ReadIris Pro 11 Mac Asian
Daniel Tschudi
Thanks,
Daniel
Eric Rasmussen
As far as I know, Readiris Pro and IRISPen are the only Macintosh OCR
applications that support Chinese.
The last version of it that I bought was Readiris Pro 9 (the predecessor to
11). It¹s possible there are changes in version 11, but I¹d be surprised if
there are. It didn¹t change in any significant way between 7 and 9, other
than the switch to selling an Asian edition as opposed to an ³add-on." I
don't get the impression they actually have people working for them who are
focused on adding specific features for handling CJK languages. If they do,
they're not doing their jobs! I think they bought the code, but they don't
have the ability to improve it. It is what it is. [I'd love to be proven
wrong.] Here's what I write in the FAQ, an assessment that hasn't changed
since the version 7:
"... Readiris is a Latin-based OCR application. While the Asian edition
functions well within this limitation and provides basic Chinese OCR, to
improve it would need Chinese capabilities built into the core application.
The major drawbacks are the absence of learning for Chinese and the lack of
any sort of effective proofing scheme for Chinese, both essential features
of good Chinese OCR."
In short, if you are going to be doing large projects, I wouldn't recommend
Readiris Pro. You are much better off setting up Windows on your Mac (one
way or another), and then installing one of the true Chinese OCR
applications (after making sure you've got a scanner that works in your
Windows setup -- not all do, especially in VirtualPC -- I've had good luck
with Canon). I think DanChing ("Max Reader") and PenPower (sold in the U.S.
via AsiaZest, which PenPower now owns) are the leaders (especially if you
are handling "traditional" texts), but there are others, like Han Wang.
I think, when I get my Intel-based Mac (next year?), I'll give PenPower a
try. I already own DanChing and have been happy with it, but it's worth $169
to me to see if PenPower is better. The thing about DanChing that drives me
crazy is the tiny font size (can't be adjusted) used in some of the proofing
features. I'm sure it's not a problem for native readers, but I have a tough
time distinguishing the itsy-bitsy candidates.
I get the impression that IRISPen is more useful, but it's not really for
large projects, and it still doesn't have learning or Chinese-specific
proofing tools.
Try a true Chinese OCR application and you'll immediately see what I'm
talking about.
Eric
Daniel Tschudi
Daniel
Eric Rasmussen
> So, which particular Chinese OCR product for Windows would you recommend.
> Thanks very much for your suggestions!
I don¹t have much to add to what I said earlier. I¹ve only used DanChing
4.5. I see there is a new version, 5.0 ($249 direct from NewSoft), so maybe
the candidates (font size) problem I was talking about has been improved.
I chose DanChing because it has an English interface and it also does
Japanese. But I've found I don't really need the English interface, and I've
never actually needed to do Japanese OCR. So I think I would try PenPower
first if I had it to do over again. It's cheaper ($169 at AsiaZest), at
least. But I don't know. I spent some time trying to find a comparison or at
least reviews and screen shots of these products, but didn't have much luck.
Eric
TenThousandThings
>
> > So, which particular Chinese OCR product for Windows would you recommend.
> > Thanks very much for your suggestions!
力新丹青 NewSoft DanChing ("Max Reader") or Presto! DanChing
蒙恬認識王 PenPower Chinese OCR
Thinking about it some more, I guess I'd put it this way:
PRICE: 認識王 is the clear winner. Do you get what you pay for?
FEATURES: 丹青 also supports Japanese OCR. This may account for its
higher price.
INTERFACE: The current versions, 丹青 5.0 and 認識王 3.1, both seem to have
optional English interfaces. Neither has English Help or manuals.
ACCURACY: I'm going to carefully check anything I'm doing, so the
proofing and learning schemes are more important to me than accuracy.
Whether one application is 92% accurate and the other 88% (or
whatever) is not that big of a factor for me. It saves a little time,
but not as much as having an efficient means of correcting the errors
and learning from them.
PROOFING: I know from experience that 丹青 4.5 had a very good proofing
scheme. My earlier criticism was a minor point. No doubt version 5.0
is at least as good.
The real question is about 認識王's proofing scheme. I don't know the
answer to that. There are two basic steps:
1. Identifying errors: Mechanisms for comparing the original scans to
the OCR results. 丹青 does this well with both vertical and horizontal
text.
2. Correcting errors: 丹青 4.5 had room for improvement in this area.
See below.
LEARNING: I don't have enough experience with this to comment on it.
Both applications have the capability. How well it is integrated into
the proofing process is a key factor.
The big problem in correction and learning is dealing with obscure
characters, which tend to occur in people's names or place names. 丹青
was very good at coping with common characters, both in recognizing
them and suggesting corrections for them when the OCR result was in
error.
The time-consuming problems arose when dealing with relatively obscure
Big 5 characters the programmers simply didn't deal with. The OCR was
always wrong, no matter how perfect the scan. I'd place these in two
classes: [1] characters that recur throughout a given text, like in a
recurring name, and [2] characters that don't. The former need to be
learned by the application. I got good at ignoring them and then later
doing find-and-replace, but that's because I'd already read the texts
and knew I could do it efficiently in those cases. The latter are more
troublesome, because learning and/or find-and-replace doesn't really
help with them. So the correction process needs to be good at moving
beyond the application's suggested common-character alternatives to
the more obscure possibilities. In my limited experience, 丹青 4.5
didn't do this terribly well and so that's why I think I will give 認識王
a try when I get an Intel-based Mac.
But it's also likely that it's an intractable problem, and no consumer-
level application can do much better than 丹青 did. Each set of
corrections/learning is unique to each project, because variables like
the typeface, the quailty of the printed text, and the clarity of the
scan all affect the OCR results, not to mention factors like character
variants.
Eric
Daniel Tschudi
--Daniel
Eric Rasmussen
>> INTERFACE: The current versions, 丹青 5.0 and 認識王 3.1, both seem to have
>> optional English interfaces. Neither has English Help or manuals.
Actually, I stumbled across an English manual for 丹青 5.0, here:
http://www.twinbridge.com/detail.aspx?ID=54
Unfortunately, the authors of the manual decided to use all English OCR
examples. This means some of the best features for proofing Chinese text are
not illustrated properly.
If you look on page 16 (PDF 22), you'll see the tiny "Suggested Characters
Window" (#3) that I was complaining about. Other than the size not being
adjustable (in 丹青 4.5 at least), it is a good idea -- see the Preferences
for it on page 20-21 (PDF 26-27).
Another interesting point is that TwinBridge selected NewSoft's 丹青 to sell
online, rather than PenPower's 認識王 -- even though they sell all of
PenPower's other OCR products. I wonder if that represents a quality
assessment? Maybe not -- it could just be because 丹青 does Japanese, and is
thus more versatile.
ER