KEMBAR78
Getting Tweet Content Examples

Getting Tweet Content Examples

Oxford Spring School, 2022

Christopher Barrie

Introduction

  • Why get tweets?

    • Testing API call (again…)

    • We can investigate text beyond keywords

    • We can investigate additional tweet characteristics (metrics, annotations)

    • We can investigate user characteristics…

Getting tweets

  • As barometer of public opinion


Getting tweets

  • As barometer of public opinion
    • Coverage?

Getting tweets

  • As barometer of public opinion
    • Bias?


Getting tweets

  • As barometer of public opinion
    • Advantages versus surveys?

      • Real-time (“always on”)

      • More accurately/realistically measured (?)


Sampling tweets

  • By keyword

Geolocating tweets

  • By user-reported location

Getting tweets

  • As measure of individual attitudes

)

Getting tweets

  • As measure of individual attitudes


Getting tweets

  • As measure of individual attitudes

Sampling tweets

  • (Here) reliance on hacked lists
  • Continued streaming

Sampling tweets

Design considerations:

  • Attrition via takedown
  • Tweet honesty?
  • Alternative sampling strategy?
  • Geolocation…

Geolocating tweets

  • V small pct. of tweets are geolocated (.1–1%)
  • Have to rely on coarse-grained location
  • Or infer from “friends” (whom an individual follows)

    • Blunt force: frequency of locations of friends network

    • More precise: spatial label propagation

      • Get location of friends

        • Get locations of friends of friends…

      • Predict user location from distribution

Coding tweets

  1. Coding criteria
  2. MTurk

    1. Use labels as outcome

    2. Use to train supervised ML algorithm

Getting tweets

  • As a measure of popularity/engagement


Getting tweets

  • As a measure of popularity/engagement


Getting tweets

  • As a measure of popularity/engagement

    • Accurate measure of human attention?

    • Accurate measure of endorsement?

Group work

  1. Replicate Flores (2017) using get_all_tweets() and another event of interest
  2. Formulate code needed to request that data
  3. (If access granted) make the API call to get data
  4. Consider limitations + design considerations