Network Breizh Data Day

networks_bdd.R

library(tidyverse)
library(ggraph)
library(igraph)
library(tidytext)
library(proustr)

bdd <- read.csv("breizhdataday.csv", stringsAsFactors = FALSE)

sw <- data.frame(word  = c(as.character(proust_stopwords()$word),
                           "https","rt", "d'une", "t.co"))
# Bigrams
tm_bdd <- bdd %>%
  unnest_tokens(bigram, text, token = "ngrams", n = 2) %>%
  separate(bigram, c("from", "to"), sep = " ") %>%
  filter(!from %in% sw$word) %>%
  filter(!to %in% sw$word) %>%
  count(from, to, sort = TRUE) %>%
  rename(size = n) %>%
  filter(size >= 3)

bdd_graph <- graph_from_data_frame(tm_bdd, directed = F)
bdd_clust <- cluster_edge_betweenness(bdd_graph)

#Draw
ggraph(bdd_graph) + 
  geom_edge_link()+
  geom_node_label(aes(label = name, 
                      color = as.factor(bdd_clust$membership)), 
                  show.legend = FALSE) +
  scale_color_viridis_d(option = "A") + 
  labs(title = "Bigrammes dans #BreizhDataDay",
       subtitle = "data depuis Twitter",
       caption = "@_colinfay") + 
  theme_graph()


parse_mention <- function(user, vec){
  a <- jsonlite::fromJSON(vec, simplifyDataFrame = TRUE)
  res <- a$user_mentions
  res$user <- user
  select(res, from = user, to = screen_name)
}
safe_parse <- safely(parse_mention)
edges_users <- map2(bdd$from_user, bdd$entities_str,
                    ~ safe_parse(user = .x, vec = .y) ) %>%
  map("result") %>%
  compact() %>%
  reduce(bind_rows) 

bdd_graph_user <- graph_from_data_frame(edges_users)
bdd_clust_user <- cluster_edge_betweenness(bdd_graph_user)
ggraph(bdd_graph_user) + 
  geom_edge_link(arrow = grid::arrow(angle = 10, unit(0.1, "inches")))+
  geom_node_label(aes(label = name, 
                      color = as.factor(bdd_clust_user$membership)), 
                  show.legend = FALSE) +
  scale_color_viridis_d(option = "D") +
  labs(title = "Mentions dans #BreizhDataDay",
       subtitle = "data depuis Twitter",
       caption = "@_colinfay") + 
  theme_graph() 

parse_hashtag <- function(user, vec){
  a <- jsonlite::fromJSON(vec, simplifyDataFrame = TRUE)
  res <- a$hashtags
  res$hash <- "BreizhDataDay"
  filter(res, text != "BreizhDataDay", text != "breizhdataday") %>%
    select(from = hash, to = text)
}
safe_parse <- safely(parse_hashtag)
edges_hash <- map2(bdd$from_user, bdd$entities_str,
                    ~ safe_parse(user = .x, vec = .y) ) %>%
  map("result") %>%
  compact() %>%
  reduce(bind_rows) 

bdd_graph_hash <- graph_from_data_frame(edges_hash)
bdd_clust_hash <- cluster_edge_betweenness(bdd_graph_hash)
ggraph(bdd_graph_hash) + 
  geom_edge_link()+
  geom_node_label(aes(label = name, 
                      color = as.factor(bdd_clust_hash$membership)), 
                  show.legend = FALSE) +
  scale_color_viridis_d(option = "B") +
  labs(title = "Hashtags dans #BreizhDataDay",
       subtitle = "data depuis Twitter",
       caption = "@_colinfay") + 
  theme_graph()