[bksvol-discuss] Kurzweil Chat Transcription Available
- From: "Linda Adams" <ladams@xxxxxxxxx>
- To: <bksvol-discuss@xxxxxxxxxxxxx>
- Date: Fri, 20 Jul 2007 17:56:13 -0600
Hello, everyone. Attached are the Word and text version of the Kurzweil chat
transcription. This will be long because I thought it best to keep Stephen's
presentation and responses intact.
Linda Adams
July 17, 2007
The Kurzweil 1000
** A New Web Site for Chat Archives: More information regarding this will be
on the Bookshare Discussion list and the Bookshare volunteer list.
** Replacements for Many Books in the Collection: Let Monica Willyard know if
you see any books in the Bookshare collection that need rescanning. She is
keeping a list of them. There is a problem with the book called SEEING VOICES.
There are a lot of carets in it and other types of punctuation that are rather
extraneous. Monica was pleased with the feedback. The replacement scans are
called BSO's.
** A request was made for a book by Anne Rule (not sure of her name's spelling)
called TOO LATE TO SAY GOOD-BYE.
** Pat reminded us that with this new Talkcom software, no extension is needed
for the name that you give your recording of this evening's chat session. To
pause the recording, you can do an alt P. To resume recording, you do the alt
P again.
** Stephen Baum, the Chief Technology Officer of Kurzweil Educational Systems,
made his presentation:
My title now is actually Vice-President, and that's largely because the chief
technology officers at companies are the people who basically make sure the
Internet is running, make sure that e-mail is working, and do all the
technological things that help a company run smoothly. I'm really a developer.
I write code. I am one of the people who has worked for a number of years now
on Kurzweil 1000. I do somewhat more than write code; I manage other people
who write code, and I spend a fair amount of my time for a developer in the
field talking to customers because it is from customers that we get our best
ideas. Since Bookshare.org ahs been around, I have been on both e-mail lists
that Bookshare runs. They've gotten busy enough that I have to admit that I no
longer read every message, but nonetheless, I do pay attention, and I pay
attention largely because this is a community that uses optical character
recognition technology in interesting ways, and those ways have helped me to
drive the direction that our product would be. I have been doing this for a
lot of years. I started with Kurzweil in 1982. The company itself started, I
believe, in 1976. They introduced a reading machine for the blind. Some of
you may have been around long enough to be familiar with that machine. It was
extremely expensive. It recognized characters at the rate of roughly three per
second. It had an interesting-sounding voice--a voice which you could become
accustomed to and learn to hear at a fairly high rate of speed but which at
first didn't really necessarily sound like English. A lot of people's first
impressions of that speech engine were to ask what language it was reading. It
was, I believe, 1976, which means that reading technology for the blind has
been around for 30 years, and of course, in many respects, in all sorts of big
and interesting ways. Now one of those ways is pretty straight-forward; it has
gotten a whole lot less expensive. It has also gotten a whole lot faster and a
lot more powerful. It would be nice to be able to say that that was due to
tremendous and wonderful software designed by people like me, but it really
wasn't. It's due to the fact that computer technology has gotten dramatically
faster, dramatically more common, and dramatically less expensive. We continue
to sort of ride that curve. They're less and less expensive, and the kind of
computer that you would carry around in a cell phone is considerably more
powerful than that original Kurzweil reading machine for the blind, and of
course, a whole lot less expensive. So as the technology has changed, the
products have improved, and the products have changed quite a bit as well.
I wanted to talk briefly in a sort of very theoretical way about what makes
products change because products come on the market, and if they're at all
successful, they start having new releases. You start doing something else.
So what motivates those changes? One thing, of course, is feedback from
customers. Customers who buy the product get in touch with us and say, "The
product is really nice as it is, but I really wish it did this." Or they say,
"This was a really bad idea to implement such and such this way, and why didn't
you do it that way?" One thing that has made Kurzweil Educational Systems, I
think, as successful as it has been is that first off, we have a customer base
that talks a lot. Secondly, we have people who pay attention to that customer
base. We've also done releases frequently enough so that the customers know
we're paying attention. We make improvements often. Those are based on our
customers. Something that doesn't happen to us quite as much but does happen
as well is another type of feedback which is negative feedback. You get people
who look at your product and who choose not to be product customers. For one
reason or another, they don't buy it. A really good company will go out and
particularly solicit opinions from those people to find out exactly what went
wrong as a company, what is wrong with our product, and what's wrong with our
customer service that caused someone not to buy a particular product. So that,
too, becomes something that happens.
Now beyond customers and non-customers, another thing is that technologies
change. We don't work in a vacuum. Even when we made our own computers, made
our own scanners, and wrote our own OCR and our own text to speech, even back
then when we were really pretty self-sufficient, we were not operating in a
vacuum. We were using chips and technologies and different hardware designs
from lots of different people. The technical ecosystem continues to change,
and if anything, it changes at a faster and faster rate, so we end up having to
pay attention to that. It's something that's kind of simple, but here's a very
old example as to a way that things changed. Reading machines used to do just
that. You'd put a piece of paper on the scanner, you would push a button, and
it might have provided navigation aids so that you could go forward and
backward in the page, but basically all you could read was the page that was
currently on the scanner. The thing that made that change was not customers
insisting on something better but rather the simple fact that all of a sudden,
storage technologies like hard discs began to exist and began to become
inexpensive enough that it was possible to store more than one page at a time.
