The Braille Monitor February, 2001 Edition
by David Andrews
From the Editor: The following is another in Dave Andrews's series of comparative examinations of access technology. For some years Dave directed the International Braille and Technology Center at the National Center for the Blind. He has moved on from that position, but he continues his willingness to share with the rest of us what he has learned from his knowledgeable comparisons of competing products. This time it is reading systems. This is what he says:
As computers become an ever-increasing part of our lives, we can divide what we do with them into a few categories. As blind people we use computers to read or write text, to communicate with others using e-mail and other methods, to surf the World Wide Web, and to keep track of things. This article will concentrate on one of these areas, the reading of text--more specifically, the reading of printed paper documents such as books, magazines, bills, and letters. In particular we will compare and review the Openbook produced by Arkenstone--a division of Freedom Scientific, Inc. and the L&H Kurzweil 1000 produced by Kurzweil Educational Systems Group--a division of Lernout and Hauspie Speech Systems, N.V.
Unless you follow the machinations of the access-technology field, these names may be unfamiliar to you. The L&H Kurzweil 1000, which we will call the K1000, was formerly known as the Omni 1000. It was developed by Kurzweil Educational Systems, Inc., a company founded by Dr. Raymond Kurzweil, who is familiar to most of us who are blind. Dr. Kurzweil sold this company to Lernout and Hauspie Speech Systems in the fall of 1998. L&H is a Belgian company involved in speech and language technologies worldwide.
Open Book, also known as "An Open Book" and "An Open Book Ruby Edition," was developed by Arkenstone, Inc. Earlier this year the holdings of Arkenstone were acquired by Freedom Scientific, a new access-technology conglomerate which also includes Henter-Joyce, Inc., and Blazie Engineering. Now that you know who's who,we can begin.
For this review we looked at version 4.9 of Open Book and version 5.0.02 of K1000. At the time of this writing, September, 2000, Freedom Scientific is selling version 4.02 of An Open Book Ruby Edition. However, in mid-August it released a public beta of version 5.0, numbered 4.9. While it is not normally our practice to review beta copies of software, we decided to do so in this case for a couple of reasons. First, this is a public beta, which means it will receive wide distribution aimed at uncovering major bugs or defects. It is unlikely that we will see additional features or major changes between versions 4.9 and 5.0. Except for major fixes, what we now see is what we are likely to get. Freedom Scientific expects to release version 5.0 in late September or early October of 2000. [It has, in fact, now been released and is available.] If we were to review version 4.02, it would be outdated by the time you read this. The K1000 folks are also working on an upgrade but don't expect to have something out until around the first of the year. It should also be noted that there are other adaptive reading systems on the market; however, Open Book and K1000 dominate the market in the United States, hence our concentration on them.
When one works with these two systems, it soon becomes obvious that there are more similarities than differences between the two because both programs are fairly mature; that is, they both have been out for a while and have had time to develop large, similar-feature sets. The PC-based reading software market is quite competitive; and, if one major player comes up with a good feature, the other one is likely to implement something similar in its next release. Yes, there are differences in philosophy, approach, and implementation; and we will discuss them. But the similarities are more numerous.
Both programs are self-voicing applications. This means that you do not need a screen-review program such as JAWS for Windows or Window-Eyes to use them. You also do not need a dedicated hardware-based speech synthesizer. While you will need a sound card, such as a Creative Labs Sound Blaster or equivalent, each package comes with one or more software-based speech synthesizers. Open Book comes with IBM's ViaVoice Outloud 4.0, and K1000 comes with RealSpeak and TTS3000 from Lernout and Hauspie. ViaVoice Outloud is based on the same technology as Eloquence, which comes with JAWS for Windows. However, its performance seems somewhat sluggish to me; that is, it doesn't respond to commands as quickly as the JFW Eloquence does. Further, with a top speed of 350 words per minute, it isn't fast enough for some power users. But at the price it isn't bad.
RealSpeak is a new speech engine developed by Lernout and Hauspie--Kurzweil Educational Systems Group's parent company--which became available with version 5.0 of K1000. Personally, I have very mixed feelings about this speech engine. The technology apparently uses digitized human speech sounds to create the synthesized speech. On the plus side its inflection and flow are quite human-like and an improvement over any synthesized speech I have heard to date. On the negative side its performance or response to keystrokes is very sluggish. It takes a good deal of system resources, and overall I find individual letters and numbers difficult to understand. Is it saying "tab" or "tap"? Or is it saying "53" or "63"? Overall RealSpeak sounded muffled; it didn't have good high-frequency response, which in large part accounts for the difficulty in understanding some things.
