snips-nlu-rs VS rustling-ontology

I am in shop for a speech recon library that works offline and can fit on a phone (Android).

I used to use Snips AI (https://snips.ai/), which worked well until it was acquired by Sonos. Now the portal is down and I can't modify the model anymore.

Looking for something ideally written in C or that can target C for portability. Free/libre and copyleft preferable to avoid the acquisition trap again.

Tapping into the vast pools of knowledge of HN; could you please suggest alternatives, preferably ones you have experience with?

Thank you.

Hey guys, I'm trying to write something that parses intents. I came across this: snips-nlu.

Yeah there is almost nothing NLPish out there for Rust, especially multi-lingual. That being said, Huggingface supports a number of models for Rust. Might be overkill for POS tagging (if there is even something, didn't check). I thought https://github.com/sonos/rustling-ontology included one but seems I misremembered

The problem is, this crate brings in rustling-ontology where there's a build script that trains all the models. The problem is, it retrains every time I change my code. It's making my development a pain and I feel like throwing my machine in the bin, honestly lol. Literally, every time I do a cargo c, I have to wait a couple of minutes for it to retrain all the models. I tried to trim it down, I made a fork and patched it where I removed all the languages (except for english), but there's still that noticeable delay where it retrains even when I only add a warn! somewhere. Is there anything else I can do, or try?