Visualizing LDA results

Purpose

This vignette creates .json file used by LDAvis.

Creating .json

library(parallel)
library(LDAvis)

###################################
# need:
#    var_phi
#    var_theta
#    document token counts,   
#    vocab (colnames(var_phi)), 
#    var_corpus_word_frequencies
###################################

var_num_cores = 4
var_parallel_cluster = makeCluster(var_num_cores)

print ("load from file")
vtheta = readRDS(file="./data/topicDocs.Rds")
vphi = readRDS(file="./data//topicWords.Rds")

print ("theta: ")
print (dim(vtheta))
print ("phi: ")
print (dim(vphi))


doc_lens = unlist(readRDS(file="./data/doc_lens.Rds"))
vocabulary = colnames(vphi)
wf = readLines("./data/freqs.txt")


print ("creating json object")
obj_json <- createJSON(
  phi = (vphi), 
  theta = (vtheta), 
  doc.length = doc_lens,
  vocab = vocabulary,
  term.frequency = wf,
  cluster = var_parallel_cluster
)

print ("writing to file")
write(file ="./ldavis.json", obj_json)

Running LDAvis

Note that if you want to see it locally you either need to have a web server or use Mozilla Firefox, otherwise there will be a CORS error as the json data is being loaded. This can be fixed by renaming the json file as a .js file and loading it in as a script.

library("servr")
library("LDAvis")

serVis("./105k_ldavis.json", out.dir="./view")