Christopher D. Byers

ANSWER: Anytime I get a question about a category of hardware or software, my first instinct is to ask a little more about the reason for the question. I am always concerned when someone focuses on the type of software rather than its intended purpose. If you were to ask about purchasing a mega-sports utility vehicle, the salesperson would ask questions like where you plan to use it, why this extra-large SUV instead of a standard-sized SUV, van, truck or station wagon. The salesperson already understands you need some transportation to move from place to place, but hasn’t yet learned what other needs you have beyond that. Do you need to transport large items on a regular basis? Do you drive in places where four-wheel drive is necessary? Well, you get the idea. The same pattern is true with computer technology.

From your question, I gather you’re interested in shortening the time it takes to prepare documents. There are many different technological options, including voice recognition, that can help accomplish this goal. The most-important thing going into this process (or any other technology purchase for that matter), is to be able to articulate your goals, determine if they are reasonable and keep focused on those goals during implementation to ensure success.

Many of my comments and observations, since they’re not based on personal experience, are taken from a session at the American Bar Association’s Techshow in March. Tom O’Connor, a consultant with Courtlink in Bellevue, Wash., and Steven Jones, an attorney from Little Rock, Ark., were the speakers. Their presentation was full of practical tips about using this technology.

Speech recognition has been around for many years. Initially, it was developed by companies like Kurzweil for the vision-impaired and disabled people. The early efforts were hindered by substantial processor and memory requirements and the limitations of the recognition algorithms. Algorithms are the rules used by voice technology to understand and translate spoken words into digital text or commands that can be used on a computer. Until about two years ago, the technology was limited to “discrete text” products. This meant that in order for the computer to recognize your words, you _ had _ to _ say _ them _ very _ slowly. This often meant multi-syllabic words were not interpreted correctly. Instead, they would appear as two separate words.

The three companies vying for market share are Dragon System with its Dragon Naturally Speaking products, IBM’s line of Via Voice programs, and Lernout & Hauspie’s Voice Xpress. If the name Lernout & Hauspie is not familiar, that’s because it is a Belgian-based company that bought Kurzweil two years ago. Lurking largely on the horizon is Microsoft, which has a project code-named “Whisper” that is intended to integrate voice-recognition technology into its entire line of products through the operating system. As best I can tell, this may debut with the next version of Windows NT, to be called Windows 2000.

The Pentium III chip from Intel Corp., which makes the processor chips for the majority of office-grade computers, is optimized for voice recognition. The other trends in this area to be aware of are the blurring of lines between the different applications, such as telephone, voice mail, e-mail and cell phones, with voice recognition. (As an aside, watch for products later this year that combine the Palm Pilot, a hand-held calendar and Rolodex product, with your cell phone. Talk about combining useful applications!)

IBM has a product marketed to disabled people called Home Page Reader, which uses speech recognition to work with your Internet browser. This should make its way into general use in the next few years as well.

The other major trend is to have speech-recognition capability available in a wide range of products. In addition to word processors, you can expect to see it connecting to case management programs and other law office tools over the next two or three years.

Improvements

The current crop of products has benefited from improved algorithms as well as cheap RAM. These allow continuous speech dictation, which means you can dictate in a more natural manner. They require much less training than earlier versions; the initial training takes about a half-hour. The recognition will improve as you work with it because it is constantly learning and watching you work.

That doesn’t mean, however, that there aren’t problems with the technology. These programs rely on user voice files. A bad cold could make you unrecognizable by your computer. A change of environment such as home, office and hotel may require significant retraining. One tip Jones shared at Techshow was to consider doing separate user files for different locations so you don’t mess up your office setup, which is most frequently used.

During the ABA session, Jones attempted to demonstrate Dragon Naturally Speaking with WordPerfect. Among his other complaints about this technology are that it never “goes to sleep” (the command that turns the microphone off) when you want it to — such as when the senior partner of the firm comes in to talk about a crisis involving the firm’s largest client. He also noted these programs create a pest even worse than typos. He referred to them as “word-o’s.” The reason they are so troubling is they can elude your word processor’s spell-checker. The voice-recognition programs will never misspell a word. These word-o’s are the result — words that are correctly spelled but are not even close to what you actually said.

The current algorithms used in these programs are context-sensitive so they are better than their predecessors in terms of getting it right the first time. However, Jones made an interesting observation. He said that your happiness and success with voice recognition will depend on your style of dictation. For single-draft writers, which I am, speech recognition can be a hindrance because you’re unable to ignore your mistakes. For those who like to get it down in rough form and clean it up later, this can be a good tool. However, he warned, this type of dictation is much more prone to word-o’s that are hard to edit since they are correct words but not the ones you intended. For instance, if you meant to say “influential,” it might be interpreted as “in the flu season.” One tip from an audience member was to create and save a sound file of your actual dictation along with the document so you can listen to your original thoughts again if you have trouble understanding what has appeared on your screen.

General tips

During Jones’ and O’Connor’s Techshow session, there were several tips, regardless of which product you select. They suggested you use the voice editor frequently to incorporate your own often-used words, such as client names, into your user file. They noted you can dictate directly into Corel’s WordPerfect 8 but not into previous versions of WordPerfect. In earlier versions, you can use the commands from WordPerfect such as Open File. He noted this technology is more effective for longer documents, but can be used for short documents only when the voice-recognition software already is loaded. The implication of his comment was that these programs take a long time to start, even on faster computers.

