Sentence Sanity Checker

Is there a way to get sentence sanity score using some NLP tool?
I wish to achieve the following:

I wish to some kind of meaning score against each English sentence. Is it possible to that by any means?
My use case is as follow:
sentence-1 : During the pendency of the proceedings, applications for impleadment were
meaning_score: 6/10 (lets say)

Sentence-2:allowed on 4 March 2011 and the operation of release orders passed by the
Union on 19 January 2011 was stayed.
Meaning_score: 7/10

But when I combine them I get a new sentence which has higher meaning score
Combined sentence: During the pendency of the proceedings, applications for impleadment were allowed on 4 March 2011 and the operation of release orders passed by the Union on 19 January 2011 was stayed.
Meaning score: 10/10

Is it possible to achieve something like this

I can’t think of any existing dataset for this off the top of my head, though I could be missing something. If you want a model that can do what you’re describing, you’re going to have to come up with some training data for it, and a model architecture that’s suitable.

how about using sentence similarity method? Any points over that. Lets say I generate sentence embedding for my input sentence and try to find most similar sentence( lets say I get some similar sentence). If I am able to get that similar sentence with some kind of confidence score it might help.
@mattg your views