neurosynth_wordcloud.R

% This script generates wordcloud from neurosynth terms
% @author Lena Dorfschmidt (ld548@cam.ac.uk)
% It is advised to 'clean' your neurosynth terms, i.e. delete anatomical terms and potentially duplicates

library(wordcloud)
library(scales)
# Change this path to your directory
basepath = '~/Documents/GitHub/neurosynth_analysis/'

# Load in table of neurosynth terms: column 1: words, column 2: weights
words <- read.csv(paste0(base.path, 'MI_terms_neurosynth_no_atomical.csv',header=FALSE))
nWords = 50 # choose how many words to display

# Split the terms into positive and negative weights for wordcloud generation:


words.pos = words[words$V2>0,] # choosing only positive weights
freq.pos = words.pos$V2/sum(words.pos$V2)*100 # transform into frequency
pdf(paste0(base.path, 'wordcloud_pos.pdf')) # save image as pdf file
wordcloud(words=words.pos$V1, 
          freq=freq.pos, 
          max.words = nWords, 
          min.freq = min(freq.pos), 
          random.order = FALSE,
          scale=c(3,.2), 
          colors = '#C90C0C')
dev.off()

words.neg = words[words$V2<0,]
words.neg = words.neg[((dim(words.neg)[1]-nWords):dim(words.neg)[1]),]
freq.neg = abs(words.neg$V2)/sum(abs(words.neg$V2))*100
pdf(paste0(base.path, 'wordcloud_neg.pdf')) # save image as pdf file
wordcloud(words=words.neg$V1, 
          freq=freq.neg, 
          max.words = nWords, 
          random.order = FALSE, 
          min.freq = min(freq.neg), 
          scale=c(2.5,0.05), 
          colors = '#0C13C9',
          family='Helvetica') 
dev.off()