Perhaps I'll explain this a bit better…
Words can be converted into a vector representation by a word2vec
algorithm [1]. After conversion words will be a point in a high
dimensional space. Relations between words will then be vectors
between such points. Similar relations (or related relations) ca
Microsoft has unveiled an idea about a grammar and style tool for
Word. [1] I proposed something similar for detecting problematic
grammatical constructs in the content translation tools.[2] It is a
couple of years ago now, and I closed the task.
[1]
https://venturebeat.com/2019/05/06/microsoft-d