dist2() with RWMD method bug/error?

[linneaturco]

I'm trying to use dist2() with a RWMD model method to calculate distances, but continue to receive an error that the RWMD method is an "unused argument". All the documentation for the package still contains the ability to use an RWMD model as the dist2() method, so I'm wondering whether there's a problem on my end?

I've included code below which I hope might be helpful, and many thanks for any help you can provide.

text_corpus <- corpus(txtstatesyears$text, 
                      docvars = docvars(txtstatesyears))
text_tokens <- tokens(text_corpus)
feats <- dfm(text_tokens, verbose = TRUE) %>% 
  dfm_trim(min_termfreq = 4, min_docfreq = .25, docfreq_type = "prop") %>% 
  featnames()

text_tokens <- tokens_select(text_tokens, feats, padding = TRUE)

text_fcm <- fcm(text_tokens, context = "window", count = "weighted", weights = 1/(1:5), tri = TRUE)

glove <- GlobalVectors$new(rank = 50, x_max = 10)
word_vector <- glove$fit_transform(text_fcm, n_iter = 10,convergence_tol=0.01, n_threads = 8)

wv_context <- glove$components
averaged_word_vectors <- word_vector + t(wv_context)

rwmd <- RelaxedWordMoversDistance$new(text_fcm, averaged_word_vectors)

un.rwmd.l <- list()
for(i in 1993:2018){
  print(i)
  sub.i <- corpus_subset(text_corpus, year == i)
  tokens <- word_tokenizer(tolower(sub.i))
  it <- itoken(tokens)
  v <- create_vocabulary(it)
  vectorizer <- vocab_vectorizer(v)
  ir_dtm <- create_dtm(it, vectorizer)

  rwmd_dist <- dist2(ir_dtm, method = rwmd, norm = "none")
  rwmd_norm <- (rwmd_dist-min(rwmd_dist))/(max(rwmd_dist)-min(rwmd_dist))
  rwmd_norm_sims <- 1 - rwmd_norm

  diag(rwmd_norm_sims) <- 0
  colnames(rwmd_norm_sims) <- docvars(sub.i, "country")
  rownames(rwmd_norm_sims) <- docvars(sub.i, "country")

  un.rwmd.l[[i]] <- rwmd_norm_sims
}

[dselivanov]

Hi @linneaturco . Thanks for report, it seem I need to update documentation. Can you try rwmd$sim2(x)/rwmd$dist2(x) functions from RelaxedWordMoversDistance class? see examples in ?RelaxedWordMoversDistance

[linneaturco]

Thanks @dselivanov. I am able to run rwmd$dist2(x) and I should have realized that the dist2() documentation was obsolete for RWMD sooner.

[dselivanov]

see example in #326