How We Took Our Table Extraction Accuracy From 10% to 84%. A Case Study.
How we measured our own table extraction accuracy against a real, diverse document corpus, including the parts that didn't go well. Full methodology, real code, and the files that still aren't perfect.