If you decide to use speech-recognition software, it is generally recommended that you purchase a very fast computer with a large hard drive and a lot of RAM memory — something in the neighbourhood of a 366- to 400-MHZ processor with at least 128 MB of RAM. If you’re using it on an NT network, you will want even more RAM. The good news is that it has become quite inexpensive in recent years. You can anticipate more problems with a laptop since the microphone is closer to the computer, which makes some noise itself and which can cause interference. Jones suggested getting a high-quality, noise-cancelling headset and microphone to improve your chances of getting the right words into your documents.

Choices, choices

Each major player in the speech-recognition market has more varieties of its products than Howard Johnson’s ice cream has flavors (or so it seems). Depending on which word processor you want to use with it, your options may be more limited and your choice clearer. Dragon Naturally Speaking from Dragon Systems (www.dragonsys.com; (617) 332-9575) comes in a standard edition ($109); a preferred edition ($229), which includes the Dragon Naturally Mobile option for transcribing text from a hand-held unit; the professional edition, which adds more macros and features suited to longer documents ($695); and the legal suite, which adds a 230,000-word legal dictionary and includes all the features of the professional edition ($895). Dragon Systems also offers a legal suite with the mobile option and recorder for $1,195.

Corel bundles the personal edition with its WordPerfect 8 Legal Suite ($499; an upgrade from other versions of WordPerfect costs $229). You may want to upgrade from the version in the legal suite if you want the mobile capability. One of the differences between the versions is the quality of the headset and the ability to make your own scripts or macros for specific tasks. From comments I’ve heard, Dragon products work best with WordPerfect.

IBM (www.software.ibm.com/speech; 800-825-5263) offers ViaVoice in several versions as well. There’s a home edition of ViaVoice 98 for $49.95 and an office edition for $89. These allow you to dictate directly into Microsoft Word or their own SpeakPad. An executive edition allows direct dictation into many Windows applications in addition to Microsoft Word. This edition can be used in a network environment with other people sharing the same PC for voice recognition. IBM also offers an add-on legal dictionary for $149.

Lernout & Hauspie (www.lhs.com; (617) 238-0986), which bought Kurzweil, offers a standard ($49.99), advanced ($79.99) and professional ($149.95) version of its product along with Voice Xpress for Legal ($249). Its products are optimized to work best with Microsoft Word. Microsoft Corp. owns approximately 8 percent of the company. Lernout & Hauspie also has signed an agreement recently with Novell to further develop its products to work in the Novell network environment.

For an excellent overview of how another lawyer uses voice recognition, take a look at Jim Eidelman’s article, “Talk to Your Computer: You Can Practice with Speech Recognition Software,” in the November/December 1998 issue of Law Practice Management Magazine, published by the ABA’s Law Practice Management Section. One of his suggestions is to purchase a headset that can serve as both the microphone for the speech-recognition program and for your telephone so that with the press of a button, you can switch from dictating to answering calls. This might solve the problem of needing to turn off the microphone if someone walks into your office.

The other accessory he discusses is portable dictating machines that now work as remote entry for the speech-recognition programs. Depending on which manufacturer’s speech-recognition software you choose, you can select the portable transcriber best able to work with that brand.

Hardware alternatives

If you’re looking at voice recognition and don’t have the hardware, the money for new hardware or the time or patience to train the software, there are some other interesting alternatives. Several companies, such as CyberSecretaries (www.cybersecretaries.com) and CyberTranscriber from SpeechMachines (www.speechmachines.com) are Internet-based service bureaus that allow you to call their service or use a dictating machine to dictate a letter. They then use voice-recognition software along with secretaries who proofread the results before returning the file by e-mail to the word processor of your choice. On a per-document basis, these services are expensive. However, if you weigh the convenience when you need it against the hardware, software and time investment of doing it yourself, it may be worthwhile.

Another alternative to keyboarding is the CrossPad by IBM (www.ibm.com/businesscenter/legal). This looks like a letter-sized pad but with a notable difference — when you write on it, it can save that text as a file for your word processor. I haven’t actually seen or used one of these, but they seem intriguing for lawyers still huddling around their legal pads for initial document drafts. I also could see where they could be useful for note-taking at client meetings since they are much less-disruptive than pulling out a laptop.

Most attorneys who use voice-recognition software regularly seem to use it primarily for getting the basic content into their word processor rather than formatting. While it’s a lot better than it was, it still requires a time investment. I wonder whether attorneys who haven’t invested the time to learn basic keyboard skills by now will have the patience to invest the time necessary to train and maintain a voice-recognition software program.

Carol L. Schlein is president of Law Office Systems, a Montclair-based training and consulting firm specializing in law firms. She formerly chaired the Computer and Technology Division of the ABA Law Practice Management Section. A lecturer for ICLE, she can be reached at (973) 746-6454 or carol@losinc.com.

Questions for Carol Schlein on law office technology may be faxed to New Jersey Lawyer at (732) 750-0010 or mailed to “Law Technology Questions,” New Jersey Lawyer, Koll Corporate Plaza, 485B Route 1, Suite 100, Iselin, N.J. 08830.