Text Mining and Visualization

Topic Modeling 261

11.6.6 Comparison of Inference Techniques

In order to examine the similarity between the two model outputs, we can use a distance

measure using Hellinger distance to calculate how similar topics are to each other. In ad-

dition to the topicmodels package, the package clue is required to compute the distance

similarities.

clue: used for creating and analyzing cluster ensembles and includes methods for measur-

ing proximity (http://cran.r-project.org/web/packages/clue/index.html)

The following code is used to calculate the similarity between the two models (LDA

Gibbs and LDA VEM):

#COMPARISON USING HELLINGER DISTANCE:

library("topicmodels")

library("clue")

#Set the working directory

setwd("C:/Users/Documents/R/DATASET/")

#Load the models ...