Does OCR for rapid Document creation already exist in Writer, Impress and Draw?
If not, is it possible to add this feature?
Thanks in advance for any help.
Regards, have a nice day.
There are many third-party programs to do this. IMHO such a feature, though very useful, is not a priority target for LO. Given the difficulty of the topic, adding it to LO would severely impact the reliability of the suite by diverting rare developer resources which are already insufficiently to address all reported bugs.
I see how OCR can facilitate initial Writer document entry but I have some difficulty about Draw (graphical design component) and Impress (presentation component) where graphics and geometric layout are of prime importance. Anyway, OCR is involved with “text” not with shape. Consequently it can’t rebuild a vector description of an image. Maybe in a few years or decades.
LOL for your last sentence.
Ok, but maybe embed something like Tesseract, or provide a plugin for it?
In the early 2Ks, ABBY Fine Reader was able to recreate the document layout, other than just the contents.
Unfortunately, it’s not an Open Source product.
Thanks for your help, regards, have a nice day.
If the document is text only, this can be done “easily” because layout is rather simple: you have optional header and footer; the rest of the page is supposed to be composed of “connected” lines forming a paragraph with spacing above and below.
However I have seen many OCR apps which handle long documents page per page where user has to “inter-connect” the pages to provide uninterrupted text flow.
I posted this very question on fosstodon several months ago.
AFAIK, there is no open source GUI application which will do what you are looking for.
I use AbbyFineReader on macOS, but it is a non-open-source, paying application. The current owners of FineReader are also attempting to push a subscription based model for their program.
Clearly, there is a demand for such a program, at least in a business environment, and none of the OSS programs I’ve tried so far on my Linux machines come anywhere near close in functionality. That is the sad state of affairs currently, unless there is something I don’t know about or is somewhat obscure.
As for integration into LO, I’d argue that it isn’t really the aim of an office suite to provide this functionality. I can’t think of any other office suite that includes a multilingual GUI OCR capable module.
@iplaw67 Could you tell which FOSS OCR programs you tested? I queried my Fedora repo with “ocr”. It returned gocr, ocrad, ocrmypdf (which do not seem GUI but can easily be “integrated” in a script or invoked from scanning software like xsane). There are also R-tesseract, paperwork, tesseract.
I have used none of them but I’m interested by a review.
I tried Tesseract, but it was ages ago.
It has an interesting history by the way, in beginning it was hp patent.
Have a look at Wikipedia about it maybe.
For text recognition alone, I have used Lios, which provides a GUI and, if memory serves, integrates with tesseract, but it knows nothing about preserving layout, and even the character recognition is somewhat inconsistent.
OCRFeeder was probably the most capable, but still struggled with correctly identifying and allowing the the user to define, specific areas of a complex document for ocr processing. This is where the proprietary software offerings have a distinct advantage. The problem is that real world PDF documents that are scanned images often contain artefacts, handwritten comments or signatures, overlaid images on text and tables, and I’ve yet to find any open source equivalent that allows me to achieve, in the same time and with the same minimal effort, an exported editable word processing document that maintains the layout of the original. Clearly, the investment made by proprietary software companies in making the ocr and export process as efficient, accurate, and easy as possible, is where the advantage is to be seen.
2 Likes