Using BERT embeddings with c-TF-IDF for Topic modeling in Russian language
Getting embeddings, tokenizing and lemmatizing large chunks of data takes some time, and if you just want to see how it works, you can download already preprocessed dataset of russian news here -> Google Drive.
Or you can just scroll down to the end of the notebook below and play with the visualisation
Name | Link |
---|---|
ru-BERTopic main notebook |
Visualization implemented with plotly provides semi-interactive figures. You can play with it in the notebook above. Hovering over a point representing a single document reveals its topic and id for deeper digging-in.