When optical character recognition (OCR) software was introduced, its ability to scan and recognize characters and text seemed like magic. Today, we take accurate scanning and recognition almost for granted. This review examines OpenBook, version 7.0, from Freedom Scientific and Kurzweil 1000, version 8.0, from Kurzweil Educational Systems, the two leading adapted OCR software packages on the market.
Both programs do far more than scan the printed hard-copy page. They search for and download books and magazines from the Internet, send e-mail and faxes, convert files from all sorts of other formats, and make manipulating and interacting with written material almost as efficient for people who are blind as for those who are able to read print. Essentially, the creators of both programs have attempted to look at all the myriad directions from which literacy can be approached with a computer and then have built some semblance of each of them into the programs.
Between the Lines
To determine accuracy, we used a variety of text styles and formats. For a simple test, we printed a sample of 12-point black text on white paper using the Arial and Times New Roman fonts. For greater difficulty, we introduced bold, italics, underlining, and tiny print to our printed material and decreased the font size. We added deliberate errors to our test pages, such as "tht" for "that" and "tke" for "the." For more advanced scanning, we used magazines printed on multicolored paper, books with diagrams, newsletters with columns, double-sided business cards, and more.
Testing was done in Windows XP and Windows 98. The scanners we used were the HP IV and HP 2C from Hewlett-Packard and the Epson 1650.
OpenBook, Version 7.0
Installation and Documentation
OpenBook comes with IBM's ViaVoice synthesizer, which provides speech throughout the installation. If a previous version of OpenBook is found on the hard drive, you are prompted to delete it before continuing. OpenBook's help system consists of a contents and tree-view setup. The tree view contains topics that constitute the OpenBook manual. You can use the arrow keys to open each topic. Pressing Enter will have the program speak the selected topic and subsequent text. OpenBook's documentation can also be found in print and braille and on Freedom Scientific's web site. A Key-Describer mode is included for new users who are not acquainted with OpenBook's command structure. You press the Key Describer key and then any other key to hear that key's function. Both the number pad and conventional Windows hotkeys can be used to navigate OpenBook. Keyboard layouts for Jaws for Windows and Window-Eyes are included for users of these screen readers.
A Simple Scan
OpenBook had no trouble scanning the various text styles and formats as long as they were larger than 4-point font. To scan 4-point print or smaller, it was necessary to change the scanner's resolution to at least 400 DPI (dots per inch) to decipher what was on the page.
OpenBook did not correct our deliberately misspelled words, However, this version does fix a previous problem with the omission of long dashes from scanned pages. In previous versions, two words separated by a dash would appear as one long word; this version recognizes dashes correctly.
Open Book can now be used to scan documents directly into Microsoft Word. If you want to add information to a Word document that is only available in print or to edit a scanned document in Word, you just choose the "Scan with OpenBook" option from the File menu in Word. OpenBook then opens, scans the page, and inserts the text in Word at the cursor.
OpenBook comes with the Omnipage, Recognita, and Fine Reader scanning engines. Although these engines are supposed to contain distinct features, such as different recognition speed and different abilities to recognize languages, we found little difference during our evaluation. For example, OpenBook took over 40 seconds to scan and recognize text, no matter which recognition engine we used.
Scanning in Depth
OpenBook made few errors in our magazine with multicolored paper. Pictures were ignored, but the text around and underneath them was read accurately. However, parts of the text in one column were sometimes read in the middle of text of another column.
OpenBook did a good job of reading newsletters and business cards. Few changes needed to be made to the default settings to read the text. Nevertheless, zeros were often misread as Os, and the "com" at the end of e-mail addresses was often mistaken as "corn." We also found scanning documents with multiple languages to be a problem. While OpenBook recognizes a wide array of languages—including French, German, and Finnish, to name a few—ViaVoice is unable to toggle among them while reading text.
More Than Scanning
OpenBook has the ability to save files as MP3s. You can then download these files to a portable MP3 player and listen to them on your daily commute to work or on vacation. OpenBook also supports the DAISY file format, so you can download your favorite books from Bookshare.org and other online sources.
