MobileRead Forums - View Single Post

tomsem · 10-07-2010, 06:46 PM

I'll admit that I haven't used this feature much. The first couple of PDFs I checked seemed to work fine, but the 3rd has issues. I can get definitions for words that use a heading font, but those in the text body don't show often, or bring up a definition for a word the cursor is not next to. In each case where I checked, I was able to do a text search to find words where definitions didn't appear, and the highlight is in the right place, so the text is in there.

What I suspect is happening is that the document was OCR-ed and that the word boundaries aren't lining up well, and Kindle is asking to look up something that starts with whitespace (e.g. " document" or " " instead of "document"). I can see the area of the screen where the definition would appear redraw, so it is attempting some lookup. Since this is a new feature, and it involves dealing with PDFs in a variety of PDF versions and authoring systems in uncountable combinations, Kindle QE may simply not have looked at enough different PDFs to encounter this case, or they found it, but the bug was deferred. It is also possible that the Adobe PDF library that Kindle uses has a bug or inherent limitation that prevents this from working in some cases.

In any case it would not hurt to report the problem to kindle-feedback@amazon.com. Unfortunately I've found that this email address has some restrictions on file attachments, so you may not be able to attach the 'problem' PDF. In my case, I got the PDF on the web and can give them an URL to download from.

I have sent several PDF bug reports and feature requests since getting my K3. One of the automatic responses I got asserted that PDF viewing was 'experimental'. Of course, this adjective is not used in connection with PDF features in any of the documentation or marketing information, but it does convey a sense of the priorities Amazon has concerning them.

Update: I just checked my 'problem' PDF in Acrobat. I first tried to OCR the text on the assumption that it had been OCRed. But it was unable to do so because the page contained only 'renderable text', which I take to mean it actually was generated directly. Indeed, the PDF producer is 'Mac OS X 10.6.4 Quartz PDFContext' and the PDF version is 1.3 (Acrobat 4.x). It also uses a bunch of embedded fonts, most of which are not on my system.

I ran 'PDF Optimizer' to save as PDF v1.4, but it didn't seem to make any difference.

10-07-2010, 06:46 PM	#4
tomsem Grand Sorcerer Posts: 6,955 Karma: 27060153 Join Date: Apr 2009 Location: USA Device: iPhone 15PM, Kindle Scribe, iPad mini 6, PocketBook InkPad Color 3	I'll admit that I haven't used this feature much. The first couple of PDFs I checked seemed to work fine, but the 3rd has issues. I can get definitions for words that use a heading font, but those in the text body don't show often, or bring up a definition for a word the cursor is not next to. In each case where I checked, I was able to do a text search to find words where definitions didn't appear, and the highlight is in the right place, so the text is in there. What I suspect is happening is that the document was OCR-ed and that the word boundaries aren't lining up well, and Kindle is asking to look up something that starts with whitespace (e.g. " document" or " " instead of "document"). I can see the area of the screen where the definition would appear redraw, so it is attempting some lookup. Since this is a new feature, and it involves dealing with PDFs in a variety of PDF versions and authoring systems in uncountable combinations, Kindle QE may simply not have looked at enough different PDFs to encounter this case, or they found it, but the bug was deferred. It is also possible that the Adobe PDF library that Kindle uses has a bug or inherent limitation that prevents this from working in some cases. In any case it would not hurt to report the problem to kindle-feedback@amazon.com. Unfortunately I've found that this email address has some restrictions on file attachments, so you may not be able to attach the 'problem' PDF. In my case, I got the PDF on the web and can give them an URL to download from. I have sent several PDF bug reports and feature requests since getting my K3. One of the automatic responses I got asserted that PDF viewing was 'experimental'. Of course, this adjective is not used in connection with PDF features in any of the documentation or marketing information, but it does convey a sense of the priorities Amazon has concerning them. Update: I just checked my 'problem' PDF in Acrobat. I first tried to OCR the text on the assumption that it had been OCRed. But it was unable to do so because the page contained only 'renderable text', which I take to mean it actually was generated directly. Indeed, the PDF producer is 'Mac OS X 10.6.4 Quartz PDFContext' and the PDF version is 1.3 (Acrobat 4.x). It also uses a bunch of embedded fonts, most of which are not on my system. I ran 'PDF Optimizer' to save as PDF v1.4, but it didn't seem to make any difference. Last edited by tomsem; 10-07-2010 at 06:58 PM.