July 2006 marks the 30th anniversary of the first large public demonstration of optical character recognition (OCR) technology in front of a large group of people who were blind. Since Ray Kurzweil's first demonstration of his then-revolutionary system, the development of products that address the complexity of converting the printed word to meaningful speech has continued uninterrupted. As we enter the fourth decade of this technology, it is useful to understand exactly what products are available and how they compare with one another.
This article focuses primarily on the two most popular packages that are intended for and marketed to people who are blind or have low vision in the United States: Kurzweil 1000, version 10, and OpenBook, version 7.02. A sidebar on OCR Alternatives considers some off-the-shelf options.
How We Tested
For this evaluation, we at the AFB TECH product evaluation lab compared several important characteristics of these products. First, how well do they recognize the text on a variety of printed pages. Once text is recognized, how do these products compare in their ability to follow the formatting of printed material. We also tested several examples of file conversion. This is an increasingly important function of OCR systems, since, in many instances, it is the only way to read inaccessible electronic files, such as untagged PDF documents. We also evaluated the installation of each product and share some thoughts on customer support.
Hardware Requirements
Since these OCR systems are intended to be installed on a personal computer, the user must provide several important hardware components. OCR technology can place much higher demands on a computer's processor than some common applications, such as word processors and web browsers. Each company has established minimum specifications for computers that will support their respective products. It is essential to understand these requirements before you purchase and install this technology. In our opinion, the quality of the experience you will have will be greatly enhanced by using a computer that is substantially more powerful than the minimum that is suggested by the software vendor.
In addition to the computer, a scanner is required to capture the text page for processing by the OCR program. As with computer specifications, specific models of scanners are recommended by the OCR vendors. Because scanner technology can change suddenly, do not assume that the favorite brand of yesterday is still a good bet today, and, conversely, brands that were once off-limits may now be recommended.
Many seasoned OCR users may recall with fondness scanners of yesteryear. Several of these scanners allowed the placement of a book along the front edge of the scanner, which eliminated the need to disassemble the book to ensure a flat placement of the page on the glass. A new book-edge scanner from Plusteck, Optic Book 3600, is now available. Several technical support professionals have told us that it can provide scans in five seconds, once it is configured and installed properly. Several important steps must be followed to ensure smooth scanning, so if the Plusteck is of interest, we strongly suggest that you contact your vendor before you purchase or install it.
The final consideration that will influence your experience will be the configuration of the OCR program as the only assistive technology on the system or its inclusion as part of a larger group of assistive technology programs. Both Kurzweil and OpenBook products can operate alone—providing their own speech output—or with a screen reader.
Installation
Before installing the OCR program, you will want to ensure that your PC is ready for the software and that your scanner is properly installed and ready. Scanners typically use the conventional installation "wizard," which you may already be familiar with from other Windows installations. We understand that in some instances problems can result from installing the software and connecting the scanner in the wrong order. Check your scanner's installation instructions carefully. Some scanners also include their own software package, which may include utilities that are intended for use with Windows. If you do not wish to install these utilities, consulting with your vendor's technical support staff may be prudent.
We believe that individuals with even basic computer experience should be able to install these programs independently. Both products require that you enter a key for authorizing the product. It is a good idea to be sure that the key is available in a format that you can read easily. Kurzweil 1000 provides several registration options, including by telephone, conventional mail, or online. The product number is provided in braille and is also scanable. OpenBook provides product registration numbers in print and braille. Products purchased directly from Freedom Scientific are already registered when you receive them, according to a Freedom Scientific sales representative.
Kurzweil 1000
Kurzweil 1000, version 10, is priced at $995. The package includes the product CDs and several additional CDs that contain a variety of classic literature. Full braille documentation is available upon request, including the product guide and the installation and quick reference guides. A cassette package of the documentation is also included, along with a print directory of the included classic literature.