It also has some artifacts typical of digital synthesized speech. The system requirements are a 300 MHz processor and 64 megabytes of RAM above the basic requirements for K1000 itself. People with inadequate memory or a slow processor have reported excessive speech stuttering, an observation that my experience bears out. I did most of my testing with a Dell Pentium 300 MHz computer with 192 megs of RAM. While my memory was more than adequate, my processor speed was just adequate, and I found using RealSpeak an unsatisfying experience. Its natural inflection is a technological achievement worth noting, one which we will benefit from in the future, I am sure. However, for now most of us will probably prefer another speech engine.
Self-Voicing Application Considerations
While it can be convenient to have self-voicing applications, especially if you do not have a screen reader, using them in conjunction with a screen reader can also get complicated. The Open Book install program does a good job of detecting whether or not you have a screen reader running. It will speak or not speak, whichever is necessary. The install program will also install appropriate files to put your screen reader to sleep while using Open Book as long as you use JFW or Window-Eyes. Open Book has an informative file on screen-reader compatibility in its Help folder. It is possible to get Open Book to do all its own speaking or to use your screen reader to do this. It is also possible to get different speech engines to share the same sound card and more, although it can get a bit tricky. The file tries to explain it all.
K1000 can be used with or without a screen reader. In its default mode it will be self-voicing and will take over the numeric keypad. You can get it to be quiet and relinquish control of the keypad, or you can put your screen reader to sleep. However, you will have to do this configuring yourself; K1000 does not automatically install any sleep-mode files as does Open Book.
An alternative approach, which avoids potential conflicts, involves launching either program with a hot-key command, a feature both systems offer. You first unload your screen reader from memory, then execute the hot key command. You have to relaunch your screen reader when you exit Open Book or K1000, a task you can also perform with a hot key. Nevertheless, both Open Book and K1000 are designed so that they can be used by computer neophytes. The potential conflicts between self-voicing applications and screen readers are numerous and complicated. In addition to these programs, other self-voicing applications include PW Web Speak and IBM's Home Page Reader. Everybody, including application developers and screen-reader developers, needs to cooperate and work out a system that is automatic and seamless for the end user so that all these programs can work together.
Several common features shared by both these programs include a speech-friendly and simple install process, automatic scanner detection, multiple OCR (optical character recognition) engines; two major control systems, including one using the 17-key numeric keypad and a second using standard Windows commands; support of multiple speech engines; file management and library systems; multiple scanning modes including scan and read, batch scanning, image scanning, repeated scanning, support of automatic document feeders, etc.; editing of recognized text; importing and exporting of multiple file formats; bookmarks; dictionaries; thesaurus; OCR error correction lists; spell checking; pronunciation dictionaries; identification of paper currency; deleting, inserting, and renumbering of scanned pages; visual display settings designed to assist low-vision users; ability to scan from within another application; find-and-find and replace? text; the launching of other programs from within Open Book or K1000; and more.
As you can see, this is quite a long laundry list of common features and includes virtually everything that a scanning/reading program needs. However, there are some differences between the two programs, even with the commonly held features. I will discuss some of these, including those that are noteworthy. I will also talk about features unique to each program.
Open Book and K1000 will both perform an automatic install, that is, using a default set of choices which are right for most people. Also either will allow you to do a custom install, making installation choices yourself. Of the two programs K1000 has the simpler, more automatic install even in the custom mode. It presents you with an option, and, if you want to change it, you hit the Enter Key. If you want to accept the choice, you can either hit the Escape Key or do nothing, and that choice will then be accepted. Open Book's custom-installation process uses a more standard Windows approach, allowing you to make choices from controls such as dialog boxes and check boxes.
In the past most people purchased Hewlett-Packard scanners for use with programs like Open Book and K1000. However, some of the newer HP scanners do not work well with Open Book or K1000. Like most things with computers, a standard which isn't completely standard has emerged for communicating between scanners and the software applications they use. This standard, called "TWAIN," stands for a "technology without an interesting name," an acronym I always get a kick out of. There are many good TWAIN scanners on the market. Some work with Open Book and K1000, and some don't. I conducted my tests with an Epson Perfection 1200U, a TWAIN scanner using a USB connection. You should check with Freedom Scientific or Kurzweil to see if your scanner or proposed scanner is compatible.
Until recently programs like Open Book or K1000 came with one optical character recognition engine, the software that takes an image from a scanner and turns it into actual text. In version 4.5 Kurzweil Educational Systems Group upped the ante, adding a second engine, and now Arkenstone has upped it again, adding a third. Version 4.9, soon to be 5.0, of Open Book ships with CAERE MTX, FineReader, and Recognita engines while K1000 includes RTK and FineReader OCR engines. Kurzweil tells us that it added a second recognition engine because in some instances the Fine engine offers improved accuracy at the cost of longer processing times. I found this true during my tests in at least one instance. Using RTK, I got quite a good scan of a hardcover book. There were a few mistakes, though. When I switched to the Fine engine, I got perfect text, although the processing time was approximately 10 seconds a page longer.
