Please note that these materials are provided for historical purposes only. The information presented is out of date and may be neither accurate nor useful. External hyperlinks may no longer be valid.
Voice Technologies
ntroduction
Voice technologies include speech recognition and voice response. Other voice technologies, including speech verification, used for security, and speech synthesis, used for translating text to voice, will not be covered in this paper. Speech recognition technology replaces a computer keyboard with a microphone and software to change the spoken word into typed characters--ideal for court staff with physical impairments or who lack typing skills. Voice response systems interact with callers through a telephone, asking questions and accepting responses to provide general information about the court, explain case and calendar status, schedule hearings, accept pleas, and receive fines or other payments to the court. Both speech recognition and voice response technology can improve public access to the court and enhance staff productivity.
Speech Recognition
escription
The primary components of a speech recognition system are a microphone, a PC card (with a digital signal processor, or DSP chip), and specialized speech recognition software. The personal computer, using its speech recognition software, interprets information received through the microphone, and presents it to other computer applications as if it had been entered on a keyboard.
As the user says a word, the computer analyzes the sounds, or combinations of phonemes, and compares them to a dictionary of tens of thousands of words. When a match is made (this is called word spotting), the program passes the characters to the word processor or other PC application. At the same time, the computer lists a number of other possible matches in a pop-up dialogue box on the screen. If the word selected is correct, the speaker continues. If an incorrect word is chosen, the speaker selects the correct choice from the list. If the correct word is not on the list, the speaker can enter spell mode, and add the new word to the dictionary. The most impressive feature of speech recognition software is its ability to apply artificial intelligence software to learn from its mistakes. The incidence of incorrect word selection diminishes rapidly as a speaker continues.
ackground
Great strides have been made in speech recognition technology since the late 1970s. An array of different approaches have been used: dynamic time warping, spectrogram interpretation, hidden Markhov models, and statistical word spotting are examples. The single greatest advancement lies in the increasing power and declining cost of the microprocessor chip. While accuracy has remained at around 95 to 97 percent for discreet, or single-word recognition, costs have declined from tens of thousands to a few hundred dollars, and dictionaries have grown to over 60,000 words. Speed has increased from about 30 words per minute to 90 words per minute, approaching the normal speaking pace of about 120 words per minute.
As word spotting gives way to phrase and concept spotting, speech recognition will make further gains. Speaker-independent continuous speech recognition will become a practical method of capturing the court record and producing instant transcripts. The next generation of PC applications will contain standard speech recognition features, making the microphone a viable alternative to the keyboard for the human/computer interface.
pportunities
Speech recognition makes staff more productive. Information can be entered into a computer while the user is doing other tasks--opening mail, shuffling through files, and the like. For inexpert typists, speech recognition systems can minimize the need for secretarial support in transcribing dictation and drafting documents, thus allowing the disabled to work at jobs that would be impossible to perform productively without using a keyboard. Speech recognition will improve telephone voice response systems and public kiosks by providing a more natural interface with the computer. It may be the long-term alternative to present labor-intensive methods of capturing the court record and producing transcripts.
imitations
Speech recognition technology today is less than perfect. Like any other skill, the technology requires time to learn, concerted practice, and regular use. Accuracy in a well-functioning system may be from 95 to 97 percent, but with an untrained user or in a noisy environment, the results may be less satisfactory. For skilled typists, a keyboard is still the fastest way to enter information into a computer system. Faster and cheaper microprocessors may change this equation in the next few years.
osts
Speech recognition systems are becoming more powerful and less expensive. Windows-based software products that use standard 16-bit audio cards range from $395 (with a 5,000-word vocabulary) to $695 (with a 30,000-word vocabulary) to $1,695 (with a 60,000-word vocabulary). When the software is packaged with digital signal-processing hardware, a vendor may charge $3,000 for a system with a 30,000-word vocabulary.
Voice Response
escription
Voice response systems use computer hardware and software connected to a telephone system. Audio menus and scripts are developed to guide a caller to the appropriate information. Using a touch-tone telephone keypad, the caller selects from a menu of services provided by the voice response system.
The first menu may offer several general choices. Prerecorded messages explaining options are stored in the computer and played as requested. The caller continues to respond to more detailed choices offered by the system until the request is specific enough to be met. Information about the court can reside in a stand-alone system; case status and calendar data can be provided if the system is connected to the court's case management system. The most sophisticated systems allow the caller to make payments with a credit card or electronic funds transfer, schedule or continue a proceeding, or request that a document be sent to a fax machine.
ackground
Voice response systems have been in operation for some time. As early as 1989, Michigan callers could get court information automatically over the telephone. Voice response systems in Michigan now provide a wealth of information on a wide range of court-related topics, all automatically. Callers in Arizona can enter their case number and find out if child support checks are in the mail. Child support information is similarly available in Florida and Missouri. In California, traffic violators obtain information about their tickets, sign up for traffic school, and pay fines with their credit cards--all with a touch-tone telephone through a voice response system.
Some attorneys in Florida schedule hearings for particular judges automatically by entering their attorney code, case number, and desired length of the hearing. The system then gives the attorney a choice of possible times and dates. If one is acceptable, the hearing is automatically added to the judge's calendar.
As speech recognition and voice response technologies merge, systems will respond to voice commands instead of telephone key entries. Faster and easier access to court information will be available without human intervention.
pportunities
Voice response technology improves public access to the court, and is an effective method of providing answers to commonly asked questions. Staff productivity is improved when less time is spent on routine requests and actions, saving resources for more unique and complex questions and problems.
Voice response technology improves the quality, consistency, and fairness of responses to the public. Certain services can be made available around the clock, any day of the week. The court users save time and expense while transacting business from their homes and offices instead of traveling to the courthouse. Voice response technology can overcome language barriers, because scripts can be created in any language.
imitations
Voice response technology also has a number of limitations. Most inexpensive systems require a touch-tone telephone; upgraded systems that accommodate rotary dialing are more expensive.
The complexity and depth of menus also are a problem. Given too many choices, the user may forget earlier alternatives while listening to later ones. If the script is broken into smaller menus with just a few options, navigation can take too long and frustrate users.
Security is another issue. If the system is used to schedule actions or take credit card payments, additional safeguards may add to the cost and complexity of the system. Integration of the system into existing computer, financial, and telephone networks can also be a challenge.
Most important, many people dislike talking to machines. They prefer interaction with humans to the sterile, disembodied monotone of most voice response systems. As technology improves and as people adapt to this technology, these concerns may diminish.
osts
Voice response systems that integrate with multi-line telephone systems and with the user’s computer applications range from $25,000 to $35,000. Over 100 PC-based voice response packages are available and can be purchased from $395 to $10,000, exclusive of telephone installation and line charges. Bank-processing charges should be considered for systems that allow financial transactions. There are also programming costs to keep information current.
Assistance
The Technology Information Exchange Service (TIES), operated by the National Center for State Courts and funded by the State Justice Institute, provides a number of services to courts:
- a Court Technology Bulletin highlights the latest developments in voice technologies;
- a Court Technology Database offers information on courts using voice technologies across the country;
- a vendor-neutral Court Technology Laboratory is available to test and compare products;
- the Technology Information Service shares information collected from industry publications, evaluation reports on packages in operation in courts, and data on vendors who sell systems.
To receive information on TIES or any of the other technology research, education, or consulting programs of the National Center for State Courts, call Court Technology Programs at (757) 259-1544.
This document
was developed under a grant from the State Justice Institute.
The points of view expressed are those of the authors and do not
necessarily represent the official position or policies of the
State Justice Institute.
Copyright 1995 National Center for State Courts