search

facebookresearch / StarSpace

Learning embeddings for classification, retrieval and ranking.
MIT License
3.94k stars 531 forks source link

Unsupervised topic classification #199

Closed helgasvala closed 5 years ago

helgasvala commented 5 years ago

Hey!

For topic classification, I want to compare StarSpace to LDA, which is unsupervised so in order to have a fair comparison I want to make my own embeddings for StarSpace first. My question is, once I make the embeddings, can I somehow divide them up to as many topics as I set as parameters in the LDA? How could I extract clusters from the embeddings?

Thank you so much!

ledw commented 5 years ago

@helgasvala Hi, you could try standard clustering methods on embeddings trained from StarSpace. I'd suggest to try a different set of number of clusters to form, in order to compare with topics from LDA.