Now Open Book includes a third engine, Recognita, but I am not sure why. Recognita, while inexpensive, is not considered one of the better performers on the market, such as the CAERE, RTK, or Fine engines. In my limited tests I was never able to get better results with Recognita, so besides its marketing value it seems to offer little else at this time. Jim Fruchterman, President of Benetech--the nonprofit company doing development on Open Book for Freedom Scientific, told me that they had hoped for better performance from Recognita. He also said that it offers support for additional languages and that an upcoming release is supposed to offer better OCR performance. Let's hope so.
Open Book and K1000 have two major user interfaces. One uses the 17-key numeric keypad found on the right side of a standard 101- or 103-key PC keyboard. These user-interface systems were originally developed for PC-based stand-alone reading systems, which both companies sold at one time. Both programs now also include a more standard Windows-style user interface employing a menu bar and pull-down menus. These systems also have a variety of hot keys for common commands or operations in addition to the use of standard Windows commands for cut, copy, paste, selecting text, and the like. Additionally the programs supplement the standard Windows menuing system with a variety of function-key commands to perform a number of common operations. Some of these commands use shift, control, or alt function key modifiers for other tasks. K1000 in particular has a rich, if sometimes overwhelming, set of commands for navigating and editing documents, managing files and folders, etc.
Open Book and K1000 both state that their keypad interfaces do not support all commands and functions present in the programs, although they do allow you to execute most of them. They are present for the sake of compatibility and consistency with past versions; however, most users are probably better off using the Windows-style menus and commands. I found myself using keypad commands for a few things, but the Windows interface is richer and more familiar to experienced users. The Open Book keypad command structure, as I have noted in the past, can be a little confusing with all the going up and down menus and in and out of choices. Likewise, I find the three-layer approach that K1000 uses for its keypad commands--layers for reading, settings, and file management--to be a little overwhelming with the number of keys that must be remembered or found.
The keypad interface for the Open Book was extensively described in a review published in the January, 1995, Braille Monitor, and the K1000 interface was extensively described in a January, 1998, Braille Monitor article. Much of the user-interface information in these articles is still valid, so I won't go into a lot of detail here. These articles also provide good information on the purpose and components in a PC-based reading system and discuss why a person would want to choose Open Book or K1000 over an off-the-shelf OCR package such as OmniPage Pro or Text Bridge. Monitor reprints are available from the NFB's Materials Center for $2. You can also find them on the NFB's Web site, <http://www.nfb.org>, or on NFB NET, <http://www.nfbnet.org>.
Speech Engines and Synthesizers
Open Book supports a variety of SAPI and SSIL speech synthesizers. K1000 will automatically detect any speech engine that is SAPI-version-4 compliant. What does all that mean, you might ask? SAPI stands for the Microsoft Speech Application Programming Interface and is a standard for speech engines and applications to communicate. It provides a common method for a screen reader or a self-voicing application to use a speech engine with your sound card. Both IBM ViaVoice Outloud and RealSpeak are SAPI-compliant speech engines. K1000 also ships with Lernout and Hauspie TTS3000 and FlexTalk from AT&T. However, FlexTalk is a SAPI-version-3 application and doesn't work well with SAPI version 4 and will not install if you have a SAPI-version-4 engine already present.
Both programs also support the SSIL standard developed by Arkenstone. This stands for the Speech Synthesizer Interface Library, a de facto standard used by screen review programs and self-voicing applications to communicate with hardware synthesizers such as the DEC-Talk, Double Talk, and the like. Because they developed it, Open Book supports the full range of synthesizers for which SSIL drivers exist. On the other hand, K1000 supports only a few of the most common SSIL synthesizers. Unfortunately this did not include my Audapter from Personal Data Systems. Kurzweil has mentioned the possibility of additional SSIL support in the past, and I for one would like to see it. They did provide me with instructions for interfacing my Audapter with their program; however, this process isn't for the faint of heart. You would need either to be an experienced computer user or to have the assistance of one.
Open Book allows you to change speech engines or synthesizers on the fly, but you can choose only from the ones you have previously installed. It also allows you to set different voices for reading text and menus, as well as choosing a voice for emphasized text. While these three voices can be different, they must all be from the same synthesizer or speech engine. While K1000 has only two voice choices for reading voice and system voice, each of these can use a different speech engine or synthesizer.
