![](bg2.png)
Functions (F), Events(E) and Connectors (and, xor,
or). Natural language expressions describe the inner
process semantics of functions and events.
Depending on the sequence of words and on its
semantics, different meanings may arise. For
example, “Define Software Requirements with
Customer” has a different meaning than “Define
Software Requirements for Customer”. The former
expression means that software requirements are
defined in cooperate manner with a customer; the
latter one indicates the customer as output target of
the performed activity.
Semantic annotation of EPC models constitutes
an appropriate solution to tackle the problem of
different meanings but raises the problems to
determine the process semantics of lexical terms
used in natural language expressions and to link
them to reference ontology instances. The process
semantics of lexical terms represents process tasks
executed on process objects, for instance.
Linking lexical terms to reference ontology
instances requires the availability of corresponding
reference ontology instances. In case of non-
corresponding instances, semantic annotation also
entails to populate a reference ontology with new
instances derived from lexical terms.
Natural language expressions follow naming
conventions or standards that represent guidelines
for naming EPC functions/events (Schütte, 1998). If
a naming convention is used, it is clear what a
lexical term expresses.
For example, a typical suggested naming
convention for an EPC function is the template Task
followed by a Process Object that specifies the
semantics of natural language expressions such as
“Define Project Plan”; “Define” has the semantics
of a task, the lexical term “Project Plan” indicates
the semantics of a process object. We analyzed
about 5,000 EPC functions/events in engineering
domains. We experienced that the suggested
recommendations do not fully cover the semantics of
used natural language expressions. These
investigations addionally triggers the neccessity of a
formal notation that enables to express or specify a
guideline. Formalized guidelines enable verifying
natural language expressions against a set of
predefined conventions.
Based on our investigations for clarifying the
semantics of natural language expressions used for
naming EPC functions/events, we introduce a set of
guidelines that extend traditional naming
conventions. The introduced guidelines are
expressed through semantic pattern descriptions.
Semantic pattern descriptions represent semantic
templates that bridge the gap between informal and
formal representation. Formal representation refers
to concepts specified by a reference ontology.
Semantic pattern descriptions are either defined for
EPC functions or for EPC events. A semantic pattern
description has a pattern template, that is represented
in the form Function(Context)[Task; Process
Object; Parameter] for an EPC function.
Context is the name of an engineering domain
the analyzed EPC function is associated with, the
concept Task represents the activity being performed
on an instance of the concept Process Object, the
concept Parameter captures an instance of a Process
Object optionally for executing an EPC function. To
represent the semantics of a function consists in
instantiating a semantic pattern template by binding
a lexical term to the variables Context, Task, Process
Object and Parameter. For example, the instantiated
semantic pattern template Function(Software) [Task:
”Define”; Process Object: ”Quality Goal”] defines
the process semantics of the EPC function “Define
Quality Goal” in the context “Software”.
Same or similar knowledge in EPC
functions/events can be expressed in alternative
ways due to the freedom of modeling (e.g. usage of
synonyms, abbreviates etc.). Consequently,
alternative natural language expressions may refer to
same process semantics. For example, the EPC
functions “Define SW Requirements” and “Define
Requirements for Software” refer to the same
process semantics.
For that reason each semantic pattern description
defines a set of lexical structures and analysis rules.
A lexical structure is defined by a sequence of word
classes (e.g. Verb, Noun, Preposition). The lexical
structure LS:= [Verb
Task
] [NounGroup
ProcessObject
]
captures a natural language expression such as
Define [Verb
Task
] Quality Goal
[NounGroup
ProcessObject
]. A lexical structure
instantiates a semantic pattern template by applying
analysis rules. They define how to map lexical
structures onto the reference ontology concepts
specified in a semantic pattern template.
An instance of a reference ontology concept has
an unique identifier which may have several textual
counterparts. For example, a process object with the
OID=123 has the textual counterparts {“Software
Requirements”, “SW Requirements”,
“Requirements for Software”}.
To separate meaning from its lexical
representation the reference ontology is split into
two layers. Layer one represents the lexical
knowledge base that captures commonsense
vocabulary used in natural language expressions;
SEMANTIC ANNOTATION OF EPC MODELS IN ENGINEERING DOMAINS BY EMPLOYING SEMANTIC
PATTERNS
107