vgherard / kgrams Goto Github PK

The CRAN policy contains

• Packages should not attempt to disable compiler diagnostics, nor to remove other diagnostic information such as symbols in shared objects.

yet packages

... kgrams ...

attempt to do so.

Do correct before 2021-11-26 to safely retain your package on CRAN, and as this will require manual check, do allow time for the CRAN submissions team to do so.

When a language model is memory intensive and therefore slow to build, it would be useful to be able to save the model and then read it into R for further work at a later date. I can't see how to do this. Is it possible?

This should be possible with clever choices of preprocess() and tknz_sent(), plus some post-processing step for output text

Oddly enough, when called from {testthat} it appears to run normally...

@vgherard, thank you for this very interesting package.
Is it possible to predict new words by using a model building with k-grams package ?
I know it is possible with the package sbo.

e.g.:

library(kgrams)
f <- kgram_freqs(much_ado, N = 2)
dict <- attr(f, "dict")
dict
#> A dictionary.
length(dict)
#> [1] 0

^{Created on 2021-07-17 by the reprex package (v2.0.0)}

With RcppParallel

Hi there,

I noticed kgram_freqs objects (and also language models I suspect) cannot be saved with R saveRDS(). Once kgram_freqs is loaded back with readRDS() it appears to have invalid C pointers, leading to errors when trying to use the object.

Is there a way to store a kgram_freqs on the disk for later reload?

Thanks!

See the rOpenSci package development guide.

TBD before submission to rOpenSci, c.f. the presubmission inquiry ropensci/software-review#452.