Of course pretty soon, it was not just more than one page but more than one
file. We got into file systems. We got into names, and we did all sorts of
other things. At any rate, it is still the case that the environment changes,
both the hardware environment and also the software around us, and for that
matter, the other sorts of products that are out there.
Another kind of thing that influences change is pretty straight-forward; it is
competitive threats. We do what we do, but we don't do it entirely in a
vacuum. There are other companies that make reading software both in the
commercial market, and in a sense, by the way, even commercial vendors are
competition for us because you of course can buy and use products like ABBYY
FineReader or OmniPage, and we have to pay attention to what those products
offer and do as well. I don't dislike competition. Competition is a great
thing. It means there's somebody out there besides me who is thinking long and
hard about exactly what a product should do, and we borrow or steal from each
other all the time. So it's nice, basically, when you have healthy
competition.
Finally, there are a couple of things that are what I call product imperatives.
One of those is really straight-forward called bugs. I'm sure we're all
familiar with those. You test it, and you have other people test it, and you
think it is useful and works well, and you send it out into the field, and
something truly horrible happens because you didn't test it enough or just
because people work in so many different environments that there are problems.
So when there are bugs, there are things that we need to fix.
There are other things that aren't quite bugs but which are incomplete
features--features that lead onto new possibilities. A good example of one of
those in Version 11 is form recognition. Form recognition is a really cool
idea that I have wanted to have in the product forever. Finally, there were
technologies available that gave me a start in that, and yet I will be the
first to say that it is only a start. It gives you the possibility to fill out
a form, print that form, and send it back to the person who sent it to you to
begin with; however, it is far from ideal, will not work on all forms, and is
missing a number of obvious things. Those obvious things lead me on to what
are likely to become features in a future version 12.
** Standard Features of Kurzweil
I am going to mention a number of the new features, but I will be happy to go
into more detail as people ask questions. One of the things I want you to be
thinking about is, "Who is this particular feature written for?" What did I
have in mind, and what was the problem I was attempting to solve, because that
might give you a sense as to where you might want things to go in the future.
As a quick aside, since I've been paying attention to Bookshare.org, there are
a number of features that quite clearly had Bookshare in mind.
1. The obvious one, of course, is that you can search the Bookshare repository
and even get a list of the new books in the Bookshare repository.
2. You can download books if you wish, and you can decrypt and open them
within Kurzweil 1000.
3. That led us in turn to technologies like DAISY and eventually to the
ability to create and to open DAISY documents.
4. Another thing that it let us do is something that those of you who are
scanning or validating books do all the time which I call production editing.
It is the kind of editing you do when you're not necessarily going to read
every word of every page, but production editing is when you want to clean up a
document as best you can as quickly as you can, featured like ranked spelling
and like poorly recognized page statistics with that in mind. Finally, the
scanning itself, making the scanning as good as possible.
** Version 11
1. Form recognition would be put at the top of the list if you were a
marketing person. For the first time, a blind user has a shot at being able to
take a paper form, concert it to something electronic, fill it out, and print
the filled-out version, all without sighted assistance. The software will
automatically identify various types of fields in a form including check boxes,
text boxes, and character boxes. It will give you a couple of different
mechanisms to explore and fill out the form. It will let you insert a
signature into a field if you wish, and finally, it will let you print out a
filled out copy of the original form. Once you've printed it out, of course
you can put it in the mail and send it to the person who originally sent you
the form. As I've said, this is a kind of incomplete feature in that form
recognition itself is, to put it mildly, an imperfect science. I think all of
you know that OCR itself (optical character recognition) is far from perfect.
Form recognition is even further from perfect. One of the problems is that
nobody gets a college degree in how to create a form. Instead, basically
amateurs go out there and make forms. The forms that you get, even ones that
you would think would be from people who really have a lot of practice, like
people in the government, are often really poorly filled out. So often, it is
a problematic thing. It is just something that becomes difficult. So it's
very hard to figure out how to recognize a form. We're using the same
technology that is in OmniPage, and frankly that technology just isn't quite
good enough. Something I would love to add in the future is a mechanism that
would allow people, sighted or blind, to fix the form recognition as best they
can so that they can fill out a form better than they can today. Right now,
the mistakes that it makes you cannot unmake; you're kind of stuck with them.
2. Scanner Button Support. A feature that is probably not as high up on the
marketing list but which I think is really useful is scanner button support.
For a number of years now, scanners have come with buttons. Blind people, at
least blind people using a Kurzweil, have tended to be told, "Don't touch that
button no matter what you do because if you do, something inaccessible will
come up on your screen, and you may or may not be able to do anything with it.
Now you can potentially press a button and have Kurzweil 1000 come up if it's
not currently active. If it is already active, have it scan another page. You
can also configure it so that a different button will bring up our fax
application, and a third button might bring up our photocopy application. This
is actually a really useful feature because it is often the case that you are
putting pages on the scanner one at a time, and your hands are by the scanner,
not by the keyboard, and so it's really nice to be able just to use a button of
the scanner in order to scan a page.
3. New OCR Recognition. Just about every release we do, we go out and look
for the latest and greatest in OCR technology. As most of you know, we no
longer write our own. In fact, no one at Kurzweil Educational Systems has
written OCR technology at this company. It is something I used to do, but we
don't do it anymore; rather, we buy OCR technology, and what we are using at
the moment is the latest version of ABBYY FineReader, Version 8, and the latest
version of ScanSoft. The company's name is Nuance, and their the ScanSoft OCR
engine is now at Version 15. These are the same as are in the commercial
products, and they are significantly better than versions in the past. The
improvements this time around are fairly significant.
