PDF to Editable Word Doc

Yes.

You need an Optical Character Recognition Program or OCR. TextBridge is one. If you've ever purchased a scanner you may have one of these programs.

When the hard copy was scanned in, it was scanned in as an image. What you need to do is convert it to text; that's what OCR programs do.

Now, OCR reads sort of like a dyslexic child, so you'll likely find some patterned errors; certain letter combinations that appear like other letters. You will need to proof.

If it were me, I'd keyboard the texts in, and edit as I typed.

I'm with Medievalist. Most likely, as far as Acrobat and Word are concerned, there are no words to edit. The document was likely scanned as a series of page images. OCR software is needed at this point to derive the characters out of the images.

Actually, there are a bunch of freeware PDF-to-text converter programs available, some of which have been around for quite a while. I used one for some documents several years ago, and it worked as well as an OCR scan of a paper document would, though it was a bid kludgy. It's on another computer where I work, and I can't recall the exact name of the program, but I just groooogled "pdf to text converter free" and got a lengthy list of such utilities.

caw

Blacbird: I would not be surprised to find out that there's an optical character recognition engine at the heart of those applications.

I'm no programmer, but it seems to me that the basic concept is the same.

blacbird said:

Actually, there are a bunch of freeware PDF-to-text converter programs available, some of which have been around for quite a while. I used one for some documents several years ago, and it worked as well as an OCR scan of a paper document would, though it was a bid kludgy. It's on another computer where I work, and I can't recall the exact name of the program, but I just groooogled "pdf to text converter free" and got a lengthy list of such utilities.

caw

Click to expand...

I suspect that some of those are OCR apps, and others are lifting the text from the txt resource embedded in the postscript header of the .pdf.

But image-based .pdfs just have the postscript binary, and lack the txt resource in the header.

Can you select text from the .pdf? Or is it a series of embedded images?

FalconMage said:

Blacbird: I would not be surprised to find out that there's an optical character recognition engine at the heart of those applications.

I'm no programmer, but it seems to me that the basic concept is the same.

Click to expand...

I'm quite sure it is. The point I was trying to make is that you don't have to physically print and then physically scan a PDF to get it converted to editable text, which seemed to be implied by earlier posts.

caw

Medievalist said:

Can you select text from the .pdf? Or is it a series of embedded images?

Click to expand...

Good question. I'm not sure, at least in the case of the one I used. I got it for a specific business-related purpose, it worked okay, and still resides on my work computer, but I don't recall having tried to select portions of text from the PDF. I'll take a look at it during the coming week and see if I can ferret out more details. If I find anything useful, I'll get back to this thread with a report.

caw

blacbird said:

I'm quite sure it is. The point I was trying to make is that you don't have to physically print and then physically scan a PDF to get it converted to editable text, which seemed to be implied by earlier posts.

caw

Click to expand...

No, of course not. OCR software is designed to "read" .pdfs that have embedded images. The images embedded in the .pdf are usually .tiff but they may be .png or .jpg.

The software can be pointed to a directory; better packages let you train it to read a particular typeface etc. in a document by providing a way for you to correct a live scan.

blacbird said:

I'm quite sure it is. The point I was trying to make is that you don't have to physically print and then physically scan a PDF to get it converted to editable text, which seemed to be implied by earlier posts.

caw

Click to expand...

No, of course not. That would be stupid. The scanning process creates images of the hardcopy; the original scan operator had several options about what the next step was; they may have chosen to have OCR performed and output to .pdf (since they don't necessarily know what word processor a customer might use) or they may have chosen to have the images dumped to a .pdf file.

OCR software is designed to "read" .pdfs that have embedded images. The images embedded in the .pdf are usually .tiff but they may be .png or .jpg.

The software can be pointed to a directory; better packages let you train it to read a particular typeface etc. in a document by providing a way for you to correct a live scan.

But we still don't know if the .pdf is a container for images or a text-based .pdf.

Medievalist said:

If it were me, I'd keyboard the texts in, and edit as I typed.

Click to expand...

This. Between the time it would take to carefully read the output and then correct the errors in your manuscript, and the changes you'll find you want to make because you're a better writer now, it will probably be faster to retype it.

If it were me, I'd even sit down with the printouts of the old material and do a proper read-through and edit, THEN retype it all in.

As a matter of fact, I recently came across some old stuff I need to do exactly this with, assuming that after I do the read-through I still want to have an editable version of the manuscript......which I very well might not. LOL!

Medievalist said:

If it were me, I'd keyboard the texts in, and edit as I typed.

Click to expand...

I've done both OCR+edit and keyboard the text directly. My experience is that it depends some on the cleanliness of the source document and nature of the text, but in general, the time required to complete either task was roughly equal. But I am a hellish good and fast typist, and a good proofreader. For someone not good with the keyboard, the PDF-to-text conversion may offer advantages.

