Closed jiqicn closed 1 year ago
Performance of the topic modeling function is not ideal.
With the sample dataset containing around 370 reviews (nearly 1200 chunks), I got this:
Regarding the official document of BERTopic (here), it seems that no more efforts could be done from the topic modeling package part, except using GPU for calculating the embeddings.
Performance of the topic modeling function is not ideal.
With the sample dataset containing around 370 reviews (nearly 1200 chunks), I got this: