Thanks for teaching me something new!
So Chromium is based on Blink, which is LGPL – a less viral GPL. Hence, it can serve as a dependency in closed-source software.
As to the shared heritage of these well-established projects – I don’t know how else to interpret it other than a testament to the complexity of building a decent browser engine.
Btw, quick shout out to Orion, a rare WebKit browser by the makers of Kagi that’s apparently coming to Linux as well. I’m a monthly supporter. Even though I still mostly use Vivaldi, it’s been coming along really nicely. Proprietary software but idc. I appreciate their unspoken mission statement: pay or be the product. (No-one should be a product, obviously, but that’s capitalism.)
ML engineer here. My intuition says you won’t get better accuracy than with sentence template matching, provided your matching rules are free of contradictions. Of course, the downside is you need to remember (and teach others) the precise phrasing to trigger a certain intent. Refining your matching rules is probably a good task for a coding agent.
Back in the pre-LLM days, we used simpler statistical models for intent classification. These were way smaller and could easily run on CPU. Check out random forests or SVMs that take bags of words as input. You need enough examples though to train them on.
With an LLM you can reframe the problem as getting the model to generate the right ‘tool’ call. Most intents are a form of relation extraction: there’s an ‘action’ (verb) and one or more participants (subject, object, etc.). You could imagine a single tool definition (call it ‘SpeakerIntent’) that outputs the intent type (from an enum) as well as the arguments involved. Then you can link that to the final intent with some post-processing. There’s a 100M version of gemma3 that’s apparently not bad at tool calling.