Sample from the Weatherrescue Ben Nevis project

The ‘Meteorology of Ben Nevis’ project published 1.6 million weather observations in the Transactions of the Royal Society of Edinburgh - around 2000 pages of printed tables.

These observations have already been transcribed, by the Weather rescue Ben Nevis project, and this makes them a great choice for testing new transcription methods. We have combined the published images and transcribed data to make a validation dataset.

Ben Nevis project sample image, alongside the text blocks produced by Textract. Red and blue chequerboard patterns show tables Textract has found on the page.

• Script to run Textract
• Textract output (as JSON)
• Script to plot comparison figure
• A closer look at the block layout
• A closer look at the tables detected