R LDA Topic Model How to get posterior for delta

Question

I ran LDA using the R package topicmodels and I have been trying to get the value for delta which is, in my understanding, the parameter of the dirichlet for words over topics. However, I was not able to access the value. I only managed to get the initial value using

LDA@control@delta

or

slot(LDA@control,"delta")

I know how to get alpha (parameter of Dir for topics over documents) for the posterior distribution, which is simply slot(LDA,"alpha") but how do get the delta?

Thanks a lot!

Answer 1

topicmodels uses a list of control parameters for the sampling method, here Gibbs sampling. By default values of alpha = 50/k and delta = 0.1 are assumed in control_LDA_Gibbs - you may, of course, specify other values. Maybe you have not specified your controls correctly. In any case, here a short example of code that should information on the deltaprior in the output. I hope that helps and solves your issue.

library(text2vec)
library(topicmodels)
library(slam) #to convert dtm to simple triplet matrix for topicmodels

ntopics <- 10
alphaprior <- 0.1
deltaprior <- 0.001
niter <- 1000

seedpar <- 0 

docssubset <- 1:500

docs <- movie_review$review[docssubset]

#Generate document term matrix with text2vec    
tokens = docs %>% 
  tolower %>% 
  word_tokenizer

it = itoken(tokens, ids = movie_review$id[docssubset], progressbar = FALSE)

vocab = create_vocabulary(it) %>%
  prune_vocabulary(term_count_min = 10, doc_proportion_max = 0.2)

vectorizer = vocab_vectorizer(vocab)

dtm = create_dtm(it, vectorizer, type = "dgTMatrix")

control_Gibbs_topicmodels <- list(
                            alpha = alphaprior
                            ,delta = deltaprior 
                            ,iter = niter
                            ,burnin = 100                           
                            ,keep = 50
                            ,nstart = 1
                            ,best = TRUE
                            ,seed = seedpar

)


ldatopicmodels <- LDA(as.simple_triplet_matrix(dtm)
                      ,k = ntopics
                      ,method = "Gibbs"
                      ,control = control_Gibbs_topicmodels 
                      )

str(ldatopicmodels)

ldatopicmodels@control@delta