and one with the start times all aligned. These are both lacking ID data, but interesting to see variety.
x=time, y=tweetcount up to 100.
Conversation
What about some cross between a histogram and heatmap to show changes in tempo of tweeting throughout the thread? Or the velocity/acceleration of tweets over time? I think themes for each thread would be interesting too but not just a word cloud.
1
1
Something like TF-IDF would be a good starting place for each thread. Finding similarities and differences from threads (and the seed topic) could be interesting. I would be interested to see which threads are more similar/different across topics.
1
Each thread? Wouldn’t that be a natural fit for the whole palooza to detect top *shared* themes?
This Tweet was deleted by the Tweet author. Learn more
Yeah 100x100=10k tweets = ~100k words... I’d guess min is a million words
1
The people doing the threads are probably more deliberate about their language though. Even presence and frequency would probably be valuable. Probably worth a quick experiment. I might have time to put this through some basic NLP tools in the next few days.
1
1
Go for it,should be fun and i suspect the thread authors would get a kick out of seeing themselves in a tf-idf mirror


