Minimal example of historical data parsing using layoutparser

Setup

• Set up google cloud vision API

  • Create a google service account service by following this guide
  • Download a key (should end in .json) to a safe location

• Or set up tesseract

  • Follow instructions listed here

• create a clean conda/venv environment

• Install layoutparser using the instructions on their github ; their website seems out of date

Input

Random historical table picked from 1951 UP census

Processing

DocParse.ipynb runs layoutparser and extracts lists to produce a table. Here, we're interested in the 2nd numerical column (which is population).

Output

Some errors, but much preferable to manual entry.