4. Appointment Calendar. We've written another talking application that comes
with the product, and that is an appointment calendar. You can invoke it from
Kurzweil 1000 through the File Launch dialog, or you can run it separately with
a hot key. It is relatively simple, but it does, I hope, much of what one
would want from this sort of application. It is not a replacement for
Microsoft Outlook if that's what you already use, but if you're one of those
people who finds Outlook just unbelievably complicated, you might find this
simpler and relatively helpful. There are things that it doesn't do that would
be nice but which would have delayed us dramatically if we had tried to add
them. An example of something that it doesn't do is that it does not
synchronize with other applications.
5. Scan and Recognize from within Microsoft Word. Something that people have
asked for a long time is the ability to scan and recognize from within
Microsoft Word. People do this on a regular basis to do all their scanning
because you lose all the options or many of the options that are in the
Kurzweil 100 User Interface; however, if you happen to be in Microsoft Word,
and you happen to need to scan a page or two into your current document, it's
very useful to be able to do so, and you can. It will not bring up Kurzweil
1000; it just uses some of the underlying technology in Kurzweil 1000. It will
insert the text into your current cursor position and one page at a time
basically. There are even a few options that you can change from within
Microsoft Word in order to affect the recognition.
6. Notes, Bookmarks, and Marking Last Reading Place in All File Formats in
Kurzweil. One of the most popular features of Kurzweil 1000 over the years has
been our ability to create notes, multi-level bookmarks, and of course, keep
track of where you stop reading. This has only been available in the past,
though, in documents that were saved in our native format, which is KES or, to
a certain extent, in the DAISY III format. If you can open a file, we can in
Kurzweil 1000 also maintain this sort of information for that file regardless
of its format. Here are a couple of words about how that is done: Basically,
there's a database on your PC that we are writing information into based on a
key. The key is the name of the file and the number of characters in that
file, and the information that we save is notes and where those notes are,
bookmarks, what level they where the bookmarks are, and your last reading
position. So when you open a document in Kurzweil 1000, the first thing we do
is look at the original name of the file, look at its size, and see if we have
additional supplemental information in our database that we now want to add to
that file in memory so that we can provide those features to you. The reason
that I go into this sort of detail is to point out what you can't do. You
cannot bookmark and note a text file or a Word file and send it to a friend and
expect those bookmarks and notes to be there. Those are in a database on your
computer only. We're not actually modifying the file because we're not
affecting the format of the file. We don't even necessarily know how to; we're
just writing this kind of information to a database.
7. Open and Play Types of Audiofiles in Kurzweil. Another new type of file
that you can open and play in Kurzweil 1000 is that you can open certain types
of audiofiles and play them. That includes Wave files, MP3 files, WMA files,
and DAISY II audiofiles. Now one thing that is kind of neat is that these are
just like other kinds of files, and you can create notes, and you can keep
track of the last reading position, which is actually something that is kind of
fun with an audiofile. You can play that file faster or slower. You just use
the usual speed up or slow down buttons to do so. It's worth noting that the
natural speed of the files is pegged at 150 words per minute, which in this
case doesn't really mean anything, but it is worth knowing that 150 is that
normal speed. You don't really want to play Beethoven's Ninth at double the
speed, though, as it turns out, you could.
8. Automatic Saving of Document Scanning and Recognition Settings. Suppose
you've been scanning a document, and you want to do as good a job as possible,
so you've used Optimize Scanning, and maybe you've used it more than once on
various different pages, and you've played with a number of settings and come
up with what you think would be optimal scanning and recognition settings.
Then you're interrupted, you leave your computer, you maybe exit from the
program, you come back, you open that document again and you want to scan more
pages, and you remember that you didn't save your scanning recognition
settings, and you have to re-create them. With Version 11, you can have K-1000
automatically create and load settings files that are associated with a
document. The default value for this new setting, which is in the
Configuration Settings Dialog, is disabled. But besides disabled, there are
two other values for it. One value is gives you the ability to save scanning
and recognition settings. You can have it keep track of basically almost
everything except verbosity and configuration settings. When you do that, you
can do such cute tricks as have one document read with a British voice at a
particular speed and volume and another document read in an American English
voice with a different speed. So basically any settings can be applied to any
document, and when you switch the current reading document, it switches those
settings.
9. Writing Audiofiles to CD in Kurzweil. We've been able to create audio
DAISY documents for a few releases now, but then typically, you would want to
write those documents onto a CD so that you could use them in a portable player
that uses a CD. Now you can write those documents directly onto a CE inside of
Kurzweil 1000. There are features to create a list of files that you want to
write onto a CD, to actually do the writing of the CD, to erase the CD, to get
a status of the writing, basically that kind of thing. It is fairly
straight-forward and fairly simple. It is a feature that only works if you
have Windows 2000 or up. It will not work on Windows 98 or ME. That, by the
way, is also true for the new version of ScanSoft. We are actually shipping a
new and an older version of ScanSoft so that people who are installing on
Windows 98 or Windows ME will still get a version of ScanSoft.
10. Oxford Dictionary and Bilingual Dictionaries. We have come for some time
with the American Heritage Dictionary, the Fourth Addition. You can also, if
you like, purchase the Concise Oxford Dictionary from us. With this version,
we also include twelve pairs of bilingual dictionaries. These are very useful
if you want to get a short definition in English for a Spanish document, for
example, or get the French translation for an English word. There is something
like five or six languages, all paired with English.