The two programs each have their own proprietary file format, with an .ARK extension in Open Book's case, and .KES in K1000's. These formats allow you to preserve text and bookmarks. Both programs import and export files in a variety of formats. Open Book imports its own files, text files, RTF files, HTML documents, and Word and WordPerfect files. K1000 imports its own files as well as Word, WordPerfect, RTF, HTML, and Grade II Braille files--which it back-translates as it imports them. It bases conversion on the file extension, and, if it is unable to determine the proper format, it presents you with a list to choose from. This list, as well as the export list, is based on any available file converters already on your system. It exports in these formats as well as in Microsoft Excel. It will also search your system for any file converters you may already have and in my case came up with an additional 150 file formats I could save as--many of which I had never heard of.
K1000's support of Grade II Braille is a useful addition. It has a built-in version of NFBTRANS inside itself. As you may remember, the National Federation of the Blind put NFBTRANS into the public domain in 1992, and it has gone through improvements ever since. K1000 will import and back-translate Grade II files automatically or output to a Grade II file or printer. It is easy to print directly to your Braille embosser; all you have to do is set up a generic text printer under Windows 95 or 98. It took me about two minutes to do so--including finding my Windows 98 CD-ROM. Formatting of documents is not perfect; NFBTRANS has little formatting information to base its decisions on. However, I found its work adequate for quick and dirty personal use. Open Book, like K1000, has the ability to launch another application from within itself, passing text to this application. You can use this function to launch a Braille translation program such as Mega Dots or the Duxbury Braille Translator. In fact, during its install process Open Book can search for launchable applications on your system. On mine it found Microsoft Word, Word Pad, and Turbo Braille--a DOS-based Braille translation program I sometimes use.
Open Book supports the Library System it has used for file storage in several previous versions of the program through a dialog box off the File Menu. K1000 has a complete file-management system, which is a variation of the file tree used in Windows Explorer and other programs. It uses a grid of drives, folders, and files, through which you navigate with the arrow keys, but dispenses with the "open" and "closed" states that Explorer uses. It took me a little while to get used to but worked well once I got the hang of it.
Open Book has a wider variety of scanning modes such as Scan and Recognize, Batch Scanning, Batch in Background, Express Batch, Express Batch in Background, Preview Scan, Orientation Only, and more. Most people will use Scan and Recognize or one of the Image/Batch modes if they have an automatic document feeder. The Preview Scan is unique to Open Book and scans just the top of a page, facilitating identification of a document.
K1000 simplifies the process. You can Scan and Recognize text, repeatedly scan based on a period of time, and also scan an image and later recognize it. Both programs will work with a number of graphics-file formats, including files from a computer-based faxing system. We didn't test this capability, although we did test the recognition of a regular paper-based fax--read on for the scanning and recognition test results.
Either program allows for the editing of recognized or imported text, either from within itself or by calling another application. The use of standard Windows editing commands makes this a straightforward task for most experienced users. The presence of spell-checking, a dictionary, and a thesaurus in both products makes these programs competent editors. They do not have the full feature set of WordPerfect or Microsoft Word but will do fine for many basic editing and correction tasks.
Both programs allow you to set bookmarks within text. However, for these bookmarks to be retained, you must save in each program's proprietary file format. K1000 has used bookmarks extensively in its manual file, and they provide a good way to navigate. Further, the program has a function that will extract a summary of a document based on bookmarks you set or on a list of keywords you assign. With some planning and forethought this could be a useful outlining and study tool.
Identification of Money
If you are like me, occasionally you will be in a store and a clerk will thrust a stack of unidentified bills into your hand. You may not have the time to have them identified so you can fold them properly. Well, Open Book and K1000 can come to the rescue. Both programs contain utilities that will identify paper money. Open Book contains Buck Scan, developed by Noel Runyan of Personal Data Systems--a longtime Arkenstone dealer and systems integrator; and K1000 contains Money Talks, a utility written by Stephen Baum and released into the public domain. Some users have reported having problems identifying money. If a bill is excessively worn or wrinkled, identification may be difficult or impossible. Open Book also advises you to hold the bill on the scanner glass by hand, not to put it on the glass and shut the lid as you would do for a regular full-sized document. Because of its small size a bill may move when you shut the lid without your knowing it. Buck Scan has the advantage of making repeated scans easy, and it works quickly. The K1000 utility works more slowly than BuckScan and must be rerun each time you want to identify a bill.
One area that Open Book in particular has concentrated on is visual appearance, especially for low-vision users. K1000 has done work in this area too and has additional plans for the future. Open Book in Version 5.0 has added a Low-Vision menu. It has settings for background, foreground, and cursor color; font style and size; and spacing between characters, words, lines, and sentences. The range of choices for these settings seems quite broad, giving many low-vision users the flexibility they need to customize the display for their use. Open Book will also print text with the display settings you choose. It has an "Exact View," which is an image of the paper document, not a representation that has been passed through the optical-character-recognition process. You can change the size of this view, which may be appealing or useful to some users.
