Code for making graph

datacharter_responders.R

# tweets are assumed to be gathered for this

library(stringr)
library(dplyr)
library(tidyr)
library(ggplot2)
library(ggthemes)
library(ggrepel)
library(Rtsne)
disc <- read.csv("discussion.csv", stringsAsFactors = FALSE)

who <- disc %>% select(status_id, all) %>% 
  unnest_tokens(word,all, token="tweets") %>%
  filter(str_detect(word, fixed("@"))) %>%
  group_by(status_id) %>%
  mutate(responded = gsub("@","", word),
    response = lead(responded)) %>% ungroup() %>%
  filter(!is.na(response), responded != response) %>%
  count(responded, response)

whowide <- who %>% spread(response,n, fill=0)

tsne <- Rtsne(whowide[,2:ncol(whowide)], dims = 2, perplexity=33, verbose=FALSE, theta=0.0, max_iter = 500, check_duplicates=FALSE)
tsne2D <- as.data.frame(tsne$Y)
tsne2D$who = whowide$responded

arcs = who %>% left_join(tsne2D %>% rename(resdx =V1, resdy=V2), by=c("responded"="who")) %>%
  left_join(tsne2D %>% rename(resrx =V1, resry=V2), by=c("response"="who")) 


ggplot(tsne2D, aes(x=V1,y=V2, label=who)) + 
  theme_void() +
  ggtitle("2 dimesional (TNSE) #DataCharter discussion map, who was responding to who") + 
  geom_curve(data=arcs, aes(x=resdx,y=resdy, xend=resrx, yend=resry, label=NULL, alpha=n)) +
  geom_point(size=1.2) +
  geom_text_repel( size=2.5)

write.csv(disc %>% 
            mutate(status_id = as.character(status_id)) %>%
            select(status_id),"ids.csv", row.names = FALSE)