Code to create your collaboration network in R

collab_network.R

## Visualizing authorship networks: code for this blog post:
## https://mathewkiang.com/2018/06/17/my-collaboration-network/
library(scholar) # devtools::install_github("jkeirstead/scholar")
library(visNetwork)
library(tidyverse)

## Constants
MIN_TIME <- 60 * 5
MAX_TIME <- 60 * 30
PROJ_WEIGHT <- .8
NODE_WEIGHT <- 1
my_scholar_id <- "eD9_J3wAAAAJ"

## If we've already pulled from Google Scholar, don't do it again. The sleep
## timer makes this an unreasonably long process.
if (!file.exists("./scholar_pulls.rda")) {
    ## Pull publication history of myself
    mvk_df <- get_publications(my_scholar_id)
    
    ## Drop the Opioid Trends preprint so we're not double-counting
    mvk_df <- mvk_df %>% 
        filter(pubid != "-yGd096yOn8C") %>% 
        mutate_at(vars(one_of("title", "author", "journal", 
                              "number", "cid", "pubid")), 
                  as.character)
    
    ## Fix UTF-8 encoding -- once here and once later.
    mvk_df <- mvk_df %>% 
        mutate(author = gsub("é|\xe9", "e", author))
    
    ## The summary df does not have a full author list so we need that. Set a
    ## generous sleep timer to stay below rate limit. It seems like one 
    ## pull every 15 to 30 minutes is fairly safe.
    authors <- NULL
    pubid   <- NULL
    for (p in mvk_df$pubid) {
        print(p)
        authors <- c(authors, getCompleteAuthors(my_scholar_id, p))
        pubid   <- c(pubid, p)
        Sys.sleep(runif(1, min = MIN_TIME, max = MAX_TIME))
    }
    
    ## Save it so we don't have to pull again.
    save(authors, pubid, mvk_df, file = "./scholar_pulls.rda")
}

## Now munging data
load("./scholar_pulls.rda")

## Split out every author into their own column 
full_authors <- tibble(pubid, f_authors = authors)
full_authors <- full_authors %>% 
    ## Fix encoding
    mutate(f_authors = gsub("é|\xe9", "e", f_authors)) %>% 
    ## Fix Rob's middle initial
    mutate(f_authors = gsub("W Moeller", "M Moeller", f_authors)) %>% 
    ## Fix Jen's middle initial
    mutate(f_authors = gsub("Jennifer Hayes", "Jennifer E Hayes", 
                            f_authors)) %>% 
    separate(f_authors, into = sprintf("author_%02i", 1:30), 
             remove = FALSE, sep = ", ", fill = "right")

## Remove all the completely empty columns
full_authors <- full_authors[, colSums(!is.na(full_authors)) > 0]

## Merge back to original dataframe
mvk_df <- mvk_df %>% 
    left_join(full_authors)

## Get a list of unique co-authors for the node list
distinct_authors <- unique(
    unlist(strsplit(full_authors$f_authors, ", ", fixed = TRUE))
    )
distinct_authors <- distinct_authors[distinct_authors != "Mathew V Kiang"] 

## Now let's make a node list
nodes <- bind_rows(
    mvk_df %>% 
        select(title) %>% 
        mutate(value = PROJ_WEIGHT, 
               group = "paper", 
               id = title), 
    tibble(
        title = distinct_authors, 
        value = NODE_WEIGHT, 
        group = "coauthor", 
        id = title
        )
    )

## Add some github side projects
nodes <- nodes %>% 
    add_case(title = "narcan", 
             value = PROJ_WEIGHT, 
             group = "software", 
             id = title) %>% 
    add_case(title = "metabeiwe", 
             value = PROJ_WEIGHT, 
             group  = "software", 
             id = title) %>% 
    add_case(title = "beiwe_data_sample", 
             value = PROJ_WEIGHT, 
             group  = "data", 
             id = title)

## Now the edge list -- reshape our merged dataframe into an edgelist 
## going from paper to co-author
edges <- mvk_df %>% 
    select(title, starts_with("author_")) %>% 
    gather(key = "author", value = "from", starts_with("author_")) %>% 
    filter(!is.na(from), from != "Mathew V Kiang") %>% 
    select(to = title, from)

## Add cases for the non-papers
edges <- edges %>% 
    add_case(to = "narcan", from = "Monica J Alexander") %>% 
    add_case(to = "beiwe_data_sample", from = "Jukka-Pekka Onnela") %>% 
    add_case(to = "beiwe_data_sample", from = "Jeanette Lorme")

visNetwork(nodes, edges) %>% 
    visLayout(randomSeed = 123456) %>% 
    visInteraction(zoomView = FALSE, dragView = FALSE) %>% 
    visGroups(groupname = "coauthor", color = "#3288bd") %>% 
    visGroups(groupname = "paper", color = "#d53e4f") %>% 
    visGroups(groupname = "data", color = "#abdda4") %>%
    visGroups(groupname = "software", color = "#fdae61") %>% 
    visNodes(scaling = list(min = 15, max = 25)) %>% 
    visEdges(width = 3, color = "grey") %>% 
    visSave("./author_network.html")