With a couple of exceptions K1000 has not concentrated as heavily on the needs of the low-vision user. First, all K1000 dialog boxes are created with a larger font than that used by Open Book and other Windows programs. This means that K1000 is able to get fewer controls in each dialog box, but their larger size appeals to many low-vision users. Second, K1000 does have settings to control the color of text, background, and highlighting; and you can increase the size of recognized text up to eight times. You can have K1000 display text or print text using the fonts and point sizes it finds in the original document. You can also have it use one font and increase or decrease the size as Open Book does.
Finally, L&H Kurzweil Educational Systems Group has announced a product (which should ship in late September) called MagniReader. This $349 product, which will be available free to all K1000 users under warranty, will scan a printed page and put its image on your computer screen. MagniReader will have a user interface that is designed entirely for the low-vision user. The size of its menu buttons and the number of buttons presented on a screen are readily configurable. You will be able to use it to scan pages and display them at varying levels of magnification in black and white or in color. OCR can be performed, and the text can be read aloud. MagniReader will come with the RealSpeak and TTS3000 speech engines. Display options will include smooth-scrolling Marquis display, magnification of individual words as they are spoken, and presentation at varying magnifications of a bitmap of the page or of the recognized text. Files cannot be opened or saved, and text cannot be edited. You should think of this program more as a CCTV on steroids than as a typical computer-based scanning and reading product.
Stephen Baum, K1000's principal developer, feels strongly that extensions to the basic reading product, such as extensive low-vision support, should be offered as add-ons. They can then be used or ignored at the end user's discretion. This is why MagniReader, which was announced at our 2000 convention in Atlanta, will be a separate product which can be integrated into K1000. Baum is considering other additions to the K1000 product, but final decisions have not yet been made. One addition that he feels confident we will see is the ability to export text to other access devices such as a Braille 'n Speak, Road Runner, or Braille Note.
Refreshable Braille Support
Support of refreshable Braille displays, or the lack thereof, is a feature that both programs have in common. Neither one directly supports refreshable Braille displays. I could get Braille output on my Power Braille at work by putting JFW in sleep mode, not unloading it completely, and running either Open Book or K1000. Direct Braille support would certainly be a nice addition to either of these programs. Open Book did offer this feature at one time and even has a setting to turn off the Spotlight, which they say helps with Braille tracking; however, they do not directly support Braille displays without a screen-review program.
While I have extensively discussed some of the common features shared by Open Book and K1000, other features are unique to each program. Probably the most notable of these in the new Open Book version, which we haven't discussed yet, is the ability to send and receive e-mail. What does this have to do with the scanning and reading of text, you might ask? Good question. Jim Fruchterman says that this is one of the most common requests Arkenstone has received from its users, even more common than surfing the Web. He adds that this capability allows people to send and receive e-mail from within a familiar environment.
The mail system is of course not as full-featured as Outlook Express or Eudora Pro but should meet most people's needs. It will allow you to send and receive e-mail, reply to and forward messages, send and receive file attachments, maintain an address book, and send messages to individuals or groups. You can establish different folders for mail and move messages into these folders. There is, however, no way to filter or move messages into specific folders automatically.
As an experienced computer and e-mail user I found it easy to set up the e-mail system and to send and receive mail. I had no trouble using any of the features after I read the instructions.
The e-mail option seems to be aimed at beginning users who may have trouble setting up or using e-mail, but you must have an e-mail account somewhere and an Internet Service Provider, ISP, or direct network connection. You must also use Dial Up Networking, a program that comes with Windows, to set up your computer for your ISP and e-mail account. Open Book will not help you with this process; you will need a screen reader or sighted assistance. Once you have a working connection to the Internet, you can set up your Open Book e-mail. You will need a little technical information like the addresses for your SMPT and POP3 mail servers, but the manual explains all of this fairly well.
The addition of e-mail in Open Book is an interesting move. On the plus side Arkenstone is attempting to meet an expressed need of many of its users. On the minus side they are trying to re-invent the wheel to a great extent. A number of very good commercial mail packages are on the market, one of which, Outlook Express, comes free with Windows. People are likely to request additional features, and this, combined with evolving and changing standards for all software, means that there will be pressure to maintain and improve the e-mail applet. Also Arkenstone has chosen not to deal directly with the whole issue of getting connected to the Internet. However, some customers will call for help with this process, so e-mail support may cost the Open Book technical support staff dearly in the long run. Also the next logical extension after e-mail is Web surfing. However, at some point Open Book with too many features would start to cannibalize the market for JAWS for Windows or JFW Lite, which is supposed to be released in the near future.
Page Layout Description and Navigation
Yet another interesting new addition to Open Book version 5 is Page Layout Description. In this mode Open Book will tell you about the different layout elements on a page, allow you to navigate from element to element, and read the text within a given element. The page elements that the system identifies include Columns, Headings, Text Blocks, Graphics, Captions, Tables, Headers, and Footers.
