JTAG

This is an application I wrote in the summer of 2003 that had the ability to manipulate, display, and classify rectangular sections of elements on each page.

Screenshots are available below (click for larger version):

 

Some of its features include:

• Ability to automatically segment and predict labels for those segments using interchangeable classifiers including:
  • k-nearest neighbours
  • logistic regression
  • gaussian naive Bayes (1 per class)
  • hidden markov model
  • maximum entropy markov model
• Selection information stored in ASCII flat files
• Ability to zoom in and out of pages, and navigate between multiple pages in the same article
• Easy to draw and label new regions. Just drag and drop with a mouse
• Ability to merge and split regions (horizontally or vertically)
• Hand-drawn regions are autmatically "snapped" to ink boundaries