Model for sentence similarity based on syntax?

Beginners
1

Are there models that capture syntactic similarity of sentences?

Sentence transformers and others capture semantic similarity, that I don’t want.
For example, here

Sentence1: What is the count of singers in Alabama who are under 30 years in age?
Sentence2: Who is the most attended singer in Alabama?
Sentence3: What is the count of registrations in Jersey of people aged under 50?

The demo gives 0.76 similarity on 1-2 and only 0.5 on 1-3, though 1-3 are more close by syntax.

2

Have you tried clustering with kmeans with different measures? I would also experiment with different tokenizers see if this helps get you closer to what you want.