Hey everyone! Hope you all had a great week. Here is a quick update on what I worked on:
What I did this week:
-
The React-based UI panel integrated with the pipeline is now merged! Now, the UI shows real-time progress bars and status updates while running in the background. The changes in the UI are being tracked in this issue: UI Discussion & Design Tracking · Issue #21 · joplin/plugin-note-categorization · GitHub (it also has the video of the current UI if someone wants to take a look).
-
I am also experimenting with using the parent notebook folder to improve clustering. I thought about doing one-hot encoding or embedding and blending the notebook names directly into vectors, but it didn't make much sense due to weak embeddings for short names and warping from extra dimensions. I'm leaning towards a two-stage post-processing approach (clustering by text first, then checking for misplaced notes in each cluster), but I'm still experimenting with this and will work more on it this week.
Plan for this week: I'm going to focus on implementing the post-processing logic to detect misplaced notes in each cluster. I also plan to introduce basic tag extraction using TF-IDF to auto-generate keywords for the clusters, along with computing centroids for each cluster.