11. Scan Signatures into a Form Field. Something that I think I've already
mentioned is that you can insert signatures into a form field. In addition to
doing that, you can scan signatures into the system because of course you'd
have to be able to do that. You can save any number of them, and insert them
at your current cursor position in any open document. There is, of course, no
reason to do so unless you are planning to print the document, but it is the
case that if you save it, say, in Microsoft Word format, that signature is
still there in the Word document.
12. Conversion Settings. edit for people who are Bookshare validators. There
are now conversion settings. These settings allow people to influence some of
the nitty-gritty details of how documents are opened and how documents are
saved in formats other than the KES format. They are rather detailed. I
generally don't go into describing them much, but I might for this particular
group because some of them were written with this group in mind.
13. Twenty-Three Relatively Modest Changes. An example of a modest change is
that in ranked spelling, you can do Control plus A to select all of the words,
then copy them. You could then paste them in a new document. Apparently, one
of the Bookshare.org volunteers would like to do that because sometimes they
validate very technical documents that are in a field that is not their
specialty. It might, for example, require someone who knows a lot of medical
terminology, and so the volunteers want to send that kind of list to someone
else so that they can tell the validators whether those particular words are
real words or not. It was a nice idea and is now a feature of the product.
A. A source of information other than our own web site for these kinds of
things is a blog that I wrote some time ago and which I confess I haven't been
updating much, but nonetheless, you can go to that blog and get all sorts of
information as to exactly that feature and various other features. The web
site is www.k1000eng.blogspot.com.
** No Version 12 This Year
One reason why we're not doing so is very simple. Much of the technology that
is in Kurzweil 1000 was written using tools that were really good ideas in
1996. They are not necessarily good ideas now. We are upgrading to brand-new
technology, and the act of doing so is going to require a lot of time and a lot
of effort. That effort at the moment is being plowed into Kurzweil 3000. When
I'm done with that, we'll be doing it for the Kurzweil 1000. When we're done
bringing everything up to brand-new standards of how development should be
done, ideally we will have something that works about as well as it worked
before, which is to say it's not a Version 12; it's just getting us up to where
we need to be as developers. Once we get to that point, then we do the fun
part, which is adding new features and moving on. So Version 12 is next year.
For the first time in the history of Kurzweil 1000, I'm skipping a year, and I
just didn't have much choice.
** Contact Information: Stephen: www.steven@xxxxxxxxx As many of you know,
our web site is usually given as www.kurzweiledu.com. I've always hated that
domain name. There is an alternative name, and it is only four letters long,
standing for Kurzweil Educational Systems, Incorporated.
My phone number is 1-800-894-5374, extension 618.
Our web site is www.kesi.com. You can get bug fixes there. You can get
further information about the product, download tutorials, and there are
various other sorts of resources at that particular site.
** Individual Concerns
1. The old Keith Bell Kurzweil voice is wanted back. Keith Bell was part of a
technology called FlexTalk. It was actually the very first SAPI text-to-speech
software engine available on the market, and we were the first company to use
it. It was originally written for SAPI One, and then it was upgraded to SAPI
Two. It came from AT&T, and the bad news is that AT&T then split itself up, as
you may remember. This was a while ago. One of them was AT&T, and one of them
was Lucent. The code base for all this went to Lucent, and Lucent dropped
support for the product. We continued to provide it. It still worked, but
basically once we got to SAPI Three, Four, and Five, it just wasn't a
technology that worked reliably. It's really kind of a pity when old
technologies just kind of vanish, and they don't give you a way to bring them
back because I agree with you: It was a really pretty good speech engine, and
there is nothing quite like it today.
2. Trouble with too many broken bars, backslashes, and making tables where
there shouldn't be tables. Stephen's Response: I don't know that I could
imagine a setting to fix this, and by the way, I'm sure you could get advice
from other people here. Oddly enough, of course, since I'm a developer, I'm
not necessarily a user, and I may not have the best ideas in terms of how to do
things. You are probably scanning books. Probably what you are getting is a
shadow in the middle of the book. The shadow occurs because scanners have a
relatively modest depth of field. As the surface of the book rises above the
glass of the scanner, it gets very dark very quickly, and it gets fuzzy very
quickly, and you start picking up a lot of miscellaneous garbage and noise that
comes in often as characters. Pressing the book down harder can make a
difference. I might add another thing. If you're scanning a hard-bound book,
it's a good idea to take the slip cover off the book because it, too, has text
on the inside of the slip cover. That text often gets picked up as kind of
miscellaneous stuff. There is also a setting that might help, but it is
sometimes a little bit on the dangerous side. That setting is the ability to
ignore questionable regions. In Version 11, that setting is disabled by
default. In previous releases, I think it was enabled by default. But in
Version 11, you may find that you get somewhat better results if you turn that
on. The one other thing that I might suggest is that switching recognition
engines to ScanSoft may do a little better for you if the problem you are
getting is picking up a lot of miscellaneous backslashes, etc. ScanSoft in
general is somewhat better at recognizing miscellaneous spots of noise as
characters.
3. A person hopes to see improved recognition in new versions more than
calendars, etc. Stephen's Response: Since I don't do character recognition,
about all I can do is provide tools that make it more obvious when a scan
itself is bad, and an example of those is one you just mentioned, which is the
ability to have it tell you when recognition drops below a particular level.