Installation
Installing Kurzweil is a smooth operation that begins automatically when you place the CD in the computer. Whether or not you are using a screen-access program, the voice guidance provided by the installer is excellent.
Kurzweil 1000 is a large program, and the installation may feel longer than that of other programs. One important note: IBM ViaVoice is a component of the Kurzweil package. Many individuals have reported that this component can conflict with the voice of a screen reader. For this reason, we suggest that you address this potential problem with a Kurzweil customer support representative.
Getting Started
Kurzweil 1000 has a well-developed user interface that is based on many years of development. The beginner can listen to prompts that describe the steps involved in using the product, as well as information on how to learn about the functions of the keys that are required to scan and read a document.
Kurzweil reading products were developed 30 years ago, before the advent of the PC. In the early products, a specialized keypad provided navigation of the products' functions. This approach of a simple keypad layout is still available to the beginner. The advanced user can gain full access to the system with the conventional Windows keyboard and commands.
The Scanning and Reading Experience
After spending only a few minutes with Kurzweil, it becomes apparent that this is a powerful program with an appropriately comprehensive interface that provides a seemingly infinite array of configuration options. When installed and launched for the first time, Kurzweil 1000 will assume that you want to scan, process, and read each page as it is scanned.
Kurzweil 1000 worked reasonably well in its default mode when it was called upon to read simple documents, such as memos, reports, and other text on white paper. It produced highly accurate and understandable results.
Moving to more complex document formats, such as magazines, proved more challenging to the technology. Many factors will affect the accuracy of scanning. Complex or unusual formats; shaded or half-tone backgrounds; and poorly printed material, such as magazines or paperback books, will require your attention. Kurzweil 1000 includes important tools such as a choice of recognition engines, brightness control of the scanner, settings for removal of speckles, and many more options. It is in the use of the tools that Kurzweil 1000 and other programs provide that the differences emerged.
Version 10 of this highly sophisticated product is perhaps the most powerful and advanced assistive technology product that is available. Many important controls have been developed over the past several years and are at your fingertips. One example is the Optimize feature. This utility will try various settings for a particular page that you have placed on the scanner. Our first try, using the default settings, produced results that were less than ideal. After we used the Optimize function, a useful, if not 100 percent accurate, document emerged.
Kurzweil 1000 allows you to select one of two recognition engines. Each has strengths and weaknesses, and the results are dependent on the kind of material that you are recognizing. The reading process can also be tailored to your liking, with headings or other visually important material announced in a distinct voice. A cornucopia of options for converting and saving documents and settings is also available. Kurzweil 1000 can access some online information as well.
Several other tests—a paperback book and magazine page—demonstrated the importance of coming to terms with the controls and settings of this program. We noted that a simple change in the scanner setup, selecting dynamic threshold, produced an immediate increase in accuracy on several formats of text that we tested. We recommend that Kurzweil change the default setting from static to dynamic thresholding, taking advantage of today's scanner technology.
OpenBook 7.02
OpenBook is priced at $995. Version 7.02 is currently shipping. Included in each new purchase is the product CD, a Getting Started cassette, a braille Quick Start guide, and a print Quick Start guide.
Installation
Whether installed alone or with the assistance of a screen reader, OpenBook provided an easy-to-follow installation process that began automatically when the CD was placed in the computer.
OpenBook includes Connect Outloud, a basic screen reader. On the basis of the information that was provided by the prompts from the installation program, it was not clear how to proceed if you did not choose to install Connect Outloud and were not using an existing screen reader. We suggest that you consult with a technical support representative before you install the program to clarify this part of the procedure.
Getting Started
OpenBook opens like other Windows programs. If you are already familiar with using Microsoft Office and similar programs, then OpenBook will feel like an old friend.
This product did not provide prompting or descriptions of shortcuts to options. OpenBook also has a heritage that includes a small specialized keypad. A simple set of keys on the keypad can provide navigation and basic operations.
