A really useful feature of BERTopic is the ability to save a model, reload it later, and run predict on new data that is then sorted into the same clusters as before. Is there anything specific about toponymy that would not allow this?
If you don't have the time/need for this feature I don't mind giving implementation a go sometime myself.
A really useful feature of BERTopic is the ability to save a model, reload it later, and run predict on new data that is then sorted into the same clusters as before. Is there anything specific about toponymy that would not allow this?
If you don't have the time/need for this feature I don't mind giving implementation a go sometime myself.