et al.,2003) implemented with Gensim and visualized using pyldavis.