And, elsewhere here, we have this interesting thread about voice-recognition software:

http://absolutewrite.com/forums/showthread.php?t=243624

I haven't tried that, but it's tempting with PDFs. Anyone else done VR to make text out of a PDF document?

caw

hudwrite said:

Hello,

I was hoping one of you had used a conversion program that produced a Word .doc as good as if you had typed the mss. directly into Word. Medeivalist says, yes, but with qualifications.

Click to expand...

Have you determined if the .pdf files are images or native/text .pdf files?

(there is no perfect conversion; 99% accuracy still means that out of every 100 characters/words/ one will be "wrong").

Yeah. I've dealt with OCR with that sort of odds. You're cruising along and everything looks fine until you get to that part where "c" is read as "e" or "g" turns into "q" or a period magically transforms into a comma.

hudwrite said:

The pdf files are scans of original typewritten pages. I have now read reviews of TextBridge and Adobe Acrobat Reader Pro and others, on Amazon, and none delivers a perfect result. I would be happy with 95-97% recognition accuracy with a few artifacts, believe me.

Click to expand...

The distinction Medi is trying to make isn't whether the pages were scanned, because it's obvious that they were scanned; the question is whether they were scanned as images or scanned as text.

If they were scanned as images, you can try to re-scan them as text (that is, an OCR scan). Otherwise, they're uneditable images of text, the same as if you'd taken a photograph of the pages. If you get them OCR scanned, they'll be text that you can copy into Word and start editing.

hudwrite said:

The pdf files are scans of original typewritten pages. I have now read reviews of TextBridge and Adobe Acrobat Reader Pro and others, on Amazon, and none delivers a perfect result. I would be happy with 95-97% recognition accuracy with a few artifacts, believe me.

Click to expand...

Acrobat Pro creates .pdfs. It's of no use to you at all for what you need.

1. Are the .pdfs text .pdfs or image .pdfs?

2. If you can select text and copy it, they are likely text.pdfs. If an entire page is selected, they are image .pdfs.

3. If they are image .pdfs, you need to use OCR software to convert the images of pages to actual text. Then you will need to proof the text against the originals or the images.

I've always wondered why I could highlight text on certain pdfs and not others. This isn't my thread but I have still learned something new.

Although I don't have the volume of words as the OP, I have the same general issue: I use a typewriter for drafting, and scan the drafts as backup. There's a number of us who do exactly this for NaNoWriMo each year, making an already sanity-limited exercise even more complicated.

There are a variety of packages out there: OmniPage always gets high marks from my fellow typists whenever this topic comes up. I haven't tried it myself, though. Both Word and the full version of the Adobe software have an OCR engine built in or installable, although it may take some digging around on searches to find it.

As was stated above, the PDF is really a "wrapper" around the actual image file itself. I don't know if OCR software is smart enough to undo the wrapper. Adobe might be your best bet, or even some image software like Gimp, which can open image-bearing PDF files, and can re-save into a variety of formats, including ones that your OCR program is bound to like. Images saved as TIFF seem to be universally accepted by OCR programs, being a format that scanners also tend to use.

No matter what, though, you're going to fall prey to the GIGO principle: garbage in, garbage out. If your typescript is at all smudged, or if the scans were of low quality, then it's very possible your recognition will be more in the 50%-or-lower range. If the typescript was prepared on an electric machine with a carbon or correctable ribbon, then the impressions should be clear and consistent. If however, you used a manual machine with a standard ink ribbon -- as I do -- your letters may vary in darkness and clarity, and it's that sort of inconsistency that gives OCR programs fits. Even having to correct every tenth or fiftieth character gets tedious very quickly.

In the past, I have retyped my mss. into the computer: the typed page is considered a draft, I draw all over it in pen, and then type in the corrections. I've got one waiting for me that I'm going to try to read into the computer with Dragon Dictate software when I can get some time to sit down and really use it. OCR simply doesn't work for me with the state of my drafts and my terrible typing technique.

hudwrite said:

...are image scans. Some of the software I have been trying, such as Wondershare Pro, claims to have OCR and to be able to create an editable Word doc.from a docx. So far, it hasn't worked for me, in Word 2000, which translates the converted PDF doc into an uneditable Word doc. Maybe places like FedEx Office can execute different types of scans, I don't know.

Click to expand...

The "uneditable word doc" is likely just the image transferred from the PDF without doing the OCR - Word sees it as a picture. There should be an option to do the OCR function, and not just "save as .doc file."

A Word file can include images as well, and you may still be looking at the image in the Word file rather than editable text that an OCR function will output.