Looks really neat! I will try it out for my next ~/tmp weekend project. Meanwhile, I noticed that the link to the C# project is broken on this page https://seed.run/docs/adding-dotnet-core-projects . I wanted to try propose the change but I couldn't find the repository on your GitHub.
Not really a product (yet), but I'm researching and trying to create an accessible natural language understanding algorithm without the use of deep neural networks (or at least, not in any significant way). In my opinion, neural networks are used in the wrong way when it comes to natural language processing such that they make the whole understanding process too opaque. Instead, I'm trying to consider the NLU pipeline as a graph problem for which I can use any model to speed up the search but where it can work without it as well.
It's a long shot and a big topic, but I've managed to enjoy it along the way so far by coding different NLP algorithms (CYK parser, dependency parsers, tokenizer, etc.) and trying to publish some of the code (C#) I made as NuGet libraries for others to use.
The team behind it is part of a group of computational linguistics researchers in Chalmers university in Gothenburg. It's a very non-machine learning approach but I think they've had some ambition of trying a ml/rule based hybrid approach at some point. Personally I think some hybrid approach should allow the best possible flexibility.
