27 Sep
2015
27 Sep
'15
6:38 a.m.
-----Original Message-----
From: cctalk [mailto:cctalk-bounces at classiccmp.org] On Behalf Of Johnny
Billquist
Sent: 27 September 2015 13:18
To: cctalk at classiccmp.org
Subject: Re: If you OCR, always archive the bitmaps too - Re: Regarding
Manuals
On 2015-09-27 03:41, Toby Thain wrote:
I did. I found it hard to read as it has OCr'd with mixed typefaces. It is also only
basically non-technical English. Try a couple of pages from any of the VM/370R6 manuals.
I have tried to OCR without the bit maps with little success. These are especially badly
reproduced (originally not as scanned). I can read the text from the BitMaps and know its
right.
One error in the OCR and I can be scratching my hed for ages. I also don't have
problems reading them on a laprop...
On 2015-09-26 5:51 PM, Johnny Billquist wrote:
I suppose it is the eternal argument between preservation and use. I use
these documents every day. I don't care about the pixels, but the content.
Museums and the like are obviously more interested in the preservation.
I get the feeling you didn't actually check the text that I OCRed from a book.
That text is an example what I'm looking for. On 2015-09-26 23:42, Toby Thain wrote:
I've seen plenty of technical books ruined by this technique, which is
why I beg anyone doing this to not divorce the bitmaps from the OCR'd
result.
I suppose some books might be relatively immune, but technical texts
seem to be quite sensitive to poor interpretation by OCR, logically enough. On 2015-09-26 4:28 PM, Johnny Billquist wrote:
> On 2015-09-26 12:16, Johnny Billquist wrote:
>> On 2015-09-25 22:35, Al Kossow wrote:
>>> I have been going back and applying OCR to the ones on bitsavers.
>>> Are there some in particular that you have a problem with?
>>
>> Aha. I wasn't aware of that. I've downloaded copies many years ago
>> that I've been keeping locally. I'll check out the current
>> versions on bitsavers then.
>
> Al, exactly how have they been OCRed? Looking at them, it would
> appear that what you see is still the bitmaps of all the pages, but
> then you have the basic text also available for selection/searching.
>
> My issue with that is that the documents are huge, and the
> experience just scrolling through them is pretty bad.
Imho, though I am sure I am not alone:
Software which "recreates" the typography of a document from OCR
does not produce an acceptable substitute, I've yet to see a book
that wasn't ruined by it.
Just worth mentioning for anyone who might be tempted - For this
reason and others, the bitmaps must NEVER be discarded (Although of
course bitmaps can be archived in a different file if people want to
supply OCR as well.)
Look at the results in the link I posted. I was more than happy with
that result.
I will not prevent people who want pixel preservation from continuing to
have that. But for me, it is a problem. The experience in actually using these
documents are pretty poor. And, as have also been noted, information have
been lost in these scans, as they have not preserved color codings.
Johnny
--
Johnny Billquist || "I'm on a bus
|| on a psychedelic trip
email: bqt at softjar.se || Reading murder books
pdp is alive! || tryin' to stay hip" - B. Idol