Most times, by the way, when recognition goes bad for a particular page,
sometimes it really is the page because you end up getting dramatically
different print characteristics. That's particularly true for inexpensive
paperbacks. Another thing that happens is this kind of arms race between
people who create books because books get more and more new graphically rich
all the time. New fonts are developed all the time. I realize that no one is
going to scan this because you'll already have a copy of it, but in a few days,
the latest Harry Potter book will be released. The heading is unrecognizable
and comes out as gibberish. That's because the heading is written in some
deliberately spooky, odd-looking font that no OCR product on the market can
begin to recognize. So we don't do OCR. That's one of the reasons that that
kind of thing just doesn't get better, or at least that's one of the reasons
why we don't spend a tremendous amount of time on it. What we spend time on
is, "What can we do before we give the image to the OCR to improve that image
or at least give you diagnostic information as to exactly what might be wrong
with that image. There are things that we have done already in that regard.
There will be more presumably in the future to make it easier, if possible, to
clean things up. Most of the time, though, when a scan is bad, it is because
it was scanned badly. It is not resting flat on the scanner and various other
things like that. I get images from people who want me to diagnose exactly
what is wrong with their product, and quite regularly, it's because the page
isn't really being held very well on the scanner. There is that big, shadowy
gulley in the middle, or quite regularly, I see situations where pages are
folded under, and so they're seeing sideways text from another page and various
other things. Other than us continually shopping around for the very latest in
character recognition, there isn't a great deal that I can do in the
development phase to make recognition itself better except what I have already
pointed out, which is better image processing up front and some potentially
better tools on the back end to fix things up.
4. Someone wanted the flash drive to be integrated into Kurzweil in a later
version. Stephen's Response: In case you are not aware of the feature, in
"file send to" in e-mail, you can send a document or a part of a document to
your e-mail client to be sent to another application. So it's in the file
menu. The menu item is called "send to," and then in that, you would have to
select an e-mail recipient, which I think is the very last option in that list,
and that will allow you to send a document or part of a document either in the
body of an e-mail or as an attachment in a particular file format or even as a
zipped attachment, so that is a fairly rich set of features, although it may be
more keystrokes than you might like. The other thing that is tricky about it,
by the way, as your e-mail client will simply come up, and you still have to do
things like put in the e-mail recipient, typically type whatever else you want,
and hit a send key. We, at least with standard e-mail clients, can't seem to
cause the message actually to be sent. That seems to be sort of a restriction
in the technology that I sort of wish could be gotten around, so you tend to
need to have your screenreader available and come up when the e-mail client
comes up. For a flash drive, one thing you can do is that the flash drive
simply looks like a disc to your operating system, and as a consequence, you
can do a "file save as" and save to that particular disc. There is nothing
particularly tricky about saving to a flash drive. Flash drives, by the way, I
think are wonderful technology, and I think you will see more from us in the
future of a flash drive. One of these days, I would like to do, and may or may
not happen soon--you never know--but I would like to see a version of Kurzweil
1000 made available that runs from a flash drive because that way, you could
haul it from computer to computer and use it as you will without a full
install, and I think that would be kind of handy.
5. A satisfied OpenBook user wanted to know if there were advantages to using
the Kurzweil. Stephen's Response: First, OpenBook uses the same OCR
technology we do. Their product has not been updated in a few years, and so it
is an older version of that technology. But nonetheless, they do have pretty
much the same OCR engine we do. They have many of the same scanning features
but not all of the same scanning features. I believe that they have never
really done Grayscale at this point, and Grayscale does make a difference in
terms of OCR quality. In terms of general features, there are roughly, from
an internal document that describes this, there are at least 85 features that
we have that OpenBook does not have at this point. Whether those features are
important to you is potentially another kind of issue, but they are features
like a DAISY document, the ability to automatically create a cropping window in
your scanning, which by the way can really speed up your scanning if you're
doing small paperbacks. We have certain scanner resolutions that they don't
support. The list is fairly long. Particularly anything to do with production
editing, in general, the answer is, we have it and they don't. Having said
that, I don't personally think in terms of recognition quality itself,
particularly if you don't avail yourself of scanner optimization, that it is
necessarily incredibly better. I'll do something a little unusual. I'll
announce, by the way, that OpenBook, my competitor, does plan to ship a new
release this year, so I believe you will see a version of OpenBook in the fall.
That was the announcement that they made at the blindness conferences.
6. Someone asked where the summarizing feature was. Stephen's Response: If
you have a document open--and it has to have a document open--you'll find that
under the file menu, there is something called Summarize. Hit enter.
A. Bookmarks. If you have bookmarks, it will ask you if you want to summarize
based on your bookmarks. If you say yes, what you will end up with is a new
document that is open that contains a list of all your bookmarks with or
without a page number in the front. If you have leveled bookmarks, they are
organized as a hierarchy with indentation, and they read like an outline. In
fact, this is a very nice way to make, as a study skill, an outline of a book
that you are studying.
B. Keywords. If you choose not to use bookmarks, you will be asked if you
want to specify keywords to influence the summary. You can have a list of
keywords. The keyword will cause us to decide what sentences are more
important than other sentences. You don't need to provide keywords at all, but
they can help in summarizing. Then
C. Without Bookmarks or Keywords. in this case, without bookmarks, when you
hit enter, you will get a new document that generally consists of a bunch of
one-sentence paragraphs one-tenth the size of your original document. It is
not necessarily by any means a perfect summary, but it is a reasonable way to
summarize a book.
