"I need to automate the conversion of thousands of DWGs to searchable PDFs using AutoCAD 2015."
So it is a one-time Job (in this size/for thousends of files), why it isn't possible to use a version 2016.1+
for this Job?
"If I change the SHX to TTF (which I can do using code/api) I risk text reflow and changing the layout of text on every sheet - "
Good point, well. If you want to go this way (using ttf) you should change the right ttf font to the right shx font,
But there is a residual risk for special characters. If you know the content of DWGs and if they created under
restrictet rules (Standards), it should the best way if 2016SP1 is not available.
"I've tried over a dozen OCR applications to post-process the PDF and none of them provide the accuracy I need."
Yes, bigger the Papersize and smaller the Text, it is not easy to find a software who working well with this pdf.
"What OCR did you use?"
I use ABBYY, OmniPage and Adobe Acrobat!
"What accuracy did you achieve?"
~90-100%
But this statement is surely unfair because i had best conditions 😉
I could choose the best possible font
Free text, adapt text rotations and width - because I created the PDF directly before and the contents of the DWG was known to a large extent.
And the plans are only DINA4 formats.
But I remember it well: that too was hard work until everything was coordinated.
Only details I can not remember, only that everything afterwards fully automatic worked.
And: I did not have to provide this as a solution for others, I did all the work and controlled.
My way, when it is not so problematic with these files:
Change all shx to ttf for these fonts you have a fit ttf (and/or you can create your own Fonts).
In My small sample file (i will try the uppload again) used ISOCP.shx and ISOCP.ttf side by side.
I know that this example is not a practical comparison, just a simple example.
No chance, i can not upload my *.pdf files 😞