The Scanning and Reading Experience
The most striking feature of OpenBook was its successful recognition and formatting of both our basic memo as well as more complex formats such as the CD liner notes, using only the default settings. After installation, the product is ready to go without further tweaking. The behavior for both scanning and recognizing was snappy and responsive. The responsiveness of the product was especially noticeable when OpenBook was used with JAWS.
A wide variety of tools and configurations are easy to reach and understand, especially for those who grasp the concepts of the Windows operating system. Selecting and navigating the settings is as simple as pressing the Alt key and moving among the standard menus. In all instances, we found OpenBook to be responsive and well behaved.
OpenBook was not as responsive as Kurzweil 1000 in the self-voicing default mode. In one instance, the voice stopped responding when focus moved to an error dialogue box caused by an empty drive. In instances in which the system produced subpar recognition results, the program lacked some of the more advanced controls found in Kurzweil 1000. Using gray-scale scanning in Kurzweil can make the difference between success and failure with some difficult images.
Dealing with PDF Documents
PDF (portable document format) is a hot-button topic for many computer users who are blind or have low vision. For a variety of reasons, which space does not permit us to explore here, these documents can defy all attempts to read them using a screen reader and the Adobe Reader. In fact, this situation can become such a problem that for many, the term PDF means "problem document format." The most effective solution in these situations is to use an OCR program to recognize the PDF file.
PDF files can be of several varieties. Some contain no text; they can be thought of as a picture of a page of text. Other PDF files may contain text, which the computer could recognize were it not for other technical issues.
Both Kurzweil and OpenBook facilitate the conversion of PDF documents. Each program takes a different approach to the process. Kurzweil needs to be open to manipulate PDF files. Use the File Open dialogue to locate and load the file, as you would any other document. OpenBook offers a Print-to option in the File menu from the Adobe reader. If you are browsing PDF documents from a network drive or the Web, this is a particularly useful feature, since clicking on the PDF file will usually trigger the Adobe Reader on your system.
Both programs did well with a standard untagged PDF test document supplied by Adobe. The one-page letter was recognized with 100-percent accuracy by both applications. We also tested an interesting, more difficult file, a simple one-page channel guide, found at <www.xmradio.com>. This document could not be formatted by either of the programs, demonstrating that PDF documents, while usable in some instances, are not always accessible.
Technical Support
OCR systems are among the most expensive in the assistive technology arena. For this reason alone, customer service is an obvious and important consideration. In our experience, both Freedom Scientific's and Kurzweil's technical support representatives provided accurate, prompt, and courteous answers to both basic and complex questions.
Kurzweil provides toll-free support Monday to Friday from 9:00 AM to 11:00 PM eastern time. OpenBook support is available from 8:30 AM to 7:00 PM eastern time.
Freedom Scientific's technical support professionals are generalists who, in addition to supporting JAWS, the company's flagship product, answer OpenBook support questions. In every instance, we found their assistance to be of the highest quality. If you regularly integrate an OCR product with other Windows applications, the Freedom Scientific representatives are in an excellent position to assist you with all your questions.
The Kurzweil 1000 technical support staff specialize in the Kurzweil product. Their knowledge of the intricacies of their product is obvious, and their ability to explain complex issues is especially welcome.
The Bottom Line
Both OpenBook, version 7.02, and Kurzweil 1000, version 10, performed well in our tests of optical character recognition. We encourage you to view and compare the samples that we have included with this article to judge the relative performance on the limited number of scans that space permits us to include.
Deciding on which product is worthy of your first consideration, then, comes down to what kind of user you are and in what context you will use the product. For the novice user, Kurzweil 1000 version 10 may be better. For users with little exposure to a computer with speech output, the excellent prompts, reminder messages, and generally friendly feel of the process make for a successful introduction to assistive technology. When questions need to be answered or trouble needs to be resolved, customer service is available 14 hours a day from Monday to Friday toll-free. Furthermore, the ability to choose from several speech-output options, including Neo Speech, provides the best chance that you will find a voice that suits your listening preferences.
