Would it be possible to create a QA check that would work without a glossary or termbase? What I mean is, the program scans both the source and target text and tries to find any inconsistencies in terms of terminology, but without referring to term pairs stored in a glossary or termbase.

I will use a simplified example to illustrate what it would do:

For example, it spots that certain source segments contain the Dutch word "organisatie", and that in most cases there is an English word "organization" in the target text. However, the program also notices that in one source segment containing the Dutch word "organisatie", the corresponding target text contains the word "organisation". It therefore flags this segment as containing inconsistent terminology.