Discrepancy between Adobe Extract PDF and the PDF content

Forum|Forum|3 years ago
October 18, 2022
5 replies
956 views

I am a newbie in PDF and I would appreciate having more explanation about how the Adobe SDK API works through the Adobe sample extractPDF using the class ExtractTextInfoFromPDF.java.

I have a source PDF that contains this definition:

7 0 obj
<<
/Type /Font
/Subtype /Type1
/BaseFont /Helvetica
/Encoding /WinAnsiEncoding
>>
endobj

Including the following text sequence:

BT
3 Tr
0.00 Tc
/F3 10.5 Tf
1 0 0 1 302.16 776.64 Tm
<i,/ILLENEUVE > Tj
ET

And when I run the extractPDF sample via the Adobe API to get the TEXT info, I get this:

"Font": {
				"alt_family_name": "* Titlingmes New Roman",
				"embedded": true,
				"encoding": "Identity-H",
				"family_name": "* Titlingmes New Roman",
				"font_type": "CIDFontType0",
				"italic": false,
				"monospaced": false,
				"name": "*Times New Roman-Bold-3921",
				"subset": false,
				"weight": 700
			},
			"HasClip": false,
			"Lang": "fr",
			"Page": 0,
			"Path": "//Document/Sect/P",
			"Text": "VILLENEUVE ",
			"TextSize": 10.0

As you can see, the API has correctly translated "i,/" (3 characters, unless '/' in this sequence has a special meaning) into the "V" character ?

The PDF has been generated using a CANON scanner with OCR/Tagging as the search capabilty is available ont this document, except when searching for "VILLENEUVE".

It must be noted that when opening the PDF for display, the "V" letter is not clearly displayed ...

Can someone explain me the mystery (TEXT correctly extracted using the Adobe ExtractPDF API) ?

Thanks, Eric

Scan documents and OCR

This topic has been closed for replies.

JR Boulay

Community Expert

"Unfortunately, this is now still in the wrong forum. The Extract SDK is a web based service, not connected to the Acrobat SDK. Correct forum: Document Services APIs - Adobe Support Community"

==> Moved to the Document Services APIs - Adobe Support Community discussions

Acrobate du PDF, InDesigner et Photoshopographe

MikelKlink

Participating Frequently

As you can see, the API has correctly translated "i,/" (3 characters, unless '/' in this sequence has a special meaning) into the "V" character ?
The PDF has been generated using a CANON scanner with OCR/Tagging as the search capabilty is available ont this document, except when searching for "VILLENEUVE".
It must be noted that when opening the PDF for display, the "V" letter is not clearly displayed ...
Can someone explain me the mystery (TEXT correctly extracted using the Adobe ExtractPDF API) ?

Apparently the CANON scanner OCR machine could not identify the (unclear) "V" letter correctly but instead recognized an "i", a comma and a slash, and added these characters as text invisibly to the scanned image. As a consequence, "VILLENEUVE" cannot be found. (But have you tried searching for "i,/ILLENEUVE"? That might well work...)

The ExtractPDF API you called, on the other hand, appears to have ignored the text data from the CANON OCR machine but instead applied its own OCR routines to the scanned image of the page. These OCR routines appear to have been more successful and have recognized the "V".

T

Test Screen Name

Legend

Unfortunately, this is now still in the wrong forum. The Extract SDK is a web based service, not connected to the Acrobat SDK. Correct forum: Document Services APIs - Adobe Support Community