Sprint PCS
bremsdj@sprintspectrum.com
CHI '97 Workshop: Speech User Interface
Design Challenges
"Spoken digit dialing" can be valuable in situations where hands-free operation is desired - such as when driving in a car. However, in the more general sense of everyday telecommunications, "spoken digit dialing" suffers a number of disadvantages that make it less desirable than a touch-tone user interface. Specifically, recognition of spoken digit strings is less accurate and slower than recognition of touch-tones. As a result, if the user can remember the complete the digit string of a telephone number, it is typically faster (and maybe easier) to use a touch-tone keypad than to speak the digit string -- at least for most circumstances.
Much more potential for advancement of the telecommunications user interface occurs with "name dialing." Name dialing has the potential to solve a fundamental human memory problem. Specifically, it is very difficult to remember a 10-digit telephone number for every friend, associate, and/or business that that each of us calls (let alone two or three digit strings for a "home number", "work number", "cellular number", etc.). It would be far easier to place calls by speaking simple names, and remembering just the names.
While "name dialing" could go a long way toward solving the fundamental human memory problem that occurs in telecommunications, there are hurdles that can limit the effectiveness of name dialing as well. One of the most important concerns the creation and maintenance of a "personal dialing directory" to support name dialing.
First, the initial creation of the name dialing directory can be inconvenient. Often, this involves the caller speaking a name, then speaking it a second time, then speaking it a third time. Following this, the caller must speak or touch-tone a telephone number to associate a telephone number with the name, and then, perform a transaction for continuing on with the process of adding names or stopping. These steps must then be repeated for every listing in the personal directory.
If one conducts a usability test, it is not difficult to find that, as user interface designers, we can design a user interface where almost all of the participants (or customers) can perform this training procedure. This, however, does not necessarily mean that success has been attained. There are a number of more subtle problems that might still exist with name dialing, including the following:
* First, we have developed a procedure where the customer's first task is to figure out who is worthy of being in his/her personal directory, and who is not worthy. Once this is determined, the caller must take the time to "train" the system for all who are worthy, or train the system for just a subset of those who are worthy (because of time constraints, boredom, etc.). This is the start of inconvenience and problems.
* Secondly, when the caller learns of a new person who is frequently called, the subscriber must have the forethought to stop other tasks, access the "programming mode" of the voice activated dialing service and spend the time to create a new "name" or voice label for later use.
* At a later time, when the subscriber is ready to place a call, s/he must decide whether or not the party to be called is already on the name dialing list. Also, s/he must be able to remember the specific "name" that was stored, e.g., whether the name was "Tony", "Tony Tsoi", "My Director", or something else. Otherwise, the caller will need to recall the procedure for accessing the "programming mode", then use the programming mode to review names in the personal directory.
As a result of all these factors, when a real person tries to deal with the current name-dialing procedure, it is quite easy to do the following:
1. On the first day of service, create a short personal directory.
2. Within a couple of weeks, forget which names were stored and which were not stored in the personal directory.
3. Fail to take time to update the personal directory list as time passes.
4. Use name-dialing only for calls to one or two common places such as "voicemail" that are easily remembered as "on the list."
As we ask ourselves how we can solve these problems and optimize the process, we must also consider the optimal size of the name dialing directory and how it can best be managed.
For example, if we create a name-dialing service, and set the maximum directory size to be 10 names, how does this affect the user interface design process? Does this afford some type of opportunity to create automatic training for the 10 most frequently called numbers? Would limiting directory sizes to 10 names make it easier or more difficult for callers to keep track of who is in their directory and who is not? If a personal directory includes just 10 names, is it really any better than a single-digit touch-tone speed dialing button? There are a host of questions that we can ask about limited-size personal directories of this type.
What if the personal directory size is expanded to 50 names? Has the caller now lost the ability to review all of the names in the directory by speaking a simple command? Is the caller now better prepared or more poorly prepared to know who is in the directory and who is not? How can the caller review names? How can the caller be motivated to appropriately create new names when needed and delete old names that are never used or outdated?
As technology improves, we might also anticipate the question of how to design creation and management of the personal directory if it includes more than 100 listings. In fact, why not assume that we can now have a directory of 10,000 names. How should the user create and maintain the directory now? Shall we ask the user to speak a name for every number that s/he ever calls (during the normal course of call placement)- so that the system can begin creating the ultimate voice dialing directory for the user? Or, will this type of directory be feasible only if some type of text-based creation of entries is permitted, or if some type of master telephone book is created ahead of time? Finally, once again, we can ask how does the caller know who is in the directory and who is not?
One way or another any new user interface for Voice Activated Dialing must be capable of meeting caller's needs, such that callers gain the motivation to use VAD more frequently than it is used with the current generation of Voice Activate Dialing services.