|
VoiceXML: Introduction to Developing
Speech Applications
5-1
Consider a hierarchy of menus in which the first menu contains categories
and the second level of menus contains individual items. Discuss the
advantages and disadvantages of single level menus with many items versus
a two level hierarchy of menus with each menu contain only a small number
of items. Be specific in your discussion by using one of the following
as an example.
A. Presidents of
the United States
There have been
43 U.S. Presidents. Let's suppose we assign about 7 options to each
menu. This suggests a main menu with 7 options and 7 second-level menus
each with about 7 options (some menus will have only 6 options). Suppose
the first menu of a two level hierarchy of menus categorizes U.S. presidents
by the date they tool office. For example, U.S. presidents from 1789
to 1837, from 1837 to 1857, etc. The second level menu lists the presidents
serving during the corresponding time period. After selecting a president,
the caller hears information about the president.
| |
Single level menu |
Hierarchy of two levels of menus |
| Number of options caller hears (worst case) |
43 options |
7 in the main menu + 7 in a secondary menu = 14 options |
| Number of options caller hears (average case) |
43/2 = 22 (on average, caller hears ½ of the options)
|
7/2=4 in the main menu + 7/2=4 in a secondary menu
= 8 |
From the point of
view of the number of options a caller hears, the hierarchy of two levels
is best. But that's not the full picture. If the caller knows the name
of the president, the caller only needs to speak the president's name
for the single level menu case. Here is a further analysis, based on
the type of the caller's query:
| Type of query |
Single level menu |
Hierarchy of two levels of menus |
| 1. If the caller knows the name of the U.S. president
but not the time he served in office. |
+, The caller speaks the name of the U.S. president
to get the associated information. |
- , The caller must listen to a list of time periods
and select the appropriate time period, then listen for the name
of the U.S. president. |
| 2. If the caller knows both the name of the U.S. president
and the time he served in office |
+, The caller speaks the name of the U.S. president
to get the associated information. |
+, The caller first selects the timeframe during which
the president served, and then selects the president. |
| 3. If the caller knows only the time period. |
-, The caller must listen to the information associated
with each president. |
+ The caller first selects the time frame during which
the president served, and then listens to the information about
each of the presidents in that time period. |
| 4. The caller knows neither the president's name or
time period |
-, The caller must listen to the information associated
with each president. |
-, The caller must listen to the information associated
with each president |
The plus (+) indicates
an advantage. The (-) indicates a disadvantage. If most callers have
a query of type 1, then a single level menu is best for the greatest
number of users
B. Kings of England,
similar analysis as above
C. Kings of France,
similar analysis as above
D. Emperors of China,
similar analysis as above
E. Leaders of your
home nation (there must be at least 30 leaders in the list), similar
analysis as above
5-2 Write
VoiceXML code for the following form:
Simplifying assumptions:
- Record the student
name and replay it so the student knows that his voice print has been
recorded. This recording can be validated by a speaker recognition
engine at a later time.
- Use DTMF to capture
the student ID and Course ID because they are more accurate than speech
recognition.
- There is no reason
to ask the caller for today's date because the computer can generate
today's date.
- Rather than attempting
to speech recognize the caller's response to Reason, record the response
and save for later analysis (by both a dictation speech recognition
engine and a human transcriptionist if higher accuracy is required).
a
|
