# Line-numbered PDF to EPUB: not handled yet

Source: https://epubcc.com/cases/line-numbered-legal-text/ (updated 2026-11-04)

Line numbers in the margin are not handled yet: on a 50-page U.S. Senate bill the EpubCC engine writes the numbers 1 to 24 as a stack of short paragraphs before each page's text and splits about 24 words per 1,000. Other engines do worse on continuity (8-gram recall 0.09 to 0.51 against our 0.559), but none gives a clean result.

Problem: Every line numbered 1 to 24 in the left margin, a page number at the top, hyphenated words at the end of short lines, and deep indentation for each level of the bill.

- Limitation: The line numbers are kept as a run of one-number paragraphs at the start of each page instead of being removed.
- Limitation: Words hyphenated at a line end are sometimes split with a space instead of joined («serv ice», «sub ject»): 23.6 split words per 1,000 in our EPUB.
- Limitation: Indentation levels ((A), (i), (I)) are flattened; the outline of the bill is not rebuilt.
- Limitation: No chapters are detected, and the engine warns about it.
- Limitation: Our 8-gram recall (0.559) is the highest of the engines measured, which says more about how hard this layout is than about the quality of our result.