150 likes | 269 Views
“MASHABLE” by Suman Kalyan Maity Dept. of CSE IIT Kharagpur CNeRG Retreat 2014 . Outline of the talk. Adoption of # hashtags in Twitter Twitter as an evolving linguistic system. Adoption of # hashtags in Twitter. State-of-the-art.
E N D
“MASHABLE” bySumanKalyanMaity Dept. of CSE IIT Kharagpur CNeRG Retreat 2014
Outline of the talk • Adoption of #hashtags in Twitter • Twitter as an evolving linguistic system
State-of-the-art • No model which shows how #hashtags are propagated and adopted • Some works on popularity of #hashtags (mainly learning some features and predicting the popularity)
Adoption of #hashtags #sachinsachin #MissYouSachin #batkid #SalaamSachin #ThankYouSachin #SRT200 #srtforever #salutethelegend #borntoplaycricket #SachinMakesMeSenti #GoodByeSachin #SachinMadeMeSmile #mysachin #respect #legend
Adoption of #hashtags #ripnelsonmandela #mandela #nelson #nelsonmandela #mandelamemorial #ripmandela #peopleschoice #madiba #respect #rememberingmandela #ripmadiba #inspiration #mandelafuneral #legend
Quantities of interest • Temporal dynamics of no. of unique hashtag associated with an event • Temporal dynamics of total no. of hashtags associated with an event • Which #hashtags become popular? but how?
Modeling the dynamics …… • Data Challenges:- - we need underlying follower-following network - But we have only 1% random sample (disconnected graph with giant component size ~ 0(1)) - we are reconstructing the graph by considering the users in our sample as seed node and following their “follower/following links” …. Still we are not getting sufficiently large giant component Any suggestions will be highly appreciated
Modeling the dynamics …… • A user has a finite memory (O(1)) to keep #hashtags it knows • At each timestep t, - a user is randomly selected to post tweets - with prob. p, the user post a tweet with brand new #hashtag (p– avg. rate of innovation in the system per timestep) otherwise - he/she posts a tweet selecting #hashtags it knows of (preferentially selected based on the #hashtag popularity) • Any tweet posted appears instantly on the screen of the user’s followings • The followers adopt the #hashtagpreferentially according to the popularity of the #hashtag Any Suggestions?
Word level analysis Evolution of various quantities over time (~2.5 yrs) - avg no. of char or word per tweets - level of formalism “I” vs “i” “very” vs “really” “you” vs “u” (want to define some measure to detect degree of formalism/informalism)
More deeper analysis • Word co-occurrence graph - want to analyze the temporal core of the graph • Evolution of slangs • Adoption of linguistic styles (formality vs informality) Any Suggestions?
Thank you Any Questions?