You first need to adjust a few items in the Settings Dialog Box. The manual explains this completely. You next scan and recognize a page. Then you can get a summary of the layout elements on a page or go into one of two layout-description modes. The magazine page which I used in the scanning tests, for example, told me that it had eight text blocks, five graphics, four columns, and four headings. The two-page layout modes are Guided Tour and Explore Layout.
In the Guided Tour Mode, Open Book automatically takes you from element to element, announcing them as you go. You can adjust the speed it uses to move through the document. You can stop it at any time, read the text in an element, or go into Explore Mode, where you control the movement from element to element. You can read or edit text from Explore Mode and jump to Text or Exact View.
One interesting sidebar of the page layout description features is the way it gathers its information. It uses information from two recognition engines, the CAERE and FineReader Engines, deciding for each element which engine is the most sure. In other words it lets the two engines vote on all decisions. While this adds time to the recognition step, it presumably gives you a better picture of what is happening. As computer power improves and technology evolves, I can see some of the voting techniques being applied to the recognition of text, giving us still more accurate documents.
Force Feedback Mouse
Arkenstone has added a cool new feature to this area of the program. If you have a Logitech Wingman force feedback mouse, you can explore the page elements tactually. The Wingman, which costs approximately $100, is a mouse that tries to give you a feel for things by providing resistance (force feedback) as you move the mouse. It consists of a hard plastic mouse-pad-sized device with a mouse permanently affixed to its surface. The mouse moves around in about a three-inch square on the surface of the hard plastic pad but will not come off. The device connects to your computer by a Universal Serial Bus, USB, port and must also be plugged in to electrical power.
I successfully installed the device on an IBM PL-310 computer but was unsuccessful at installing it on my Dell. The computer seemed more prone to crashing after this installation, but I did not have enough time to work through all the possible issues.
With the Wingman force feedback mouse connected and the page-description features turned on, you can move the mouse around, exploring the page layout elements on the page you have just scanned. The field of movement is fairly small, somewhat smaller than the total area in which the mouse can be moved around. As you move the mouse, you encounter areas of resistance. It could be, for example, a large area denoting the edge of the document or a smaller area marking the boundary of a column of text. Each layout element--text block, heading, or graphic, for example--is supposed to have a unique feel. Also the size of the element and its position in relationship to the other elements on the page are reflected in the feedback you receive.
Does it work? Yes and no. I certainly was able to explore the general layout of the document. I did get information on relative position and size. However, the tactile distinction between different elements was only minimally discernible to me. The differences are subtle at best.
Is this a useful tool? It depends on your needs. I can certainly see that, if you do page layout, teach, do demonstrations, or work on layout with sighted people, this device might be useful. This is in large part because, once you have navigated to a specific element, you can press the left mouse button on the Wingman, and the text within that element will be read to you. The Wingman force feedback mouse is an interesting first development but one that won't change most people's lives. Like the improvements cited above with RealSpeak from L&H, I think we will benefit in the future from this work, and I hope that Arkenstone, Benetech, Logitech, and others keep it up.
Other Open Book Features
Open Book version 5.0 has a number of other new features which we haven't discussed, such as the addition of a pronunciation dictionary and an OCR correction list. Other improvements such as a third OCR engine, Preview Scan, and the Low-Vision Menu, have already been discussed. The program has two simple scanning settings: Scan for Accuracy and Scan for Speed, which are useful. You can quickly make one choice instead of having to change several things. However, overall its most important additions, in my opinion, are the inclusion of e-mail support, page layout description and navigation, and support of the Logitech Wingman force feedback mouse.
Unique K1000 Features
Like Open Book, the L&H Kurzweil 1000 system has several unique features. You can establish both a favorites list for folders and one for files. It is then easy to jump to these frequently used areas or files. K1000, as mentioned earlier, has the ability to summarize the contents of a document. It does this either by using bookmarks you establish yourself or by using keywords you give it. It will also go through a document making its best guesses about what is important, presumably based on layout. I tried this, but the results were not particularly impressive. It is an interesting and potentially useful idea though, and, depending on the document itself, could be quite useful.
Reading tables can be difficult. In version 5.0 K1000 has added settings and commands to help with this task. With the proper settings made, K1000 automatically identifies tables. A set of commands is available to read within a table, by cell, column, row, and the like.
Support of ZIP files is another unique feature K1000 offers. A ZIP file is a compressed version of one or more files. It is a way to store things so that they take up less space. Many programs and files are zipped prior to transferring them to another computer on-line because the files are smaller and contain multiple files. K1000 will treat any ZIP file like a folder, allowing you to look at and manipulate the files within the archive. It will also add files to a ZIP file.