D. You can browse, and this is in the Read menu, so you would say read, and
then browse, and Browse will typically read the first sentence of each
paragraph and then move on to the next paragraph. It will do so automatically
and just sort of skip its way down a document, reading just the first sentence,
which tends to be the topic sentence for the paragraph.
E. Finally, I find that people when they're listening do kind of zone out, and
so a third way to go through a document quickly is to crank up the speed as
fast as you can hear it and simply press the right shift button, which will
move you forward by one reading unit, and you will move forward one paragraph
at a time, and that, too, becomes a very good way to actively read quickly
through a document.
7. Someone wanted to know where to find editing tips and how to work with RTF
files easier. Stephen's Response: I've heard that in the past, and I was also
sort of amused to read that I am working on it. I don't know precisely what
the problem is. The work-around of course is that you open the file in
Microsoft Word, make a slight change, and save the document again, and then it
works. The fact that our RTF file can be read by us, by Word, and by Word Pad,
and by just about everything else I can think of that reads RTF but isn't read
by Bookshare makes me wonder what it is that Bookshare is doing when they're
reading RTF. What they have told me in the past--and maybe this is sort of the
issue--is that they have a validator who basically makes sure that there is a
preface that indicates that it is an RTF file and who also goes through and
makes sure that there is a balance between opening and closing brackets. Oddly
enough, I've looked at those documents, and I've seen opening and closing
brackets there all the time, and those look fine, too. I think what I
mentioned about editing tips was simply my blog. It doesn't really talk about
tips for editing in particular. Actually, I believe you guys have some fairly
interesting collections of tips yourselves which I cannot tell you off the top
of my head. There are some web sites that are announced from time to time on
the e-mail lists for Bookshare.org that have a fairly good list of tips. There
are a bunch of features that I said were minor features and consequently that I
didn't mention off the top of my head. I provided an address for those, which
was k1000ng.blogspot.com.
8. Someone wanted to know if there was a demo version of Kurzweil to download.
Stephen's Response: We don't have a demo you can download, at least not very
easily, simply because Kurzweil is a very big program. You can always call us
up at 1-800-894-5374, and you can ask us to send you the demo. The demos are
free. The demo CD when you install it works pretty much like the regular
product except that after an hour, it will stop scanning. You have to exit
from the program and start it again. You can save your document, so you don't
have to worry about that. After 30 days, it will simply stop working, and I
believe it even tries automatically to uninstall itself. It is also worth
pointing out that you can get this demo from our web site by filling out a
particular form at the Kurzweil 1000 page of that web site, and again, it's
www.ki.com. Finally, I think it's worth mentioning that there is a competitive
upgrade price for people who own OpenBook. That price is $595.00. Finally,
having said everything else, in case you missed my earlier announcement, there
is going to be a new OpenBook release in the fall, or at least that's what the
people at Freedom Scientific announced at the blindness conventions. It will
have upgraded OCR engines, and they mentioned a handful of new features.
** Scanner Compatibility with Kurzweil
1. One of my favorite scanners, the PlusTech 3600, has a serious problem in
that the system software that comes with the scanner, basically Microsoft
technology has no idea that it has buttons. They wrote things in very odd,
different ways, and we can't figure out how to get access to the buttons at
all. It is about a $240 scanner. Its main really exciting point is that it
has a book edge. You can actually scan a book one page at a time with the
binding of the book very close to the edge of the scanner. It is not a
panacea. You still need a decent margin on the inside margin for it to read
well, but nonetheless, that's a very nice feature. There is no scanner on the
market like that. The price point is good, and the scanner works reasonably
well. I don't, however, like its software particularly well, and it sometimes
conflicts with other scanners. Nonetheless, any many respects, it's good.
2. Except for really, really low-end scanners, pretty much all scanners
produce decent images. Where scanners differ rather considerably is speed. It
is really hard to find large-platen scanners these days. A lot of the
differentiation in price is in terms of the capacity of the document feeder and
how robust a document feeder you get. Beyond that, what you get for extra
money is a more robust scanner. Inexpensive scanners are primarily designed to
scan an occasional color photograph into a publishing package for somebody at
home. They tend not to have a long lifespan, and when they break, you simply
replace them with another scanner, and given that the price is fairly
inexpensive, it is not necessarily a bad thing to do. At any rate, at our web
site is a link called Scanner Compatibility. In there, you will see a list of
preferred scanners for Kurzweil 1000. Also, you can just review lists of
scanners by all sorts of different features. You can, for example, select
manufacturers, select a scanner, and simply learn more from that. There is
actually a lot in that particular list. It is a rather complicated page, and I
do find that people don't always use it very effectively, but it will tell you
a lot about what scanners we have tested. We don't test everything. When we
test a scanner, we have to buy it just like everybody else, and they get
expensive. For that matter, it takes a lot of time to evaluate a scanner, so
there are scanners that we don't get to.
3. For people who like portability, I really like the Canons. There are a
number of Canon scanners. There are some that are quite inexpensive. The
Canon LID-50 or 90, for example, which has the added benefit of being extremely
light, and as a consequence, carrying it around with your laptop is not a bad
solution if you occasionally have to move your scanner. It is more luggable
than portable. What is neat about it is that you don't even have to plug in
the scanner; it gets its power from the USB.
4. We've recommended Epson scanners for years, largely because their software
works really well. There are higher-end Epsons. There are higher-end HP's
that we recommend. Those are on our web site.
