My PDF converter turned every borderless table into loose numbers. The fix runs on a CPU.

Описание и применение

Конвертер PDF изначально не сохранял структуру таблиц, превращая их в разрозненные числа и ломая абзацы. После внедрения Camelot для borderless‑таблиц, разделения многострочных ячеек и Tesseract OCR для сканированных страниц, он теперь корректно генерирует Markdown‑таблицы и форматирует списки, выполняя задачу за 23 секунды на одном CPU за $0.001.

Задача

The initial PDF converter failed to preserve table structure, producing loose numbers and broken paragraphs, especially for borderless tables, dense budget tables, scanned pages, and two‑column articles.

Результат

The updated converter accurately extracts tables and scanned content, achieving near‑ML quality for key tables while remaining inexpensive ($2 per 1,000 pages) and fast (23 s on one CPU).

Рабочий процесс

• Run Actor via Apify API
• Use Camelot stream mode on numeric pages
• Split multi‑line cells into rows
• Rebuild Markdown from Tesseract OCR for scanned pages
• Map OCR bullet characters to dashes
• Discard non‑numeric two‑column prose
• Return Markdown chunks with metadata

Доказательства

My converter lost on the two things that mattered. On both tables the cells came out one per line, so the budget became a long column of loose numbers.

Источник

Camelot's "stream" mode (MIT licensed) finds those columns from the spacing alone, in 0.1 to 0.3 seconds a page on one CPU.

Источник

After the changes, on the same four files: three tables on the two-page borderless report, the budget as one nine-column table with the same rows as the ML converter, the scanned page with its list and headings. All of it took 23 seconds on one CPU, 21 of them OCR.

Источник