Finally, K1000 comes with some extras. When you purchase the program for the first time, you get a CD-ROM with hundreds of electronic books on it. These text files are of books that are now in the public domain. Also your K1000 CD-ROM contains a folder of extras. This is a collection of DOS and Windows programs, games, and utilities, some of which are freeware and some of which are shareware. There are plenty of things to play with if you are so inclined, and you may find some useful programs there.
Now let's get down to what you have been waiting for: the results of the scanning tests. I was no longer able to locate the documents I used in my last tests back in 1995, so I came up with a new set of documents, ten in all. These included a laser-printed letter; a relatively poor-quality fax; a bad photocopied page; a page from a magazine which had multiple columns, graphics, headlines, and color printing; a page from a mail order catalog containing graphics and small print; a relatively high-quality newspaper page; a paperback book; a hard-cover book; a page from a bank statement; and a one-dollar bill that I crumpled up.
While these tests are not exhaustive (they do not test how well the systems preserve formatting information, for example), they are practical. The test pages represent a cross section of the kinds of documents each of us wants or needs to read. For these tests we established a scoring system. We scanned each page and assigned the document one of the following five point values, based on where it fell in the continuum:
1. Garbage: random characters and indecipherable results.
2. Poor text: mostly garbage with a few understandable words sprinkled randomly throughout.
3. Fair text: mostly understandable text, but still a relatively high number of mistakes. Familiar or simple text can be followed, but unfamiliar or difficult material or material with numbers or computer commands may not be useable.
4. Good text: very understandable results with only a few mistakes; problems may occur only with numbers or computer commands and the like.
5. Excellent text: few if any mistakes, no more than one or two per page.
If the results fell between two numbers, half a point was awarded.
We started with the default settings for each software package and changed them only if the results were bad or if there was a specific setting for the type of page being scanned (fax or degraded text, for example). So, unless otherwise indicated, brightness, contrast, and print style settings are at their default or automatic values for all tests.
To end the suspense, here are the scores: Open Book scored a 44, and K1000 got a 41.5. Below are the document types, scores for each program, and an explanation where necessary.
DOCUMENT TYPE OPEN BOOK K1000
Laser printing 5 5
Fax 3.5 3
Photocopy 2 2
Magazine 5 4
Catalog 5 4.5
Newspaper 4.5 4.5
Paperback book 5 4.5
Hardcover book 5 5
Bank statement 4 4
Money 5 5
As you can see, many of the document types scanned identically or nearly so. The photocopy and fax were difficult for both programs. I got a somewhat better result with the fax on Open Book by switching the document type to typewritten/dot matrix. The K1000 made some OCR mistakes and missed a little text on the magazine page, which was complex. I turned on dynamic thresholding and changed the document type to "Degraded Text." I also used the Automatic-Contrast setting with Open Book. Finally, I was able to improve the OCR quality when scanning a hardcover book with K1000 by switching to the FineReader engine. Otherwise I used the RTK OCR engine.
I urge you not to base your purchasing decision on my scanning scores. The scores are for the specific documents I chose only, and more time and experience with the software might have yielded better scores for one or both systems. Further, the point separation between the two systems is not really very great when you look at the overall scores. Both companies have demos available, and I urge you to get the demos and try some of your own documents.
You should also consider other factors including price, special discounts, dealers, technical support, what your friends or colleagues use, features, and more. We hope that these tests and reviews can give you some guidance, but they won't make the final decision for you.
You would probably like to know which system to buy. I can't tell you. The right answer for me isn't necessarily right for you. While I do have access to both programs, I would be perfectly happy with either if I had to give one up.
Open Book has been around a long time and has a loyal following. The company made great advances in the PC-based reading market in the early and mid 1990's. They seemed to lose some steam for a while, but things have picked up again in the past year or so, and they have a competitive product. Arkenstone's acquisition by Freedom Scientific is likely to make some difference in the future, although at this time it is impossible to know exactly how. Both Arkenstone and Kurzweil have relied on a network of dealers. Many dealers sell both products, a practice that many think Freedom Scientific will bring to an end. Around the time of this writing, we learned that Freedom Scientific has terminated all dealer relationships with HumanWare. Other dealers are concerned that they will follow.
K1000 on the other hand is a newer product, one that has matured quickly and nicely. For several years Kurzweil was putting out two upgrades per year. As the product has matured and the development team has taken on more projects, this pace has slowed somewhat, but the program is still a competitive and feature-rich program.
