Location

Level 0, Open Space, Owen G. Glenn Building

Start Date

12-15-2014

Description

In Twitter information primarily propagates through retweet mechanism. While a massive amount of tweets gets generated everyday, only a handful of them get retweeted widely. In this study, we have investigated the impact of user-roles in retweet phenomena. We have introduced the concept of “Information Diffusion Impact” (IDI) and identified three important user roles, namely “information starter”, “amplifier”, and “transmitter”. Retweetability has been modeled using IDI impact for different user roles along with the content features like presence of hashtag, URL etc. Further, the effect of a major event on the factors affecting retweetability has been investigated. Our findings demonstrate that retweetability is significantly affected by amplifiers and information-starters and these effects change substantially due to event. We have also reexamined our model in another dataset of the Boston marathon bomb blast, 2013 and the outcome of this analysis is in good agreement with our findings from Japan earthquake dataset.

Share

COinS
 
Dec 15th, 12:00 AM

Factors Affecting Retweetability: An Event-Centric Analysis on Twitter

Level 0, Open Space, Owen G. Glenn Building

In Twitter information primarily propagates through retweet mechanism. While a massive amount of tweets gets generated everyday, only a handful of them get retweeted widely. In this study, we have investigated the impact of user-roles in retweet phenomena. We have introduced the concept of “Information Diffusion Impact” (IDI) and identified three important user roles, namely “information starter”, “amplifier”, and “transmitter”. Retweetability has been modeled using IDI impact for different user roles along with the content features like presence of hashtag, URL etc. Further, the effect of a major event on the factors affecting retweetability has been investigated. Our findings demonstrate that retweetability is significantly affected by amplifiers and information-starters and these effects change substantially due to event. We have also reexamined our model in another dataset of the Boston marathon bomb blast, 2013 and the outcome of this analysis is in good agreement with our findings from Japan earthquake dataset.