Ceci est une ancienne révision du document !
I was recently given a product key for Able2Extract 12, a PDF converter & editor. Previously, I’ve always done these sorts of tasks using various command-line tools. As extracting text from PDFs, or editing them in any way, is not a task I do too frequently, I cannot promise that I’ve tested everything the software has to offer. That being said, here are my experiences and thoughts.
Compatibility
While the software offers packages for only Ubuntu and Fedora, I was able to create a PKGBUILD that correctly installed and runs the Ubuntu .deb file under ArchLinux.
I did, however, run into an issue in Ubuntu 16.04, Ubuntu 17.10, and ArchLinux. Specifically, the application would crash with an error about the QT Fonts location. After contacting the company, we were able to resolve the issue. Apparently, the application requires the variable $QT_QPA_FONTDIR to be set to the root path. Instead of defining this system-wide in /etc/environment or in my user’s .bashrc, I instead created a bash script that sets the variable and runs Able2Extract. The script is:
#!/bin/bash
export QT_QPA_FONTDIR=/
/opt/investintech/a2ep/bin/Able2ExtractPro
I went this route because the Able2Extract package does not seem to add the bin to your $PATH variable, meaning it can be run only from the folder, or the .desktop file. After dropping the script into a folder on my PATH, I am able to run it as normal. This has the added benefit of not interfering with any other applications, should they want the same variable.
Application Interface
The layout of the application itself is very familiar (after having used software such as Adobe Acrobat), and it offers some helpful (non-intrusive) tips when starting it for the first time.
Features
The application allows you to create, edit and convert PDF files. Part of the conversion process utilizes OCR technology in order to convert PDFs to editable files such as documents (.odt), or presentation slides. It also offers the ability to create spreadsheets, CSV, HTML, images, and AutoCAD files.
I tested the Word, Excel, and HTML modes on a few recipe scans I have. Some of these files were taken via smartphone camera, and others were scanned on an actual flatbed scanner. The OCR system worked well for most of the files I tried, although one particularly badly photographed image had a few gaps where light reflections obscured the text. That being said, I could have filled in the blanks using logic, or adjusting the contrast of the image to make it more legible for myself. I was most impressed by the HTML results, as it actually added plenty of styling to the text to make it look clean and legible. If you were planning on turning PDFs into unstyled HTML files to add to a website, you should have a plan in place to strip out the inline styles. I did not see any options for the HTML converter.
The conversion options offered do allow you to handle things such as missing or unrecognized glyphs, or to set the file format for Word and Powerpoint conversions (on my system, it defaulted to OpenOffice). You can also do some document styling such as margins.
The creation tool selects an image file and turns it into a PDF - I did not see an option to select text documents or word documents (though you could create PDF files using a PDF printer or something like LaTeX). The editing tools include things like adding stamps, highlights, text, comments, etc. They also include things like redacting sections of files, deleting PDF pages, extracting specific pages, and adjusting text styles. The text style adjustment appears to work only on some PDFs - in my tests these options were grayed out. They probably work only on PDFs that were created from a text document, as opposed to image-based scans.
Results
As noted in the previous section, almost every attempt I made yielded a complete copy of the PDF. In some cases (low contrast, poorly lit, etc), there were some gaps in the resulting file. These could relatively easily be corrected or filled out (especially if you have access to the original document). The worst result came from a recipe that was in 3 columns - while the OCR system managed to correctly separate the columns (I’ve experienced some that treat 3 columns as 1 line), the character recognition of the actual text was not that impressive. The font in the PDF file was very small, and quite faint, which could have added to the lack of accuracy. The resulting file would have definitely needed proofreading and correcting (though most OCR files should be checked before deeming it finished).
Overall, the results I’ve experienced using Able2Extract 12 rivals any other OCR software I’ve ever used, and is much better than other Linux-based alternatives I’ve tried so far. Is it always perfect? No, but in every test I ran, it yielded a file that would have reduced the effort required to copy the file by hand by at least 50-60%. In most cases it would have required only a few small corrections.
Conclusion
If you do a lot of PDF work (splitting documents, OCR scans, etc), and don’t have an application for Linux to do this in, I would highly recommend giving Able2Extract a shot. Even if you have an application you use, you may not be happy with the OCR results - and then I would recommend you try Able2Extract as well.
It’s almost a perfect score - if the package worked out of the box, and if there were extra options for HTML conversions, I’d be happy to give it a 5.