citation-network-sketch.rmd

Citation Network Analysis
========================================================


1. Go to http://apps.webofknowledge.com.offcampus.lib.washington.edu/
2. click the 'web of science' tab at the top
3. Do a search...  refine search by journal... perhaps archaeolog* in 'publication title', hit 'search'
4. scroll to bottom of search results, where 'Output Records' section is
5. Select records -> 1-500 since 500 is max 
6. Select content -> Full Record & Cited References
7. Select destination -> "save to tab-delimited file (UTF-8)"
8. Inspect text file with text editor

Load the data into R...
```{r}
# get files of citations into R
# this is the folder that contains the txt files of citations
wd <- "C:/Users/marwick/Downloads/AAcitations"
setwd(wd)
fl <- list.files()
require(data.table)
arts_list <- lapply(fl, function(i) data.table(read.delim(i, sep="\t", stringsAsFactors=FALSE, check.names=FALSE, quote="", row.names = NULL)))
# rearrange data tables in list into one big data table
arts <- do.call("rbind", arts_list)
```
Rearrange the table to split out the citations into another table but keep a lookup value so the new table of citations can be linked to the article in the original table.

First, format citations to try and make them consistent. 
```{r}
require(data.table)
# format get source article details in the same format as WOS citations
ar <- vector("list", length = nrow(arts))
for(i in 1:nrow(arts)) {
# get source article details in same format as citations 
# first author (remove periods)
au1 <- toupper(gsub("\\.", "", unlist(strsplit(arts[i, `PT`], ";", fixed=TRUE))[[1]]))
# year (with spaces and comma)
yr <- paste0(" ", arts[i, PD], ",")
# abbreviated journal title (with spaces and comma)
bn <- paste0(" ", arts[i, BN], ",")
# VNNN
vn <- ifelse(is.na(arts[i, PY]), "", paste0(" V", arts[i, PY], ","))
# PNNN
pn <-  ifelse(is.na(arts[i, MA]), "", paste0(" P", arts[i, MA]))
# DOI
do <- ifelse(is.na(arts[i, AR]), "", paste0(", DOI ", arts[i, AR]))
# put it all together
ar[[i]] <- paste0(au1, yr, bn, vn, pn, do)
}
# put in a list
ar1 <- unlist(ar)
```
Second, make the table of source article and citations, trying to make the text format consistant on both sides.
```{r}
require(data.table)
# make one data table of all citations for one article
cit_list <- vector("list", length = nrow(arts))
for(i in 1:nrow(arts)){
  # reformat citations to match source
  # and improve consistency
  # get rid of comma between author and year
  c1 <- sub(",", "",  
            unlist(strsplit(arts[i, FX], "; ")), 
            fixed = TRUE)  
  # get rid of periods with author name
temp <- strsplit(c1, ",")
c2 <- toupper(unlist(lapply(seq_along(temp), function(x) {
  t1 <- gsub("\\.", "", temp[[x]][1])
  paste(t1, temp[[x]][2], temp[[x]][-c(1, 2)], sep = ",") })))

  # control flow in case there are no citations(!)
  if(length(c2) == 0) { next } else {
  cit_list[[i]] <- data.table(ar = toupper(unlist(rep(ar[i], length(c2)))),
                   cits = c2                 
                   ) }
}
# put list of dts into one dt
cits <- do.call("rbind", cit_list)
```
Now that we have the data in a table, we can create and plot the network
```{r}
# trim cits so labels are readable
cits1 <- data.table(ar = tolower(substr(cits$ar, 1, 40)), 
                    cits = tolower(substr(cits$cits, 1, 40)))
require(igraph)
g <- graph.edgelist(as.matrix(cits1))
# identify those vertices part of less than n edges
n <- 50
g.vs <- V(g)[degree(g) < n] 
# exclude them from the graph
g1 <- delete.vertices(g, g.vs) 
# remove nodes that don't have a set number of edges
g1 <- delete.vertices(g1, which(degree(g1) < n/5))
# make vertex sizes variable
V(g1)$size <- degree(g1)/10
# make label sizes variable
V(g1)$label.cex <- degree(g1)/10
# basic plot
windowsFonts(Arial=windowsFont("Arial"))
plot(g1, 
     edge.curved = TRUE,   
     vertex.color= "grey", 
     edge.arrow.size = 0.01, 
     vertex.label.dist = 0.5,  
     vertex.label.color = "grey30", 
     vertex.label.family = "Arial" 
     )#vertex.label = NA)
# use a layout
layout1 <-  layout.fruchterman.reingold(g1, niter=500)
plot(g1, layout=layout1, edge.curved = TRUE,  vertex.color= "grey", edge.arrow.size = 0.1, vertex.label.dist=0.5, vertex.label = NA)

```


```{r}
# d3 plot in web browser
# from the d3Simple package
devtools::install_github("d3Network", "christophergandrud")
require(d3Network)
d3Network::d3SimpleNetwork(get.data.frame(g1), width = 1500, height = 800, 
                textColour = "green", linkColour = "red", fontsize = 10,
                charge = -1000, opacity = 0.9, file = "d3net.html")

```


We can search for an author.
```{r}
sear <- cits[grepl("DE LEON", cits$ar), ]
# subset graph by label: http://lists.gnu.org/archive/html/igraph-help/2009-10/msg00039.html
degree(g, which(V(g)$name %in% c("A", "F", "X")))
# subset
subgraph.edges(g, V(g)[edge > 1])
```

```{r}
# export in gephi format
write.graph(g1, file="g.graphml", format="graphml")
# find location of gephi file
getwd()
```


Background and ideas:

http://kieranhealy.org/philcites/
http://kieranhealy.org/blog/archives/2013/06/18/a-co-citation-network-for-philosophy/
http://nealcaren.web.unc.edu/a-sociology-citation-network/

Choosing community detection algos:

http://igraph.wikidot.com/community-detection-in-r
http://bommaritollc.com/2012/06/17/summary-community-detection-algorithms-igraph-0-6/?utm_source=rss&utm_medium=rss&utm_campaign=summary-community-detection-algorithms-igraph-0-6
http://stackoverflow.com/a/9478989/1036500


```{r, echo=FALSE, message=FALSE, eval=FALSE}
# This chunck is to run the code and generate the PDF
# Load packages
setwd(wd) # assumes wd has been set earlier in the doc
require(knitr)
require(markdown)
 
# process .md and .pdf files (including smart punctuation and grey background of code blocks)
filen <- "citation_net" # name of this markdown file without suffix
knit(paste0(filen,".rmd"))
system(paste0("pandoc -s ", paste0(filen,".md"), " -t latex -o ", paste0(filen,".pdf"), " --highlight-style=tango  -S"))
```