For those who want to integrate OCR functionality into a work pattern in which Microsoft Office is the centerpiece, OpenBook provides a responsive, intuitive, and well-organized approach to scanning and reading many kinds of documents. It also performs well out of the box for recognizing and formatting printed matter of both basic and some intermediate complexity.
We were disappointed in the overall performance of the speech of OpenBook in its self-voicing mode. The limited choices of voices, their more computer-like sound, and a bug that was revealed when selecting drive options suggest that this product works best with a screen reader, such as JAWS, OpenBook's sister product, that controls speech functions.
For the advanced user who requires precise control of the OCR process and the widest range of tools to meet the most demanding OCR challenges, we believe that Kurzweil 1000, version 10, is worthy of first consideration. The Optimize feature enables settings to be automatically adjusted for a particular job or text document. The highly granular control of the most obscure functions of the scanning and recognition is unsurpassed. This advanced control and user configurability demands that those who perform intermediate and advanced scanning tasks with Kurzweil 1000 orient themselves to the controls and idiosyncrasies of this powerful and complex product.
Demonstration versions of both products can be obtained from the respective companies. If you have a computer and scanner available, installing and comparing these powerful and costly applications for yourself may be the best way to go.
View the Product Features as a graphic
View the Product Features as text
View the Product Ratings as a graphic
View the Product Ratings as text
Product Information
Product: Kurzweil 1000, Version 10.0.
Manufacturer: Kurzweil Educational Systems, 100 Crosby Drive, Bedford, MA 01730; phone: 800-894-5374 or 781-276-0600; e-mail: <info@kurzweiledu.com>; web site: <www.kurzweiledu.com>.
Price: $995.
Product: OpenBook 7.02.
Manufacturer: Freedom Scientific, Blindness and Low Vision Group, 11800 31st Court North, St. Petersburg, FL 33716; phone: 800-444-4443; e-mail: <Sales@freedomscientific.com>; web site: <www.FreedomScientific.com>.
Price: $995.
OCR Alternatives
So you like the idea of a system that will control a scanner and allow you to read the text that you have placed on it. You may also be thinking that you would like an alternative to the Adobe Reader to handle that in-box full of PDF files. Despite your interest, you still suffer from a case of sticker shock. Several alternatives may be of interest.
People who are blind or have low vision are not the only ones who use OCR technology on a PC. Many business applications and processes rely on this technology. Two competing off-the-shelf products vie for the bulk of the OCR business. These products also make use of the internal OCR processors that are within both OpenBook and Kurzweil.
In instances in which a sighted person will be scanning and preparing materials for a blind person, an off-the-shelf solution is not only significantly less expensive, but allows the sighted person to use the familiar visual cues and conventions of Windows programs. Be forewarned, however: These programs are not always fully accessible and require intermediate to advanced Windows skills.
OmniPage, with a street price of $349, is an off-the-shelf, Windows-based OCR program. Many individuals, most with intermediate to advanced screen-reader skills, report that they have been successful using OmniPage as an alternative to the programs we have reviewed here. They caution, and our experience confirms, that using the program is a challenge in comparison to OpenBook or Kurzweil 1000.
The second alternative is FineReader. Generally distributed online, it is priced competitively with Omnipage. It also offers a powerful OCR engine and includes some of the challenging design limitations of programs that are only partially accessible.
An alternative strategy for recognizing inaccessible or poorly tagged PDF documents may be found in Adobe Acrobat, the $149 big sister of the Adobe Reader. We have included an example of this strategy with the unsuccessful conversions of the XM Radio Channel Guide. We did not conduct in-depth comparison tests of Acrobat with other recalcitrant PDF documents. Adobe's own information suggests that the process that is used in Acrobat is less rigorous than the processes that are available in either of the blindness-specific OCR products.
Return to article, or use your browser's "back" button.
