ddangelov / Top2Vec

Top2Vec learns jointly embedded topic, document and word vectors.
BSD 3-Clause "New" or "Revised" License
2.95k stars 374 forks source link

online topic modelling #341

Open ptear opened 1 year ago

ptear commented 1 year ago

as I understand it online topic modelling might be possible with Top2Vec, by calling add_documents and compute_topics together.

1) does this maintain currently existing topics and only add new ones? or does it completely refresh the topics?

2) how does the approach used in Top2Vec differ from that used by BERTopic, which uses River?

looking at the code it seems like compute_topics does a complete refresh, but just want to clarify. if this is the case, I might look at implementing something similar to BERTopic's approach for Top2Vec