Python Translation of PDF document to JSON

Abstract: Take a publicly available, poorly formatted PDF format document with a variety of tables and turn it into consistent, accurate, hierarchical JSON. Every year the US publishes information about facilities that it maintains on its own soil and in other countries. [[login to view URL]] This information is published in a PDF format, with many pages of text and explanation, and long lists of tables. Our immediate goal is to get these tables into a reasonable hierarchical document in a machine-readable format, ideally json. This will allow us to represent the data in a web site, but this second step is not included in this bid. Only the first step of turning the PDF into structured data is covered in this bid. Complications to the task include oddly formatted page numbers, table sections that bridge different pages, and other issues that prevent a simple conversion. We strongly prefer that the processing be done in Python. We are currently only looking at converting the 2009 document, but our longer term goal is to get a system that would be able to read multiple years of this document: it is largely similar from year to year. If the bidder can demonstrate that their automated method works over the 2008 and 2007 versions of the document (without significant modification or overly special case-based code), we will add %20 to the value of the successful bid. Requirements: Project should be executed in Python. Python program should take PDF in and yield accurate hierarchical JSON data. Final data format should include the following in key/value pairs: *URL of document that created this datum *Country *Base name *All other columns in the PDF document, including None for blank columns Code will be licensed GPLv3