textmineR (version 3.0.6)

FitLdaModel: Fit a Latent Dirichlet Allocation topic model

Description

Fit a Latent Dirichlet Allocation topic model using collapsed Gibbs sampling.

Usage

FitLdaModel(
  dtm,
  k,
  iterations = NULL,
  burnin = -1,
  alpha = 0.1,
  beta = 0.05,
  optimize_alpha = FALSE,
  calc_likelihood = FALSE,
  calc_coherence = TRUE,
  calc_r2 = FALSE,
  ...
)

Value

Returns an S3 object of class c("LDA", "TopicModel"). DESCRIBE MORE

Arguments

dtm

A document term matrix or term co-occurrence matrix of class dgCMatrix

k

Integer number of topics

iterations

Integer number of iterations for the Gibbs sampler to run. A future version may include automatic stopping criteria.

burnin

Integer number of burnin iterations. If burnin is greater than -1, the resulting "phi" and "theta" matrices are an average over all iterations greater than burnin.

alpha

Vector of length k for asymmetric or a number for symmetric. This is the prior for topics over documents

beta

Vector of length ncol(dtm) for asymmetric or a number for symmetric. This is the prior for words over topics.

optimize_alpha

Logical. Do you want to optimize alpha every 10 Gibbs iterations? Defaults to FALSE.

calc_likelihood

Do you want to calculate the likelihood every 10 Gibbs iterations? Useful for assessing convergence. Defaults to FALSE.

calc_coherence

Do you want to calculate probabilistic coherence of topics after the model is trained? Defaults to TRUE.

calc_r2

Do you want to calculate R-squared after the model is trained? Defaults to FALSE.

...

Other arguments to be passed to TmParallelApply

Details

EXPLAIN IMPLEMENTATION DETAILS

Examples

# load some data
data(nih_sample_dtm)

# fit a model 
set.seed(12345)
m <- FitLdaModel(dtm = nih_sample_dtm[1:20,], k = 5,
                 iterations = 200, burnin = 175)

str(m)

# predict on held-out documents using gibbs sampling "fold in"
p1 <- predict(m, nih_sample_dtm[21:100,], method = "gibbs",
              iterations = 200, burnin = 175)

# predict on held-out documents using the dot product method
p2 <- predict(m, nih_sample_dtm[21:100,], method = "dot")

# compare the methods
barplot(rbind(p1[1,],p2[1,]), beside = TRUE, col = c("red", "blue"))