A better generator for using Keras in R for word2vec

gistfile1.R

# the example code provided by RStudio to use R for word2vec has an error if you don't have enough data.
# This fixes the error by having the generator reset when it runs out

skipgrams_generator <- function(text, tokenizer, window_size, negative_samples) {
  gen <- texts_to_sequences_generator(tokenizer, sample(text))
  function() {
    next_value <- generator_next(gen)
    if(is.null(next_value)){ #if there isn't new text from the generator
      gen <<- texts_to_sequences_generator(tokenizer, sample(text)) # remake the generator
      next_value <- generator_next(gen)
    }
    skip <- next_value %>%
      skipgrams(
        vocabulary_size = tokenizer$num_words, 
        window_size = window_size, 
        negative_samples = 1
      )
    x <- transpose(skip$couples) %>% map(. %>% unlist %>% as.matrix(ncol = 1))
    y <- skip$labels %>% as.matrix(ncol = 1)
    list(x, y)
  }
}