Knowledge graphs vs. Word embeddings

I am working on a project that utilises KG as a search engine in an internal repository of articles. (Similar to: Building An Academic Knowledge Graph with OpenAI & Graph Database | by Fanghua (Joshua) Yu | Medium)

I have been struggling to understand why custom trained word embeddings can't replace knowledge graphs in such a use case.

If someone could shed some light on this, it would be very helpful.

Thanks,
Anita