NER with unsupervised learning

I am wondering how an NER task can be performed with no/little labelled data? I have been using a rule based approach however I wonder if any of the data centric approaches can be used for this problem?

If you are working on a specific problem and you only want to detect the specific names then you may custom train your NER model on top of the pre-trained NER model by assigining a new entity in your dataset. It will solve your problem, I have trained 10+ custom NER models for my use cases.