TITLE:		AUTOMATIC ABBREVIATION GENERATION
AUTHORS:	Gregg M. Stum 
COMMENTS:	UD CIS Technical Report No. 92-01

ABSTRACT:
A common phenomenon in the English language is the substitution of
abbreviations for words and phrases. Investigations of this behavior
reveal it to be a systematic rule-based process where people have
individual preferences for the rules they use in forming
abbreviations. Automatic Abbreviation Generation (AAG) describes how a
computing system that expands the abbreviations entered by a user can
be made to reflect its user's preferences. AAG uses a rule-based
formal model consisting of a set of well-defined abbreviation rules and
a strategy that specifies rule preferences and orderings for
abbreviations and expansions. An important use of such a system is in
the field of Rehabilitation Engineering known as Augmentative and
Alternative Communication. This subfield is concerned with
facilitating the communication of individuals having a speech
impairment, and commonly uses a computerized abbreviation system as
one such tool. The traditional approach taken is to define, for each
item in a given vocabulary, one specific abbreviation. This fixed
scheme is then memorized by the user in employing a traditional
expander for their communication. These two requirements impose
unnecessary burdens on those defining and memorizing this scheme. The
goal of AAG is to reduce or eliminate these costs with an abbreviation
system that uses its model. A Scheme Constructor automatically defines
a fixed scheme tailored to the user's preferences, if that is
desired. This scheme is then used by a traditional
expander. Alternatively, a Flexible Expander using a flexible scheme
in which an abbreviation can be associated with more than one
abbreviation, eliminates the need for a fixed scheme. To the Expander,
the user gives a reasonably well-formed abbreviation of the vocabulary
item they intend as they are typing, and the Expander, applying the
model to the flexible scheme, identifies the most likely
expansion(s). The Expander can also use information from another
process, such as a syntactic parser, in rating these possible
expansions.