5. Regarding the all-in-one scanner/printer/fax machines, I'm not so sure
about specific problems with those and Kurzweil. It has been a long time since
I've seen them, but there used to be scanners from Microtech that were often
problematic. In particular, they tended, at least in the past, when they were
scanning, your speech would always stutter. So those are scanners I would
recommend avoiding. They are often really cheap, but they often have problems.
Eumax scanners are sometimes good, and sometimes they're not perfect. In
general, the problem with all-in-ones is that the scanner is often the weak
link in what they're providing, and of course they're very competitively
priced. They are often very inexpensive. We have in general found the Epson
all-in-ones to work quite nicely. The higher-end HP ones work pretty well.
The lower-end HP's tend to be just too slow to use. The main thing that I
think you really want to pay attention to is the scanner because scanning speed
in the end is going to be important to you.
6. Someone wanted to know about new Kurzweil voices in the future. Stephen's
Response: The way Kurzweil works is that we want to support every voice that
you have on your computer if it is a publicly accessible voice. JAWS' version
of Eloquence is one of the few things that is sort of out on the market that
tends to be private. Basically Freedom Scientific gets a discount for selling
a version that only works with their particular application. That is sometimes
true for a few other products as well. Speech engines that you get with other
products or that you buy from web sites like Nextup should just work with
Kurzweil. All you have to do in general is switch to them and use them. We
really are about the only thing out there that supports SSIL as well as SAPI,
DAPI, and every other flavor of speech devices, so we basically try to cover as
many speech engines as we can. To the extent that there are interesting
engines available, I'm not necessarily sure that we would add more. They all
cost us money when we add additional ones. I have been considering simply
selling more at reasonable prices so that you could go to our web site and
order any of the RealSpeak solo voices from us, which support a lot of
different languages with very high-quality voices. We haven't done that to
date, largely because we're not a company that sells inexpensive things. There
is actually a skill to that, and we're not necessarily very good at it. But it
is possible that we would do that in the future. Something else that may be of
interest is that the person or the company that makes Eloquence also is making
a new voice which they think sounds better and which they think is as
responsive as Eloquence. That might be a voice that will interest us in the
future, but to be honest, we have heard that before from other vendors, and
we're suspicious about it. A lot of times, when people think a voice sounds
better, it's because to an untutored ear, it sounds more natural, but often a
more natural-sounding voice doesn't sound as good at high speed or is just not
as understandable at high speed, and we are careful about that.
7. Someone likes being able to scan PDF documents in Kurzweil from a web site.
Stephen's Response: I wanted to talk briefly about that because this is where
we sort of get into nitty-gritty of certain features. As I may have mentioned,
there are conversion settings in Version 11. A conversion setting document
starts with a field labeled Action. Its possibilities are opening documents
and saving documents. Then if you tab, it basically has a list of formats
regarding either opening or saving documents. Under the Opening Documents one,
it has Text, Braille, RTF, PDF, and Other. If I tab after that, it has just
one setting called Emphasis, and the values are Recognition of Images or
Extraction of Text. PDF files are very complicated beasts. They can contain
pictures of text without actually having the recognized text. They could have
both a picture of the text and the recognized text, and even worse, they could
have some recognized text, but not all of the text that they're showing as an
image in a particular page. The default here for us is Recognition of Images,
and what that means is that when you open a PDF file in Kurzweil 1000, we will
extract the text that is recognized and is already present in the PDF page, we
will then recognize the images, and to the extent that we can, we will match up
the text to the images. But if we have text that doesn't match, we will simply
throw it away, and we will tend to trust the images. The other possibility,
Extraction of Text, will give you better recognition results if all the text is
there. The bad news is that we don't know if all the text is there, and so in
some circumstances, you can end up losing text. While I'm in this dialog, I
want to describe one other thing, which is for opening documents. Let's say
the format is RFT. There, the setting is Split Long Pages Enabled or Disabled.
I know for Bookshare volunteers, pagination is a very big issue, and so if you
get documents that are properly paginated in RTF and you are validating them,
it is just possible that you want to disable this setting so that we will not
split what we think of as being a really long page into more than one page. So
you now have control over that.
8. What are the pros and cons of Kurzweil over OmniPage? Stephen's Response:
First, there is an obvious disadvantage. OmniPage Pro is a whole lot less
expensive. A Kurzweil product, because it is made for a blind audience, tends
to be on the pricy side. What it gives you is a whole lot of features that are
designed with a blind user in mind. We have mentioned some of those features.
Basically the simplest one, and yet it is a powerful one, is simply the ability
to read and edit while you are scanning. So you can be scanning and
recognizing one document while you are reading another, or you can be scanning
and recognizing the same document but on a different page than you are reading
and editing. There are lots and lots of other features. By the way, I will
also point out that OmniPage is pretty feature-rich, too. The user interface
is optimized for someone who is blind. You tend to use it without a
screenreader. It does its own speaking. In many respects, it tends to be
easier to use. Of course in your case, if you are already quite accustomed to
OmniPage, that may not be a big deal for you. Another thing, of course, is
that rather than just having the OmniPage engine, you also get the ABBYY
FineReader recognition engine, so it's essentially like having two products in
one in that regard, and there are differences. Every once in a while, it's a
good idea to switch from one engine to another and see if things improve. I
will mention one other thing, which is that we have we have toll-free customer
support. The people who answer the phone are blind themselves and are users of
Kurzweil 1000. They know the product and the product base well, they know
their customers well, and they don't mind spending a fair amount of time with
people if they need that time. That's something that you tend not to get from
commercial products.
