While I know that the ctclib-kenlm-sys crate is not a general purpose binding library, I'm currently in need of such a binding and your work is a good starting point.
I have a first question, related to KenLM's new method that requires a dict. I have used kenlm's query binary and did not have to use this to do queries. So I forked your repository and made some changes (making Model public, using base_score and then computing perplexity) and on the (tiny) samples that I have tried, I got the same results than the query binary.
So I'm wondering if I should try to add these changes (in a non breaking way, obviously) insode this lib or if you'd prefer me to fork the ctclib-kenlm-sys and make something more opinionated.
Let me know what you prefer, or if there's a way of using the KenLM struct in the way I intend to do, without changing code :)
I have pushed my changes on my local fork here, and I'll publish it on crates.io so that I can use it in my projects. If you are interested in me PRing my changes to your repo let me know!
Hello,
While I know that the ctclib-kenlm-sys crate is not a general purpose binding library, I'm currently in need of such a binding and your work is a good starting point.
I have a first question, related to KenLM's
new
method that requires a dict. I have used kenlm'squery
binary and did not have to use this to do queries. So I forked your repository and made some changes (makingModel
public, usingbase_score
and then computing perplexity) and on the (tiny) samples that I have tried, I got the same results than thequery
binary.So I'm wondering if I should try to add these changes (in a non breaking way, obviously) insode this lib or if you'd prefer me to fork the ctclib-kenlm-sys and make something more opinionated.
Let me know what you prefer, or if there's a way of using the
KenLM
struct in the way I intend to do, without changing code :)