A brief application to work with google scholar and show coauthorship network (see video intro here)
You need to have shiny package installed in R, then install these packages as well, using:
install.packages(c("shinyjs", "graphics", "methods", "scholar", "igraph", "network", "stringdist", "networkD3"))
Then open app.R file inside Rstudio, now you should see a “Run App” button on upper left side of script editor (instead of usual “Run” button), click on “Run App” and you should have another window opened on the top of Rstudio editor, follow guides in app (like above) to see google scholar coauthorship network graph.
To run the application you can simply go here and download the code file, or instead copy and paste following code into an R script file and save it as “app.R” and follow above steps to run it with shiny.
##################### Information ########################
# *------------------------------------------------------
# * File-Name: Google Scholar shiny app
# * Date: 05/02/2016
# * Author: 3AT
# * Purpose: build coauthorship network
# * Input File: google scholar profile link
# * Output File: coauthorship graph, main component, name inconsistencies
# * Data Output:
# * Previous file:
# * Machine: Ubuntu
# *------------------------------------------------------
# *------------------------------------------------------
### in order to see coauthorship networks graph, copy google scholar profile of author (example: "https://scholar.google.com/citations?user=zufgVroAAAAJ&hl=en") and put it in the text box provided in the application UI, you will see coauthorship network in seconds
# this application uses scholar package in R to exctract data from google scholar
# as always, I owe most parts of the code and what I wrote to stackoverflow and google searches
# load libraries
require(shiny)
require(shinyjs)
require(graphics)
require(methods)
require(scholar)
require(igraph)
# require(network)
require(stringdist)
require(networkD3)
# Shiny app
shinyApp(
ui = fluidPage(
shinyjs::useShinyjs(),
titlePanel("Who are researchers mostly co-authoring with?"),
sidebarPanel(
helpText("see how it works"),
HTML('<iframe width="350" height="200" src="https://www.youtube.com/embed/NZ5WdBnZ-CE" frameborder="0" allowfullscreen></iframe>'),
helpText("Copy the link to your person of interest Google Scholar profiles (example: 'https://scholar.google.com/citations?user=zufgVroAAAAJ&hl=en') below and see the co-authorships network", style = "color:green"),
textInput("scholarid",'google scholar profile link',value = ""),
actionButton("submit", "Submit"),
helpText("Reference: data for co-authorships and publications are being extracted from google scholar with 'scholar' package for R"),
helpText("**: It is the first and earliest version, some issues are knwon and in the process to be solved, some not, I will be happy to hear: akbaritabar[at] gmail.com")
),
mainPanel(
h4("author's data",align="top",align="center"),
tableOutput("sctable"),
h4("Raw co-authorships Graph",align="top",align="center"),
helpText("use mouse scroll to zoom in and out; further details (including main component) below", style="color:green"),
simpleNetworkOutput("evalPlot1"),
h4("Main component",align="bottom",align="center"),
plotOutput("evalPlot2"),
h4("author name's inconsistencies",align="top",align="center"),
helpText("if you see similar name with different spellings, means google scholar data has inconsistencies that need to be resolved", style="color:green"),
textOutput("sctext")
)
),
server = function(input, output) {
observeEvent(input$submit, {
citid <- strsplit((strsplit(input$scholarid,"&",fixed = TRUE)[[1]][1]),"=",fixed = TRUE)[[1]][2]
infosc <- get_profile(citid)
pub <- get_publications(citid, flush=TRUE)
authnum_articles <- get_num_articles(citid)
authdistinc_journal <- get_num_distinct_journals(citid)
autholdest_article <- get_oldest_article(citid)
authnum_topjournals <- get_num_top_journals(citid)
## tolower does character conversion, and remove the trailing "..."
coauthors <- sub('[ ,.]+$', '', tolower(pub$author))
coauthors <- coauthors[nzchar(coauthors)] # only keep entries that aren't blank
## Add self-loops for single-author entries
adjlist <- strsplit(coauthors, '\\s*,\\s*')
lens <- lengths(adjlist)
adjlist[lens==1L] <- lapply(adjlist[lens==1L], rep, times=2) # repeat single author entries
edgelist <- cbind(
unlist(lapply(adjlist, tail, -1L)), # col1
rep(sapply(adjlist, `[`, 1L), times=lengths(adjlist)-1L) # col2
)
charr.list <- list()
charr.list2 <- list()
for (j in seq_along(edgelist[,1])) {
charr <- edgelist[,1][j]
charr<-unlist(strsplit(charr, '\\s* \\s*'))
charr <- charr[max(seq_along(charr))]
charr.list[[j]] <- charr
charr2 <- edgelist[,2][j]
charr2<-unlist(strsplit(charr2, '\\s* \\s*'))
charr2 <- charr2[max(seq_along(charr2))]
charr.list2[[j]] <- charr2
}
edgelist.clean <- cbind(as.character(charr.list),as.character(charr.list2))
coauthorgraph <- graph_from_edgelist(edgelist.clean,directed = TRUE)
# report of author name inconsistencies
name.author<-unlist(strsplit(infosc$name, '\\s* \\s*'))
name.author <- tolower(name.author[max(seq_along(name.author))])
comparisons <- stringdist(name.author,names(V(coauthorgraph)),method = "jw")
node.comparisons <- V(coauthorgraph)[0<comparisons & comparisons<0.3]
results <- list(node.comparisons=node.comparisons)
# author information to be shown on app
authordatadownloaded <- cbind.data.frame(
"number of articles" = as.character(authnum_articles),
"unique journals" = as.character(authdistinc_journal),
"oldest article" = as.character(autholdest_article),
"number of top journals" = as.character(authnum_topjournals),
"h-index" = as.character(infosc$h_index)
)
networkData <- data.frame(edgelist.clean)
shinyjs::disable("submit")
shinyjs::disable("scholarid")
shinyjs::disable("scholarid2")
output$evalPlot1 <- renderSimpleNetwork({
set.seed(333)
# using networkD3 to have zoom possibility
simpleNetwork(networkData,zoom = TRUE)
})
output$evalPlot2 <- renderPlot({
set.seed(333)
ord <- V(coauthorgraph) # node order
theta <- seq(0, 2*pi-2*pi/length(ord), 2*pi/length(ord)) # angle
theta[theta>pi] <- -(2*pi - theta[theta>pi]) # convert to [0, pi]
dists <- rep(c(1, 0.7), length.out=length(ord)) # alternate distance
## Plot
plot(decompose.graph(coauthorgraph)[[order(clusters(coauthorgraph)$csize,decreasing = TRUE)[1]]], layout=layout.auto, vertex.label.degree=-theta, main = paste("main component of:",infosc$name),
vertex.label.dist=dists, vertex.label.cex=1.1,
vertex.size=5, vertex.color='#FFFFCC', edge.color='#E25822')
})
output$sctable <- renderTable(authordatadownloaded)
output$sctext <- renderPrint(results)
})
}
)If you download this rmarkdown file and add runtime: shiny to the yaml setting before output: html_document: then you can run the shiny application and see the code and how it works in one place (Since this is a static page hosted on github it says “Shiny applications not supported in static R Markdown documents”)