9. What might be upcoming down the road? Stephen's Response: We won't be
doing a Version 12 this year, and I don't at the moment have a feature list. I
have already mentioned, though, a few things that I want to work on.
A. I would like to make forms work better than they do. Primarily what I mean
by that is that I would like low-vision users and blind users to be able to
analyze themselves the results of form recognition and be able to do things
like delete a field, insert a field, modify the name or label of a field, etc.,
so that they can do a somewhat better job of filling out that form.
B. Kurzweil on Flash Drive. Another things that I think would be a powerful
new delivery of an existing product would be to deliver the product on a USB
device: flash disc, thumb drive, whatever you want to call it. In that kind
of delivery mechanism, it would be locked to that flash drive; i.e., you need
the flash drive, and you need to plug it in in order to use the product, but
you wouldn't need to install on a computer. The consequence of that would be
that you could use it on pretty much any computer. That is particularly
powerful for the non-scanning features. For the scanning features, you could
do that, too, but I tend to think that that is likely to make you a little more
grounded to a particular sort of thing.
C. New OCR Engines. As new OCR engines become available, we analyze them. We
potentially add them, and we potentially add them. I certainly would expect
that with Version 12, you will see changes in terms of what we provide for OCR.
D. Beyond that, I'm interested in your ideas as to what we ought to do in the
future. I do collect those ideas and have a tremendous number of e-mails in a
category that I call Wish Lists and just a whole lot of things that I pay
attention to, and I try to do something with them.
10. When we receive books in RTF format from our sighted volunteers, there
aren't enough page breaks. How does Kurzweil handle the section breaks from
ABBYY FineReader? Stephen's Response: We put in code to treat section breaks
as though they were page breaks if there were no page breaks. It's a little
bit complicated, to be honest, because if there are both section breaks and
page breaks, it is a little hard to know what to do. It is possible that there
is something that we should fix there.
10. The issue of not reading correctly was mentioned again. Someone wanted to
know if Kurzweil would do Braille translation in a future version. Stephen's
Response: It's a little vague hearing "not reading correctly." That can be
OCR or other things. One thing that I'm aware of that has been reported
several times is circumstances where the speech just runs off, and you can't
stop it. You hit the control key or hit whatever you want, and it doesn't stop
until it reaches the end of a page at worst or somewhere in the middle of a
page at best. I think that it is only in SAPI Five. To be honest, I don't
think we understand what the problem is. So that's one that I'm kind of aware
of that may be what you are referring to. We were the first scanning and
recognizing product to add both forward and backward Braille translation.
11. Someone wanted to know how to get the Kate voice to stop interrupting
herself if a person is in two-page mode and the confidence level is below a
certain level. The person can't hear what that level is before Kate says,
"Recognition complete" for the second page. Stephen's Response: I would
suggest something that I do when I'm scanning. Use the verbosity settings. In
the settings dialog, you'll find a setting called Verbosity. Verbosity gives
you a long list of what are called Events. There's an event called Recognition
Progress, and there's another one called Recognition Complete. I would change
those to be either disabled or to use a chime. Using a chime is much less
obtrusive and will tend not to stomp on other speech.
12. Someone wanted to know how to keep Kurzweil from getting two copies of a
Bookshare book from the bookshelf. First off, it's news to me. I really
haven't heard that. There were circumstances in the past where we would list a
book twice, and I forget even why that was, but it was fixed quite a while ago.
I am reasonably sure that we are only downloading one copy--at least I think
we are. If that is happening, it is something that I would like to be able to
duplicate. Something that is a little bit like that is that occasionally we
have multiple entries in the Project Gutenberg, and I in fact was just told
this the other day, and it is typically because they have more than one format
available. In their case, those formats are listed in separate entries in the
table.
13. Someone asked about Text Adventure games in Kurzweil. Stephen's Response:
I'm not sure what you could do in Kurzweil with a Text Adventure game, but I
will point out that there is a fair number of Text Adventure games on the
Kurzweil CD in the Extras folder. The Extras folder has just sort of a mess of
strange freeware and shareware that we thought might be of interest to people,
and there are some odd little things in there.
14. Someone asked about how the competitive upgrade works. Stephen's
Response: I believe that the competitive upgrade at this point only applies to
OpenBook, but if you have OpenBook at any version, essentially you provide some
proof of ownership. I believe they accept your serial number. They may accept
a front page of your manual. I'm not entirely sure. Having done that, the
price, instead of the usual $995.00 is $595.00.
15. Someone wanted to find out how to keep from getting two page number
announcements when saving a text file into RTF in Kurzweil. Stephen's
Response: Here's my guess. One of the page announcements that you're getting
is the page header that happens to be at the top of pages. The other one is a
page break that is sort of randomly inserted, and probably the difficulty is
that your original document doesn't actually have page breaks at all, and we
are inserting page breaks after a certain number of lines because we don't want
the pages to get too long. It's laborious, but you typically end up wanting to
insert page breaks into those documents. That, by the way, is one reason that
Bookshare.org tends not to accept text files anymore because all too often,
they don't have page breaks at all. I believe that Word, like most word
processors, is oriented toward what something is going to look like when you
print it, which is a function of the print size and how much text you have on a
page. Then indeed, Word is going to insert page breaks. They are simply an
indication of if you were going to print this here, this is where a page break
would be. I believe that you can pretty safely ignore them as they are not
likely to be written into the output file.
The meeting ended.
Other related posts:
