Flexible Abbreviation Expansion

Gregg M. Stum, Patrick W. Demasco
Applied Science and Engineering Laboratories
University of Delaware / Alfred I. duPont Institute
Wilmington, Delaware USA

(c) 1992 RESNA Press. Reprinted with permission.

Abstract

This project is part of an augmentative and alternative communication
project directed toward reducing the burdens imposed on both
clinicians and users by traditional fixed abbreviation
expanders. These burdens include the requirement to devise and
memorize good unique abbreviations for the items that are to be
abbreviated. This is achieved by the application of a set of formal
abbreviation rules across the user's vocabulary in order to anticipate
the abbreviations that the user will enter. A flexible expander is
currently under development for the Microsoft Windows operating
environment. Being independent of any particular application this
expander offers consistent access to abbreviation expansion from
within any context.

Background

Ever since [KV82], abbreviation expansion has been identified as an
important technique to reduce the typing load of individuals for whom
typing is their primary form of communication, or for whom typing is a
significant physical task. It is important for these individuals to
produce text as quickly as possible and with the least possible
effort.

Abbreviation systems developed since then rely on the assignment of a
specific unique abbreviation to each word that the user wishes to
abbreviate. The user then enters this abbreviation and the system
replaces it with the associated expansion. This approach is easy to
implement and operates efficiently.

Unfortunately fixed abbreviation expansion imposes numerous
unnecessary burdens. The user is required to memorize all of the
abbreviations that the system recognizes. Someone must manually devise
and enter any abbreviations not already in the system. The set of
abbreviations is probably not optimized for the cognitive skills of
the user.

In Word Compansion [DLM89] there is no predefined assignment of
abbreviations to expansions. Rather, the system performs a number of
transformations on an abbreviation in attempting to infer the intended
expansion. In this way the user simply enters a reasonable
abbreviation for the expansion they intend, rather than recall one
that they had memorized.

Flexible Abbreviation Expansion is thus a hybrid of these two
approaches. It uses knowledge of how people go about abbreviating in
order to anticipate the most likely abbreviations that a user will
enter for any given expansion. This allows it to operate as
efficiently as traditional fixed expanders while bringing the power
and flexibility of Word Compansion to the user.

Statement of the Problem

Current abbreviation expanders succeed primarily in decreasing a
user's typing load. Ideally this also translates into increasing the
rate at which the user produces text. While just this is of tremendous
benefit, there is still room to significantly improve the user's
typing capability.

The first improvement is to free the user from memorizing specific
abbreviation assignments. Current systems do decrease the amount of
memorization required of the user by utilizing mnemonic coding
approaches to make the abbreviation assignments easier to
remember. Typically these take the form of: <2> = 'to', <b> = 'be',
<4>= 'for', and <bc> = 'because', and so forth. [V87][R91]. While such
approaches work very well for the most common and frequently used
items, the fact that they are still unique assignments means that once
an abbreviation is used for one item, it is unavailable for any other
item, no matter how good or appropriate it might be in that case as
well.

This leads to the second improvement being covering the user's broad
vocabulary needs.  Generally expanders come with a set of default
assignments for common vocabulary items such as function words,
articles, and frequent verbs, nouns, and phrases. These do comprise
the majority of the user's typing needs; they aren't its
totality. There is a vast reservoir of subject-specific vocabulary that
the user taps on demand. To abbreviate any of these items with a
current expander, someone must have manually assigned an abbreviation
to it. Since any of these items are used infrequently, it is not
 practical to make the effort in coming up with abbreviations for each
of these that are both unique as well as easy for the user to
remember.

A solution to this need illustrates the third improvement, tailoring
the set of abbreviations to the individual user. To overcome the
demands of manually making the abbreviation assignments, one could
consider using sets of predefined abbreviations; one for 'chemistry',
one for 'basketweaving', one for 'golf', and so one. Two problems with
this solution are not being able to insure the uniqueness of any
particular abbreviation, and that the abbreviation assignments do not
in general reflect the preferences of its user.

Another improvement is making abbreviation expansion available to the
user within any typing context in a consistent manner. Currently an
expander is tied to a specific application.  If a user has an expander
for a speech aid it is separate and distinct from one they would use
with a word processor. Not only must the user how to operate each of
these, they must also remember a different set of abbreviations with
each, or at least remember which and how the abbreviations differ.

Approach

Flexible Abbreviation Expansion makes all of these benefits available
to the user. It achieves this by using a set of abbreviations for a
given expansion, rather than a unique assignment.  This set is not
entered manually, rather it is generated automatically given the set
of items that is to be abbreviated.

An abbreviation strategy directs the generation of the set of
abbreviations associated with a vocabulary. The strategy consists of a
set of abbreviation rules that are partitioned into subsets related to
word length. That is, for a word of a specific length, there is a set
of rules to be used in producing a set of abbreviations for that
word. This set is ordered according to the user's preference in
applying the rules. This arrangement is based on the observation that
preferences vary among individuals based on word length [S91].

An abbreviation rule is a formal description of the relationship
between an expansion and an abbreviation [S91]. A simple example is
truncation, where the abbreviation is some number of the leftmost
letters from the word. Another example is salient-letter, the first
letter of each syllable in a multiple-syllable word.

