Conversations in the Eye of the Storm: At-Scale Features of Conversational Structure in a ...
summary
Constructs @reply conversations from Hurricane Sandy tweets by collecting local keyword streams, pulling each local user's full timeline via REST API, filtering to local replies and originals, and building directed acyclic graphs for ten-day before/during/after/long-after slices. Temporal analysis shows intertweet peaks at ~3 minutes, ~2 hours, ~16 hours; reply times are long-tailed and slower during the emergency. Network analysis shows the during period contains a largest component ≈30%, conversations include 25% self-replies, 73% two-person, 2% multi-person, and group threads nearly double during crises.
AI-written summary · about
chapters (9)
- 0:00Introduction to crisis conversations
- 1:32Moving beyond tweet-level analysis
- 3:42Data collection and conversation extraction
- 5:32Temporal patterns in conversations
- 8:47Network structure of replies
- 11:14Participant composition in discussions
- 13:23Content and conversational practices
- 14:49Summary of findings
- 15:32Q&A session
related talks
01h 02mReal-time Twitter analysis with Reactive Extensions
01h 05mSocial Text, Sentiment, and Tone Analysis
01h 02mLightning Talks
31mGetting Insights through Conversational Chatbots for Analytics
19mStreams of Content, Limited Attention:
10mSnowy: Recommending Utterances for Conversational Visual Analysis
39mUsing a modern data stack to explore & visualize the impact of a global pandemic on airline traffic
05mCrisis Mapping