Overall, I think K1000 is probably best for beginners. The manual explains things in great detail, including scanning and reading concepts and Windows commands, and it seems to do a little more hand holding than Open Book. The program, for example, most of the time remembers where you left off reading a file when you reload it. Depending on your perspective, though, you may find the manual long and tedious. It reads more like a reference than a narrative and takes thirty-five pages to get to the point of scanning a document. I have previously discussed RealSpeak, which you are likely either to love or to hate. Overall, I find K1000 a little sluggish in its response to keystrokes, a situation exacerbated by RealSpeak. However, the program has a wealth of features and is likely to be able to do almost anything you need. I have had good experience with its technical support staff; the developers are very accessible by e-mail, and it runs an Internet Mailing List which yields good answers. The staff monitors the list and responds quickly. Many knowledgeable users also help out. Further, several experienced users who regularly work with both K1000 and Open Book tell me that they can more quickly scan and read a pile of materials with K1000 than with Open Book.
In late November we learned that Lernout & Hauspie Speech Products NV has filed a chapter 11 petition for reorganization protection under the U.S. bankruptcy code. So how does that affect the Kurzweil Educational Systems group, its products, and the people who use them?
The following answer to this question was posted to the K1000 listserv by Stephen Baum. "The short answer is that it hasn't. We are all still here at work, and our responsibilities and goals have not changed. The engineering group is working on the next releases of the Kurzweil 1000, MagniReader, and the Kurzweil 3000. Marketing continues to market, sales continues to sell, and all of us continue to assist the customer support group in serving our customers. We were acquired by Lernout & Hauspie a little over two years ago. The acquisition gave us access to some great technology, but in other ways we have changed very little. We are a small, focused, and remarkably stable group. We look forward to continuing to make great products and to providing excellent support for them."
It should be noted that a Chapter 11 filing for reorganization is just that, protection from creditors while the company reorganizes. It doesn't mean that L&H or KESI is going away, but we will have to wait and see what, if anything, it means to those of us who use products from Kurzweil.
It is difficult to judge the Open Book manual from the beta. It was concise and seemed to cover all features. What we were provided was entirely in the Help system although Arkenstone personnel indicate there will be a separate manual on disk and in print. There will also be print, Braille, and a taped command reference. Overall I like the design and feel of Open Book. It is fairly responsive with ViaVoice Outloud and has good accuracy and a variety of tools and settings to make adjustments. Some of my comfort may come from the fact that I have used it longer. We have heard mixed reports lately about tech support although my experience over the years has been good. Further, technical support and other operations have moved to the Henter-Joyce site in Florida, and we don't yet know if this has made or will make any difference.
Arkenstone also runs a mailing list, and Jim Fruchterman is quite accessible and responsive. We haven't seen much participation from other staff members, and Freedom Scientific recently posted a message to the list stating that, if you want e-mail support, you should write to the <[email protected]> address. It further said that the list would be used for posting announcements, technical support bulletins, and the like. Until recently Arkenstone had a toll-free support number. Freedom Scientific has ended this to the chagrin of some.
If you have an e-mail address, joining one or both mailing lists offered by Arkenstone or Kurzweil can be a good idea. To join the Open Book list, send e-mail to <[email protected]>. Put the words "subscribe users" in the body of the message, without the quotes. To join the K1000 list, send a message to <[email protected]>. In the body put "subscribe K1000," without the quotes. These lists can be a good place to have questions answered, find out what is going on, get suggestions from experienced users and staff, and more. There is also a list for discussion of scanning and reading issues which isn't product-specific. It is called "scan-talk." To join, send a message to <[email protected]> and put "subscribe scan-talk" in the body of the message.
For assistance with this and other technology considerations, contact the National Federation of the Blind's International Braille and Technology Center for the Blind. Since they have and use everything commonly available, they can help you sort through the issues you need to consider. You can reach them by calling (410) 659-9314.
Whether you choose to buy Open Book from Arkenstone or K1000 from L&H Kurzweil Educational Systems Group, you can't go wrong. They are both strong, competitive products that deliver what they set out to do within the confines of today's technology, and they keep getting better.
Freedom Scientific Blind/Low Vision Group, Arkenstone Division, 11800 31st Court North, St. Petersburg, Florida 33716, Sales: (800) 444-4443 (U.S. and Canada), Phone: (727) 803-8000, Technical Support: (727) 803-8600, FAX: (727) 803-8001, TDD: (800) 444-4443, e-mail: <[email protected]>, Web: <http://www.arkenstone.org>, Price: $995, Competitive Upgrade: $500
Lernout & Hauspie Speech Products, Kurzweil Educational Systems Group, 52 Third Avenue, Burlington, Massachusetts 01803, Phone: (781) 203-5000, toll free: (800) 894-5374, sales: either of the above numbers at extension 5037 or 5359, technical support: (800) 995-9905, fax: (781) 203-5033, Web site: <http//www.LHSL.com/kurzweil1000/>, price: $995 ($1,195 with DECtalk Access32), competitive discount: $595 ($795 with DECtalk Access32)