One of OpenBook's most innovative features is the Freedom Import Printer, which allows you to recognize, format, and read PDF files and turn them into text. The ability to translate unreadable PDF files can be especially helpful for college students who attend institutions where electronic resources are often scanned as inaccessible PDF images. In addition, OpenBook identifies American and Canadian currency with its "BuckScan" feature. Simply place a bill in the rear right corner of the scanner to determine how much money is left in your wallet.
Kurzweil 1000, Version 8.0
Installation and Documentation
The Kurzweil 1000 comes with the Microsoft speech engine, which guides you through the setup process. If there is a previous version of Kurzweil installed, it will be removed and replaced with the new version.
The product manual and quick reference guide are available in the Help menu, as well as on Kurzweil's web site. The quick reference guide allows you to learn important hotkeys in a timely manner. The Kurzweil uses the function keys and the number pad in conjunction with conventional Windows menus for its commands.
A Simple Scan
In addition to the default RTK recognition engine, the Kurzweil comes with the Fine Reader engine. Although the Kurzweil had no difficulty scanning the text styles and formats we chose as long as they were larger than 4-point font, for smaller fonts it was necessary to increase the scanning resolution. The Kurzweil scanned and recognized a page in roughly 15 seconds but, unlike previous versions, it did not correct our deliberate errors.
Scanning in Depth
The Kurzweil had no difficulty scanning magazines with multicolored pages. Articles and advertisements that contained graphics were read with few mistakes. Columns also presented little difficulty. However, the Kurzweil occasionally confused numbers and letters on the address lines of business cards.
The Kurzweil recognizes a variety of languages, including Portuguese, Dutch, and Swedish. It will change reading languages more than once on a page if you change the "Language Identification" option to "Per Paragraph."
Additional Features
The Kurzweil identifies bills using its Recognize Currency feature. It identified all the bills we tried. MP3 files can be created from pages that have been scanned. The files can then be played on an MP3 player or by a media player on your computer.
With the Online pull-down menu, you can search for books, magazines, or news about the Kurzweil product itself. When "search for books" is selected from the Online menu, you can type in an author or title and then select which online sources to search. Among the list of sources are two that will be familiar to many AccessWorld readers: Bookshare.org and Web-Braille (the site that makes available braille files of books and magazines from the National Library Service for the Blind and Physically Handicapped, NLS). For example, if you select Web-Braille and type Grafton in the author box, the Kurzweil will ask for your user name and password and present the 15 titles by popular mystery writer Sue Grafton that are available for download—all in about six seconds. Select the desired book or books for download, and, voilà! they are on your hard drive and ready for reading. The Kurzweil can also convert files from braille or DAISY formats instantly, which can then be saved as any other format of your choosing.
The magazines available in Version 8 are those available on Web-Braille (all those that are produced in braille by NLS), as well as the Braille Forum and the Braille Monitor. Version 9 (to be released at about the same time as this article) includes AccessWorld in the magazine search options.
Although most computer users already have a word processor of their choice, the Kurzweil includes an easy-to-use text editor. In addition to the usual arsenal of editing tools, the most notable features of this aspect of the program are its dictionary and thesaurus. By highlighting a word in a document being read or just typing in a word, you can quickly obtain definitions or synonyms. In addition, the Kurzweil recognizes and converts PDF files into text and then reads them. However, the quality of the resulting text depends on how the file was created—photocopies will not be as clear as original files.
Some people have begun to use the Kurzweil as their main computer application. They use it to read documents, to search the web, to look up words in its dictionary, and to perform other word-processing functions. The main task for which they switch to another application is writing.
Let us say that you have just scanned a hard-copy contract for a new mortgage, and you want to send it to your brother, the lawyer, for examination. By choosing the "Send to" option in the pull-down files menu, you can compose the e-mail right where you are, and the Kurzweil will launch your e-mail client for you and send the message. And speaking of the "Send to" option, let us say that the contract is so encrypted in legalese that you would like to study it during your commute to work tomorrow. Sending a file directly to such popular devices as the BrailleNote, PAC Mate, Book Port, BookCourier, and other products is easily accomplished with a few keystrokes. Once you have scanned a document, you can make an additional hard copy of it through the Kurzweil. Or if you prefer to fax a document that has just been scanned or is already on your hard drive, you can fax it as well. All these options work well, but setting them up takes time and may be a bit daunting for novice users. Just as the Kurzweil needs to be told what device it needs to look for to find the scanned material to recognize, so must it be told which printer to send material to (presumably, the printer already in use for printing documents from other applications). Once the fax capability has been set up (do not forget to connect the telephone line to your computer), faxing from the Kurzweil is not difficult.
