X

Thank you for sharing!

Your article was successfully shared with the contacts you provided.
“You talk, it types …” So goes an ad campaign for speech recognition software. Is it that simple? Not quite. Last year at this time it looked as if speech recognition was close to being ready for prime time. All of the major speech recognition companies rolled out continuous speech recognition software with great promise, and most industry analysts predicted that speech recognition would be ready for prime time within another year, as the improved speech recognition software and a new generation of desktop hardware converged to bring dictation to the desktop. A year and another generation of speech recognition software later, the desktop dictation market remains stagnant and speech recognition companies are shifting focus away from the desktop — to telephony and Web-based speech applications, such as voice portals to the Web for cellular phone users. Why? In a word, speech recognition is not yet good enough for the general use on the desktop. The problems that have bedeviled the speech recognition industry from the beginning — the need to speak with unusual clarity, the need for a quiet working environment, low rates of accuracy, heavy proofreading and correction demands, and the need for a high-end processor, sound card and microphone — continue to hobble the industry. NEW GENERATION Dragon Naturally Speaking 5 and IBM Via Voice 8 represent the newest generation of speech recognition software. Both are available in a wide range of versions, from basic “talk and type” versions to specialized versions targeted at the legal market. Dragon Naturally Speaking offers a range of general use products ranging from Dragon Naturally Speaking Essentials, a personal-use product, at $59, to Dragon Naturally Speaking Preferred Edition, for general office use, at $199. IBM Via Voice 8 offers a similar range of versions for general use, priced from $30 for the Via Voice for Windows Personal Edition to $190 for Via Voice for Windows Pro. The “preferred” and “pro” editions integrate closely with Microsoft Office and other standard general office software, allowing users to dictate directly into documents and navigate around within the applications (“Open a new document …” or “Add a row at the bottom of this table …”) which means that a lawyer who does not type can use a computer by talking to it, with minimal need to use a keyboard. The “preferred” and “pro” editions allow a user to develop extensive custom vocabularies. In addition, Dragon offers a specialized product targeted at the legal market. Dragon Naturally Speaking Legal Suite priced at $995, has a highly developed legal vocabulary (roughly 250,000 words) and a well-developed macro set that allows lawyers to insert boilerplate text, launch applications such as time and billing programs, search documents and so on. Via Voice does not offer a specialized legal product, but does offer a legal vocabulary add-on for Via Voice for Windows Pro, priced at about $150. HOW GOOD ARE THEY, REALLY? Naturally Speaking and Via Voice have jousted for supremacy in the desktop dictation market for several years, and the current releases of both are state-of-the-art. With time and effort, either product can be used as a keyboard replacement. But “You talk, it types?” Not yet. The new releases of Naturally Speaking and Via Voice perform about equally in real-life tests. Set-up and basic training takes about 30 minutes for both products. Accuracy levels between 90 percent and 95 percent are typical, and some users can push accuracy rates into the 96 percent to 98 percent range by using high-end microphones and sound cards, more extensive training, and careful attention to speaking clearly and precisely. Throughput (dictation and correction) typically ranges from 20 to 30 words per minute. In short, the new releases are not fast enough or smart enough to serve as a keyboard substitute for lawyers who can type at reasonable speed, and the products will most likely reduce productivity for facile keyboard users. Even with the problems, speech recognition can be a godsend for lawyers with special needs — lawyers who do not type or type quickly, and lawyers who have repetitive stress injury (RSI) or other impairments. Speech recognition software shines as a disability workaround for users with RSI or moderate small motor impairments (e.g. arthritis). Impaired users, in contrast to normal users, will benefit immediately from speech recognition software — a 60 wpm typist may chafe at the limitations of speech recognition software while a user suffering from RSI is likely to find the limitations negligible compared to the pain of using a mouse or keyboard. The same is true for slow typists or nontypists. Speech recognition software can be used to open and control applications and dictate directly into applications. Speech recognition software is a resource hog, challenging even today’s fast computers with large hard drives. A 500 MHz processor and 128 MB of RAM are the bare minimum for reasonable performance and speech recognition will use between 300-500 MB of hard drive space for most users. A high quality microphone is essential for reasonable accuracy. Lawyers who are serious about speech recognition software should toss the microphone/headset that comes with the product and invest in a high-quality, noise-filtering microphone/headset. A high-quality sound card is also essential. Many laptops and desktops with “on board sound” desktops don’t work well with speech recognition software, even if they otherwise have a fast processor and plenty of RAM. Speech recognition software will continue to improve incrementally over the next few years. Naturally Speaking and Via Voice will continue to improve accuracy as recognition engines are enhanced and better computers come on line. But accuracy will not increase dramatically until speech recognition is coupled with “personal context recognition” — the ability to analyze words in the context of an individual user’s idiosyncratic writing style rather than generic “context guessing.” Speech recognition software will also become more usable as word processing “auto formatting” develops, freeing lawyers from the tedium of switching back and forth between dictating and control functions to format text. THE BOTTOM LINE While both Naturally Speaking and Via Voice perform as advertised, the latest generation of speech recognition software is an incremental improvement over the systems that existed two years ago, but the Holy Grail of speech recognition — a system capable of meeting general legal office needs — remains elusive. Both systems are suitable to meet special needs, supporting lawyers who cannot, should not or do not use a computer effectively for word processing, are able to speak clearly, and are willing to master and use speech recognition. Neither Naturally Speaking nor Via Voice is yet ready for general use in the law office — while both systems offer a way to get words onto paper, the low accuracy rates mean that the systems are not as efficient as typing, particularly for lawyers who type reasonably well. Eric H. Steeleand Thomas ScharbachSystems Services are principals of Steele Scharbach Associates, a consulting firm assisting law departments and law firms with strategic technology design, planning and implementation.

This content has been archived. It is available exclusively through our partner LexisNexis®.

To view this content, please continue to Lexis Advance®.

Not a Lexis Advance® Subscriber? Subscribe Now

Why am I seeing this?

LexisNexis® is now the exclusive third party online distributor of the broad collection of current and archived versions of ALM's legal news publications. LexisNexis® customers will be able to access and use ALM's content by subscribing to the LexisNexis® services via Lexis Advance®. This includes content from the National Law Journal®, The American Lawyer®, Law Technology News®, The New York Law Journal® and Corporate Counsel®, as well as ALM's other newspapers, directories, legal treatises, published and unpublished court opinions, and other sources of legal information.

ALM's content plays a significant role in your work and research, and now through this alliance LexisNexis® will bring you access to an even more comprehensive collection of legal content.

For questions call 1-877-256-2472 or contact us at [email protected]

 
 

ALM Legal Publication Newsletters

Sign Up Today and Never Miss Another Story.

As part of your digital membership, you can sign up for an unlimited number of a wide range of complimentary newsletters. Visit your My Account page to make your selections. Get the timely legal news and critical analysis you cannot afford to miss. Tailored just for you. In your inbox. Every day.

Copyright © 2020 ALM Media Properties, LLC. All Rights Reserved.