Once the set of abbreviations is generated, they are entered into an
abbreviation-expansion table. Each row of this table represents all of
the expansions , for any particular abbreviation. These expansions are
ordered by the preference for the rule used to produce the
abbreviation. That is, the most preferred expansion is the one for
which the abbreviation was produced by the most preferred rule. When
two expansions have the same preference, they are arranged by
increasing length. This represents the preference of having shorter
abbreviations associated with shorter expansions.

The entire set of items that the user wishes to abbreviate is divided
into separate vocabularies, where each is typically a set of perhaps
as many two hundred related words.  The user can then select a small
number of these vocabularies as being active at any given time. This
allows the expansion to focus on the particular subject about which
the user is writing, rather than the as many as twenty thousand words
that the user might know in general. Each such vocabulary has its own
table. This is done for the convenience of the user and makes the
process of expansion more efficient.

The user has a choice in how abbreviations are expanded. With one
method the abbreviation is replaced with its most likely
expansion. Should this not be the one that the user intended, they
press the Space Bar (or any designated < abbreviate>key) again, and
the next most likely expansion appears. The user repeats this, cycling
through the entire list at will, until the desired expansion appears,
at which point they resume typing with a non-<abbreviate> key. With
the second method, should the user reject the most likely expansion, a
list is presented from which the user can select the desired expansion
directly. These expansion methods are actually not distinct: the user
can press the Space Bar to scroll through the list, and can directly
select the expansion in the first method. The difference between the
two is essentially whether or not the list appears when the most
likely expansion is rejected. This is provided as a feature for those
users to whom the seeing the list is not of benefit.

One important observation about flexible expansion is that it does not
preclude the traditional fixed expansion. The table produced for any
vocabulary can be edited to add, change, or remove both abbreviations
and expansions. In this way fixed abbreviation assignments can still
be used by simply having only one possible expansion for an
abbreviation. Thus the expander can provide the benefits of fixed
expansion for the most frequently used items.

Implications

By removing the uniqueness constraint on the abbreviation-expansion
relationship, the user no longer has specific assignments to
memorize. Rather, the user can simply construct some reasonable
abbreviation for the expansion item they wish to enter. With a
suitable defined strategy the expander will have anticipated this
abbreviation and can present the desired expansion. In the worst case
the desired expansion would be one of the alternative expansions the
system identified.

Using the strategy the system can tailor the set of abbreviations for
a given set of items to any particular user. More importantly this set
of abbreviations is automatically generated and does not have to be
manually entered. Thus abbreviation expansion is made available across
the user's vocabulary in a practical manner. Being independent of
specific application, flexible abbreviation expansion gives the user
consistent and transparent access across all applications. The user
can make use of the same set of abbreviations in the same manner from
within any application and gain the same benefit.

Discussion

A flexible abbreviation expander is being developed and tested for the
Microsoft Windows environment. This typing aid gives users easy and
consistent access to a wide variety of software ranging from popular
off-the-shelf word-processing to specialized speech communication. In
addition, this is one more tool in making workplace accommodation more
complete.

Future enhancements of Flexible Abbreviation Expansion include
incorporating adaptiveness to a particular user. Generally this means
fine-tuning the operation of the = expander based on the selection
patterns of the user. For example, with a given abbreviation the user
may actually prefer an expansion that the strategy indicates as being
the third most preferred. An adaptive expander would recognize this
and override the strategy and reflect the user's preference with this
abbreviation. In another case, the system could also observe more
wide-scale tendencies and actually modify the Strategy itself. Of
course, some users will be more sensitive to this fine-tuning than
others. So the use and extent of adaptiveness would be entirely at the
discretion and desire of each individual.

Acknowledgments

This work is supported by SBIR Grant Number RN91076101 from the
Department of Education to GMS Systems for the development of the
Windows application. Additional support is provided by Grant Number
H133E80015 from the National Institute on Disability and
Rehabilitation Research, and by the Nemours Foundation

Microsoft Windows is a trademark of Microsoft Corporation.

References

[DLM89] Demasco, P., Lillard, M., and McCoy, K. (1989). Word
Compansion: allowing dynamic word abbreviations. Proceedings of the
RESNA 12th Annual Conference,282- 283.

[KV82] Kelso, D.P., and Vanderheiden, G.C. (1982). Ten-Branch
Abbreviation Expansion for Greater Efficiency in Augmentative
Communication Systems. {RESNA} 5th Annual Conference on Rehabilitation
Engineering, 3.

[S91] Stum, G.M. (1991). Automatic Abbreviation Generation. Master' s
Thesis, U of Delaware.

[SDM91] Stum, GM., Demasco, P.W., and McCoy, K.F. (1991). Automatic
Abbreviation Generation. Proceedings of the RESNA 14th Annual
Conference, 97-99.

[VK87] Vanderheiden, G.C., and Kelso, D.P. (1987). Comparative
Analysis of Fixed- Vocabulary Acceleration Techniques. AAC
Augmentative and Alternative Communication, vol.3, no. 4, 196-206.

[R91] Ray, S. (1991). The Function of Function Words and Function
Units in Augmentative Communication. Proceedings of the RESNA 14th
Annual Conference, 94- 96.

Contact

Gregg M. Stum 
Applied Science and Engineering Laboratories 
A. I. duPont Institute 
PO Box 269 
Wilmington, DE 19899 
Email: stum@asel.udel.edu