make_data_coauthors.R

library(dplyr)
library(stringr)
library(scholar)
library(igraph)
library(ggraph)
library(see)
library(tidygraph)


# Functions --------------------
find_coauthors <- function(id) {
  df <- scholar:::list_coauthors(id, Inf)
  df$id <- scholar:::grab_id(df$coauthors_url)
  df[c("author", "coauthors", "id")]
}


list_coauthors <- function(df, sleep = 0, silent = FALSE) {
  data <- data.frame()

  for (i in 1:nrow(df)) {
    if (silent == FALSE) {
      cat(paste0(round(i / nrow(df) * 100, 2), "%\n"))
    }

    Sys.sleep(runif(1, 1, sleep))

    if (!df$coauthors[i] %in% unique(df$author)) {
      data <- rbind(data, find_coauthors(df$id[i]))
    }
  }

  data
}


get_coauthors <- function(id = "bg0BZ-QAAAAJ", n_deep = 1, sleep = 3, silent = FALSE) {
  stopifnot(is.numeric(n_deep), length(n_deep) >= 1, n_deep != 0)

  df <- find_coauthors(id)
  if (n_deep > 0) {
    for (level in 1:n_deep) {
      cat(paste0("level ", level, ":\n"))
      df <- rbind(df, list_coauthors(df, sleep = sleep, silent = silent))
    }
  }

  df$author <- stringr::str_to_title(df$author)
  df$coauthors <- stringr::str_to_title(df$coauthors)
  df <- df[!df$author %in% c("Sort By Citations", "Sort By Year", "Sort By Title"), ]
  df <- df[!df$coauthors %in% c("Sort By Citations", "Sort By Year", "Sort By Title"), ]
  df[c("author", "coauthors")]
}



create_graph <- function(data) {
  data |>
    tidygraph::as_tbl_graph(directed = FALSE) |>
    dplyr::filter(name != "") |>
    dplyr::mutate(
      closeness = tidygraph::centrality_closeness(normalized = TRUE),
      degree = tidygraph::centrality_degree(normalized = TRUE)
    ) |>
    tidygraph::activate(edges) |>
    dplyr::mutate(
      importance = tidygraph::centrality_edge_betweenness(),
      group = as.factor(from)
    ) |>
    tidygraph::activate(nodes) |>
    dplyr::filter(!name %in% c("Sort By Citations", "Sort By Year", "Sort By Title")) |>
    dplyr::mutate(
      name = stringr::str_remove(name, ",.*"),
      # Other groupings: group_edge_betweenness(), group_walktrap(), group_spinglass(), group_louvain()
      group = as.factor(tidygraph::group_edge_betweenness())
    ) |>
    as.list()
}



# Get data --------------------------------

# Scrap data from google scholar
data <- get_coauthors("bg0BZ-QAAAAJ", n_deep = 2, sleep = 15)

# Save data so that it can be re-used
data |>
  filter(!coauthors %in% c("About Scholar", "Search Help")) |>
  filter(!author %in% c("About Scholar", "Search Help")) |>
  write.csv("data/data_network.csv", row.names = FALSE)



# Process data ------------------------------------------------------------
data <- read.csv("data/data_network.csv")

# Prune
# 1. Find direct relations of DM
data1 <- data[(data$author == "Dominique Makowski" | data$coauthors == "Dominique Makowski"), ]
firstlevel <- unique(c(data1$author, data1$coauthors))

# 2. Find direct relations of these first-level co-authors
data2 <- data[(data$author %in% firstlevel | data$coauthors %in% firstlevel), ]
secondlevel <- unique(c(data2$author, data2$coauthors))

# 3. Find whether these have also coauthors in the list to link them between them
data3 <- data[(data$author %in% secondlevel & data$coauthors %in% secondlevel) |
  (data$author %in% secondlevel & data$coauthors %in% secondlevel), ]


# Make plot --------------------------------

# Plot
data_graph <- create_graph(data=data3)

p <- tidygraph::tbl_graph(nodes = data_graph$nodes, edges = data_graph$edges, directed = FALSE) |>
  ggraph::ggraph(layout = "nicely") + # fr, kk, nicely, lgl, graphopt, dh
  ggraph::geom_edge_arc(aes(alpha = importance), show.legend = FALSE, strength = 0.1) +
  ggraph::geom_node_point(aes(size = degree, colour = group), show.legend = FALSE) +
  ggraph::geom_node_text(aes(label = name, size = degree), repel = TRUE, check_overlap = TRUE, show.legend = FALSE, max.overlaps = 20) +
  # ggraph::geom_node_label(aes(label = name, size = degree), repel = TRUE, show.legend = FALSE) +
  ggraph::theme_graph() +
  scale_size_continuous(range = c(2, 6)) +
  scale_edge_alpha_continuous(range = c(0.1, 0.8)) +
  # scale_color_viridis_d() +
  see::scale_color_material_d(palette = "rainbow", reverse = TRUE)

# Show plot
p
# Save
ggsave("img/collaboration_network.png", p, dpi = 500, width = 10, height = 10)