.40, 0.45) [0.45, 0.50) [0.50, 0.55) [0.55, 0.60) [0.60, 0.65) [0.65, 0.70) [0.70, 0.75) [0.75, 0.80) [0.80, 0.85) [0.85, 0.90) [0.90, 0.95) [0.95, 1.00) 150rsos.royalsocietypublishing.org R. Soc. open sci. 3:…………………………………………top 500 broadcastersall usersFigure 4. Distribution of positive sentiment fraction for the top 500 broadcasters (for = 0.75), and for all users, using (SS).0.50 0.45 sentiment fraction based on (SS) 0.40 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0 1 10 000 20 000 30 000 40 000 50 000 60 000 70 000 80 000 90 000 100 000 110 000 120 000 130 000 140broadcast score rank positive sentiment fraction negative sentiment fractionFigure 5. The relationship between sentiment fractions (as a moving average over a window of 1000 points) and broadcast score rank, for = 0.75 based on (SS).(SS) and = 0.15, the `negative sentiment strength’ and `negative sentiment fraction’ attributes for the top 5000 broadcasters were very nearly equal to the mean over all users. In addition to investigating the sentiment use of the top broadcasters, we looked for general trends relating sentiment use to broadcast rank. Figure 5 plots moving averages of the (SS) sentiment fraction attributes against broadcast rank, using a window of 1000 observations to smooth the noisy data. We0.8 0.7 sentiment strength based on (SS) 0.6 0.5 0.4 0.3 0.2 0.1rsos.royalsocietypublishing.org R. Soc. open sci. 3:…………………………………………Figure 6. The relationship between sentiment strengths (as a moving average over a window of 1000 points) and broadcast score rank, for = 0.75 based on (SS).see that from rank 1 to about rank 9000 the positive sentiment fraction decreases sharply; after this it decreases slowly in an approximately linear way. The fraction of tweets with negative sentiment appears approximately constant at this scale. Figure 6 plots similar moving averages for the sentiment strength attributes. The average strength of positive sentiment declines sharply to begin with and then slowly, whereas the average strength of negative sentiment is approximately constant. Although the local fluctuations were different, the graphs had the same general shape for all the values of 0.3, 0.45, 0.6, 0.75, 0.9 tested.4. Sentiment and evolution of communities on TwitterIn this section, we describe how we identified meaningful communities or `sub-networks’ of Twitter users, and we Lonafarnib site present the results of our analysis of how these communities evolved over time, including how their sentiment evolved. The existence of communities has been observed in all kinds of real-world networks and identifying them has been the subject of considerable research effort in recent years, much of which can be NS-018 custom synthesis traced back to a seminal paper of Girvan Newman [14]. In the vast literature on community detection (e.g. [15]), a community is often taken to be a group of users with two characteristics: (i) The community is densely connected internally, i.e. people within the same community talk to each other a lot. (ii) There are relatively few links crossing from the community to the outside world, i.e. people talk to fellow members of their community more often than they talk to non-members.4.1. How we detected communities and selected a subset for further studyBecause we wanted to find communities that would endure over time, we needed to take a longer period of data than the 7 days we analysed in ?. We can imagine online discussions that spri..40, 0.45) [0.45, 0.50) [0.50, 0.55) [0.55, 0.60) [0.60, 0.65) [0.65, 0.70) [0.70, 0.75) [0.75, 0.80) [0.80, 0.85) [0.85, 0.90) [0.90, 0.95) [0.95, 1.00) 150rsos.royalsocietypublishing.org R. Soc. open sci. 3:…………………………………………top 500 broadcastersall usersFigure 4. Distribution of positive sentiment fraction for the top 500 broadcasters (for = 0.75), and for all users, using (SS).0.50 0.45 sentiment fraction based on (SS) 0.40 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0 1 10 000 20 000 30 000 40 000 50 000 60 000 70 000 80 000 90 000 100 000 110 000 120 000 130 000 140broadcast score rank positive sentiment fraction negative sentiment fractionFigure 5. The relationship between sentiment fractions (as a moving average over a window of 1000 points) and broadcast score rank, for = 0.75 based on (SS).(SS) and = 0.15, the `negative sentiment strength’ and `negative sentiment fraction’ attributes for the top 5000 broadcasters were very nearly equal to the mean over all users. In addition to investigating the sentiment use of the top broadcasters, we looked for general trends relating sentiment use to broadcast rank. Figure 5 plots moving averages of the (SS) sentiment fraction attributes against broadcast rank, using a window of 1000 observations to smooth the noisy data. We0.8 0.7 sentiment strength based on (SS) 0.6 0.5 0.4 0.3 0.2 0.1rsos.royalsocietypublishing.org R. Soc. open sci. 3:…………………………………………Figure 6. The relationship between sentiment strengths (as a moving average over a window of 1000 points) and broadcast score rank, for = 0.75 based on (SS).see that from rank 1 to about rank 9000 the positive sentiment fraction decreases sharply; after this it decreases slowly in an approximately linear way. The fraction of tweets with negative sentiment appears approximately constant at this scale. Figure 6 plots similar moving averages for the sentiment strength attributes. The average strength of positive sentiment declines sharply to begin with and then slowly, whereas the average strength of negative sentiment is approximately constant. Although the local fluctuations were different, the graphs had the same general shape for all the values of 0.3, 0.45, 0.6, 0.75, 0.9 tested.4. Sentiment and evolution of communities on TwitterIn this section, we describe how we identified meaningful communities or `sub-networks’ of Twitter users, and we present the results of our analysis of how these communities evolved over time, including how their sentiment evolved. The existence of communities has been observed in all kinds of real-world networks and identifying them has been the subject of considerable research effort in recent years, much of which can be traced back to a seminal paper of Girvan Newman [14]. In the vast literature on community detection (e.g. [15]), a community is often taken to be a group of users with two characteristics: (i) The community is densely connected internally, i.e. people within the same community talk to each other a lot. (ii) There are relatively few links crossing from the community to the outside world, i.e. people talk to fellow members of their community more often than they talk to non-members.4.1. How we detected communities and selected a subset for further studyBecause we wanted to find communities that would endure over time, we needed to take a longer period of data than the 7 days we analysed in ?. We can imagine online discussions that spri.