There was no great coding work on our part: we simply built their command line example and then wrapped that with a bash script.

Some stuff works fine in a desktop scripting model, but there is a threshold beyond which a more batch-level solution is best.

yes, that didn’t work out on my side. I’ve now bought the Linux engine of ABBYY (150 EUR for 12000 pages a year) and written a wrapper around it that recursively and autonomously iterates my directory structure and runs the engine on all documents that have not yet been treated. I’ve made it open source:

http://www.mnsoft.org/547.0.html?&cHash=9120c122ed&tx_ttnewsbackPid=544&tx_ttnewstt_news=30

and

pdfocrwrapper.sourceforge.net.

Tested on thousands of pages, works perfectly.

HTH,

M

This sounds like pure AppleScript, isn’t it? Even if the apps don’t provide any explicit AppleScript support, you can still have it drive the keyboard for you and automate the process.

The goal behind my own script was to allow FineReader to run unattended while I was sipping tea somewhere else, and it does that just swimmingly.

One point of difference: you are using FineReader Express 8, while I am working with FineReader for ScanSnap. If anything, I would assume yours provides more functionality.

I am trying to make this work with ABBY FineReader Express 8 for Mac. Making some progress in that I can hand over to the program the PDF I want to work on; then I still have to tell the program “do it”, afterwards “save”; then next file.

Do you think there’s a way to remote control the program entirely? I always want to use the same settings, but unattended.

Thanks,

M

No, in my home life the docs are few enough that I name and file them by hand. I imagine that some of the heavy hitters in the document management field will automatically populate keywords based on content, but then they also have an annoying habit of squirreling your documents away into the deep recesses of their black box.

With available command line tools it ought to be pretty easy to do some regular expression matches with the content to attempt to categorize and then name a document.
My main concern would be the inherent slushiness of the OCR process. How do you write a simple app that looks for errors like “Citihank” and “Ciflbank” and correctly interprets them as “Citibank”?

I guess if I were writing such an app, it would display thumbnails of documents along with their proposed names, and a list of “undetermined” docs. Then you could go down the list and approve or reject the filename changes. Certainly most would be correct, but one or two would be misnamed.

If you are looking into a more substantial solution, for a small organization or business, you might be interested in talking with ABBYY about licensing their engine. This is what we did at work for a particular job: we bought a license to use their engine on a single Linux server for a specific number of pages per year.

What they provided was their SDK along with platform-specific binaries. We were able to compile a command line client to the SDK that worked wonders. We then used a bash script on a cron job to wrap the command line client, periodically checking a source directory for files and feeding them to the CLI.

To stop all dock bouncing forever, type the following in a terminal window:

defaults write com.apple.dock no-bouncing -bool TRUE

and then

killall Dock

To reverse the process, do the same thing, but use FALSE instead of TRUE.

As far as the quality, this is likely a consequence of settings in Finereader.
I launched FineReader for ScanSnap Preferences and clicked on the Scan to Searchable PDF tab.

On this tab I have selected the following:

Save mode: Text under page image
Quality: High (for printing)
Format: Automatic

I imagine your Quality setting might be on one of the lower settings right now.