Did It Really Say That?
You can allow the Kurzweil to tell you, upon scanning a new document, the percentage of predicted accuracy. (This feature can, of course, be turned off if you find it annoying.) While scanning accuracy is about 95% most of the time, the Kurzweil will undoubtedly encounter some hard-copy print that is incomprehensible. The announcement of accuracy is particularly useful in forewarning you of this event (which occurred in our tests mainly with bulk-mail advertising or product packaging of various kinds). Most books, magazines, and computer-generated documents (such as memos and bills) pose few difficulties. In our tests, however, the program did present an occasional disconcerting message, both in scanning hard-copy documents and in reading electronic files, that an error had occurred and the program needed to close. Again, this was a rare instance and was delivered with extreme courtesy ("We are sorry for any inconvenience this has caused"), but warrants mentioning. In addition to reading aloud, the program offers a variety of screen-magnification and contrast options. It also provides braille, but in our tests, the braille appeared only in messages and pull-down menus and was not available in the text itself.
Recognition Complete
OpenBook version 7.0 and Kurzweil 1000 Version 8.0 have similar features. They both do a good job of scanning and recognizing printed material and include more than one recognition engine. Both also include many additional features, some directly related to scanning and some not. As in our previous evaluation, a major difference is speed. Kurzweil 1000 has a significant advantage in that it scans and recognizes print more than twice as fast as does OpenBook. Using either program, you can break down the barrier keeping you from almost instant access to printed material.
Manufacturer's Comments
Freedom Scientific
"Freedom Scientific would like to expand and clarify a number of issues addressed in the article. When using BuckScan, the position for bill placement is scanner-specific; place the bill at the point where the scanning head begins scanning. OpenBook's default settings are capable of handling all fine print normally found in documents such as contracts or legal papers. Four-point print is roughly the same height as a braille dot and is rarely found in any type documents. OpenBook 7.02 (August 2004 release) includes great improvements in scanning and recognition speeds, as well as updated versions of all the OCR engines for even more reliable scanning. The versatility of the Freedom Import printer is not limited to Adobe PDF documents. You can also use it to recognize text in graphics or any other document or image that you can open and print using a program installed on your computer. Connect Outloud, which is included with OpenBook, broadens the access provided by OpenBook to provide speech and braille access to the Windows operating system, Internet Explorer, and Outlook Express."
Kurzweil Educational Systems
"Thank you for a thoughtful review of Kurzweil 1000 Version 8. Kurzweil 1000 ships with two text-to-speech (TTS) engines: NeoSpeech (male and female voice), and IBM TTS (large number of voices in many languages). Kurzweil 1000 also ships with Microsoft TTS (both the SAPI 4 and SAPI 5 versions are on the product CD), but it is not installed automatically. Kurzweil 1000 does not directly support refreshable braille support, but braille support can be provided by your screen reader.
"Kurzweil 1000 Version 9, available in late summer, has many new features, including online encyclopedia search, a Create a Note feature to annotate documents, and DAISY file support (text only). Also, OCR is improved with an update of FineReader 7 and the addition of ScanSoft OCR® Version 12 (the RTK OCR no longer ships with Kurzweil 1000). Toll-free customer support is available."
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: OpenBook 7.0.
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: <http://www.FreedomScientific.com>.
Price: $995.
Product: Kurzweil 1000, Version 8.0.
Manufacturer: Kurzweil Educational Systems, 14 Crosby Drive, Bedford, MA 01730; phone: 800-894-5374 or 781-276-0600; e-mail: <info@kurzweiledu.com>; web site: <http://www.kurzweiledu.com>.
Price: $995 with FlexTalk speech, $1,195 with DECtalk speech.