Princeton Gerrymandering Project: Hofeller Files

Problem

Recently, the daughter of one of the GOP’s key gerrymandering strategists released a fileset belonging to her father. The files may contain information around controversial gerrymandering tactics used in redistricting, but in their current form they are almost impossible for journalists to navigate.

Solution

Online web application to navigate and search the Hofeller files dataset. Should be a tool for journalists to identify trends and patterns in the files and easily be able to navigate to related files. The application would provide an interface to search through the files and filter through different types of data.

The project would provide much needed structure to the data set to enable effective navigation and analysis of the files. In doing this it would enable a larger audience to dig into the files and gather their own insights from related and connected files.

See our code

About the Project

Features

Search through data
- Allow users to filter the data by metadata (file type, creation date, etc)
- Keyword-based file search querying
- Logical search queries
Provide easy-to-navigate interface for journalists to view related files
- Pull related files using sentence indexing and provide links to the files
Provide preview of specific files
- Identify and display column information for excel files
- Identify and display relevant snippets of file based on search query

Technical Challenges

Parsing and cleaning over 60 thousand documents down to plain text data from various file types (Word, Excel, PDF, ect.)
Using Elasticsearch on AWS for file search and tagging
Generating recommendations of similar files using Universal Sentence Encoder and nearest neighbor searching

About the Client

The Princeton Gerrymandering Project aims to give activists and legislators the tools they need to detect offenses and craft bulletproof, bipartisan reform. Their analysis is published widely, and their work is used by legislators and reformers of all communities, without regard to partisan affiliation.