I’ve been utilizing BERTopic to output some results of my own corpus, but I am unsure whether the results (the output of the Python code) can be interpreted the same way as that produced by the traditional method which is LDA.

I know that the two methods have different mathematical/computational mechanisms in terms of operation, but their outputs pretty much seem to be the same. So I was wondering if it is right to interpret the results of BERTopic the same way as I would interpret the results of LDA implementations.

The reason I am asking this question is because I haven’t been able to find papers or tutorials specifically explaining about BERTopic outputs and how the interpretation of its outputs is similar/different to LDA. Also, papers that utilize BERTopic for research is very scarce, making it difficult to study it.

Can someone provide some answers on this matter? If there is a link I can refer to or study regarding my question, as well as specific humanities/social science research examples that use BERTopic and compare it to LDA, that would also be greatly appreciated.

FYI, I am working in *computational social science* and trying to use BERTopic for my research, if this helps.