Automatic Bookmark Organizing with Document Clustering.

Daniel Zheng, daniel.zheng@pitt.edu

Originally, I wanted to create a way to automatically clean up my bookmarks bar. I have hundreds, maybe thousands of bookmarks floating around not sorted into their proper folders.

I then thought, it would be cool to make a Chrome extension that takes list of bookmarks (URLs), automatically clusters them into folders, and rearranges your bookmarks bar!

This would entail:

• Learning Chrome's bookmarks API and making the extension
• Hierarchical clustering
• Cluster labeling (folder names)

For this project, I decided to focus on clustering and labeling. K-means and Hierarchical methods were investigated.

Dataset

For this project, I utilized a 12GB text dump of wikipedia found here. Since the clustering is unsupervised, it isn't necessary to run the algorithms on the entire dataset, though I eventually will do so. In the data folder are a small sample of JSON files of wikipedia articles.

Repository Contents

• Practicing topic extraction
• Visitor log
• Final Report
• K-means Clustering and Visualization
  • Jupyter Notebook
  • Markdown
• Hierarchical Clustering and Visualization
  • Jupyter Notebook
  • Markdown
• Class Presentation Slides
• Data, JSON files of wikipedia articles
• Images, folder of all images used in repository
• Cluster Pickle Files, saved .pkl files of clustering outputs
• Preprocessing
  • Wikiextractor, a submodule of the repo used in converting wikipedia XML dump to json
  • Reformatting script, a short custom script for some further preprocessing to convert from JSON Lines format into one dict per file.
• License
  • GPL License
  • License Explanation

NOTE: Before running any of this code, I would recommend installing all dependencies with pip install -r requirements.txt.