hOCR for pdftohtml?

[bmix]

Hi!

This is not a bug or whish, it is just a thought I was having an hour ago.

I know, this is not the main-poppler repository, but this repo got referenced in Poppler's bugzilla in an issue, that covered the mistmatch of the <b> and <i> elements (i.e. Xerces validator: Unexpected element "b". The content of the parent element type must match "(#PCDATA)".) and, since you work on pdftohtml, and I do have an account already here and am too lazy to sign up there (or the mailing-list) ;-), I raise this here. I hope you don't mind.

The XML, that gets created with the -xml switch could be, as it seems, fully replaced by the hOCR (Wikipedia) microformat, which gets more and more support (Tesseract-OCR supports it out of the box). It is also XML, since they use XHTML. The specification for hOCR-1.2 is here.

I think, it may be a a little less elegant, since they use the <title> element to store absolute positioning and font-config (and other) info, but using it, would mean one XML format less to be maintained (even, if it is a very simple one) and people, who have already written XSL-Transforms could re-use their stylesheets. Maybe, hOCR could even replace the main-output of the HTML pdftohtml creates? The CSS should be easily doable in an XSLT, that may be even included inline into the output.