Vous pouvez ajouter des informations de localisation à vos Tweets, comme votre ville ou votre localisation précise, depuis le Web et via des applications tierces. Vous avez toujours la possibilité de supprimer l'historique de localisation de vos Tweets. En savoir plus
#BernieBros are so awful on Twitter? Perhaps when you cherrypick your evidence and use very little of it.
The data on 600K tweets shows little difference in the sentiment among tweets of supporters of different candidates.
@RachelBitecofer @LarrySabato @Tanielpic.twitter.com/JAL82VvWK1
This table has data only from tweeters who follow just one candidate. I'd rather try to scan their content and bios to figure out if people following multiple candidates really plan to vote for a specific one, but that interesting NLP work is for another day.pic.twitter.com/obDMVCKdvS
1. The chance of a negative tweet coming from any particular candidate is about the same If people following both Warren and Sanders are more closely identified with one candidate (big assumption), then maybe this shows evidence that the impression comes from more Bernie tweets.
2. But I'd feel better being able to detect which people following multiple candidates are more obviously making their 1st choice clear to readers. An interesting NLP challenge but the technical challenges mean it won't be done tomorrow. ;-)
Out of curiosity how do you determine a negative tweet?
Based on ranges of values in the two algorithm's outputs. Textblob generates a number from most negative (-1) to most positive (+1). < -0.75 is very negative. -.75 to -.5 is negative.
Interesting! Did you look at (or is it possible to look at) breaking down negative tweets into negative general vs. negative toward candidates they are not supporting for?
I did do one summary table where people were replying to other people who didn't share any candidate follows. It is a much smaller dataset, but again the stats didn't differ significantly depending on which candidate the tweeter followed.
Thanks for this. I agree with others that your study deserves far more attention and discussion. Of course, you'll have to be prepared to defend your methodology to a bunch of people who are invested in pushing a narrative that's counter to your conclusions. Great work!
I'm more interested in collaborating with other data scientists, NLP users to defend and improve the models based on their feedback. So I'll probably let others talk to antagonists. ;-)
thank you so much. I wish this info was made available in an article format to share. maybe on medium dot com? I know its a lot to ask youve done so much work.
Someone from a major publication is interviewing me today, so I would expect to see something soon. Not the same as an academic journal paper, but that wouldn't get published until after the election ;-)
Twitter est peut-être en surcapacité ou rencontre momentanément un incident. Réessayez ou rendez-vous sur la page Twitter Status pour plus d'informations.
