Nice demo, really useful! Rather than a comment as the cluster title, is it possible to have one or two words that sum up that cluster? Would that require analysing each cluster, or is there a better way?
I'd love to see sentiment per comment and perhaps als using NLP or an LLM to give each comment (or each cluster) a label/phrase so that you can identify them in the visualization.
Using UMAP that kind of thing is possible - a while back I did it with speech embeddings and the cool thing was that all the similar sounding people were nearby and with a bit of a Bokeh hack it was possible to listen to the one you hovered over, so it should be similarly possible to display the item text
![Felix "Unfair" | [Stray Kids : SKZ-PLAYER]](http://i.ytimg.com/vi/Oswujxm2Ag0/mqdefault.jpg)
This is great; thanks so much for sharing.
Nice demo, really useful! Rather than a comment as the cluster title, is it possible to have one or two words that sum up that cluster? Would that require analysing each cluster, or is there a better way?
Yes I think that would be way better. I guess we could feed the comments for each cluster into an LLM and ask it to give us a name for that cluster?
I'd love to see sentiment per comment and perhaps als using NLP or an LLM to give each comment (or each cluster) a label/phrase so that you can identify them in the visualization.
Using UMAP that kind of thing is possible - a while back I did it with speech embeddings and the cool thing was that all the similar sounding people were nearby and with a bit of a Bokeh hack it was possible to listen to the one you hovered over, so it should be similarly possible to display the item text
I did look into UMAP but I thought that was more used for dimensionality reduction, so it'd be comparable to t-SNE?
Your toolbar on your editor bothers me. It is soooo tiny. 😅
Haha yeh I have the font size extremely large otherwise you wouldn't be able to see anything! But the side effect is the the toolbar is tiny!