80 ◾ Mariela Rico et al.
for improving the representation of the real entity semantics. Based on the outputs
of Processes 1 and 2, these features can be detected. Since not all of these features
must be made explicit, answering the following questions could help to identify
those that must be explicit.
◾ Are there any implicit features in the representation of an entity that may be
inferred by a human agent but cannot be inferred by a machine agent? If the
answer is yes, can these features be inferred incorrectly in contexts dierent
from the one considered? If the answer is yes, these features should be made
explicit.
◾ Are there any entities whose representations and/or meanings may change
based on the context in which the entities are considered? If the answer is
yes, are the representations and meanings completely explicit in the ontol-
ogy? If the answer is no, the representations and meanings should be made
explicit.
◾ What are the quality dimensions used to represent a feature? Are they the
same regardless of the context in which the feature is considered? If the
answer is no, are they explicit in the ontology? If the answer is no, these qual-
ity dimensions should be made explicit.
3.5.4 Process 4: Make Features of Each Entity Explicit
After the features and their representation dimensions have been identied, they
must be made explicit. For this purpose, integrating existing and widely accepted
ontologies should be considered. Examples of such ontologies are:
◾ OWL–Time ontology [16] for modeling most of the basic temporal concepts
and relations, i.e., a vocabulary for expressing facts about topological rela-
tions among instants, intervals, and events, along with information about
durations, dates, and times
◾ ISO 3166 Country Codes Ontology (http://www.daml.org/2001/09/coun-
tries/iso-3166-ont) for modeling ocial country names
◾ A portion of an ontology implementing ISO currency codes published in
Standard ISO 4217:2008, such as the PCS ontology for the representation of
currencies and funds (http://www.ifpi.com/pcs/)
◾ A portion of an ontology implementing ISO 80000, the successor of ISO
31, for modeling physical quantities and units of measurement, e.g., the
United Nations Centre for Trade Facilitation and Electronic Business (UN/
CEFACT) (http://www.unece.org/cefact/codesfortrade/codes_index.htm)
When it is not possible to reuse an ontology to improve the representation of a fea-
ture, it is necessary to identify whether the feature is simple or complex. A simple
feature does not exhibit multiple qualities and is associated with a one-dimensional
Toward Semantic Interoperability between Information Systems ◾ 81
representation in human cognition [15], e.g., the weight of an object. us, two
elements should be added to an ontology: (1) a term denoting the representation
dimension and (2) a relation between this term and the term that represents the
simple feature.
A complex feature bears multiple qualities and is associated with a set of integral
dimensions that are separable from all other dimensions [15]. In an integral dimen-
sion, it is not possible to assign a value to an object in one dimension without giving
it a value in another. For instance, color can be represented in terms of the integral
dimensions of hue, saturation, and brightness. By contrast, weight and hue dimen-
sions are said to be separable. Each integral dimension is associated with a simple
feature. To improve the representation of a complex feature, the following elements
should be added to the ontology:
◾ A term representing the set of integral dimensions and a relation between this
term and the term that represents the complex feature
◾ For each integral dimension, a term representing it and a relation between
this term and the term that represents the set of integral dimension
◾ For each term representing an integral dimension, a relation between this
term and the term that represents the corresponding simple feature
In addition, for each term representing a one-dimensional representation or inte-
gral dimension, a term representing the unit of measurement of the dimension and
a relation between these two terms should be added to the ontology. is term
aects the granularity of the dimension but not its structure. For example, a weight
dimension has positive real numbers as values regardless of whether the metric
units are kilograms or tons.
3.5.5 Process 5: Designate Bridge Term for Each Entity
e intended uses of an entity in the context considered should be represented
byterms called bridges because they allow linking dierent meanings and rep-
resentations of the same entity in dierent contexts. ese terms should also be
interpreted as representing contextual features because the intended use depends
on the context in which the entity is considered. us, it is necessary to deter-
mine whether an existing term designates the intended use of each entity; if such
a term is absent, it must be added. Bridge terms should also relate to the elements
that represent the entity whose intended use they represent. An entity may be
represented by a single element or a set of elements. In the former case, a relation
between the single element and the bridge term should be added. In the latter,
the most representative term should be chosen and then a relation between this
term and the bridge term should be added. As the bridge term represents a con-
textual feature, it should also relate to the term that represents its representation
dimension.
82 ◾ Mariela Rico et al.
3.6 Application Example
Suppose there is a collaborative relationship between a packaging industry (sup-
plier) and a dairy industry (customer). Both trading partners must exchange the
information shown in Table3.2 to reach an agreement on a replenishment plan.
e structure and semantics of this information are initially reected in an EBD
ontology as shown in Figure3.9. To improve the representation of entities in this
ontology, the proposed method is applied.
3.6.1 Process 1: Identify Entities and Their Features
e entities whose information must be translated are: (1) trading partners that
assume two roles: supplier and customer; a relevant feature of a trading partner is
its address; (2) a replenishment plan that refers to the agreed plan between the trad-
ing partners; some plan features are the time period during which the plan is valid,
the products to be exchanged, the quantities of products, and periods within the
horizon during which these products are exchanged, among others; (3) the prod-
ucts involved in the replenishment plan (manufactured by the packaging industry
and the packages containing the dairy industry products). Product features include
trademark, type, and size.
Table3.2 Examples of Necessary Information for Devising
Replenishment Plan
Horizon: 6/04–31/05 (Day/Month)
Period
Product
Identification Trademark Type Size Quantity
7/04–13/04 20320101 yy Carton 1000 4400
20320102 2900 2880
20070231 Plastic 196 1600
20070232 250 1800
20320101 zz Carton 1000 2200
20320102 2900 8064
20070232 Plastic 250 1800
20070235 1000 6500
14/04–20/04 … … … … …
Toward Semantic Interoperability between Information Systems ◾ 83
3.6.2 Process 2: Identify Ontology Elements
A trading partner assuming a supplier role is represented by the Agent, Organization,
and Supplier terms, their properties, and the relations between these terms. A
customer is represented by means of the Agent and Organization terms, their
properties, and the relations between them. eir addresses are represented by the
Address term and the hasAddress relation. e replenishment plan and its features
are represented by EBD, EBDItem, and EBDItemsCollection terms, their proper-
ties, and the hasItems, and hasItem relations. EBD refers to documents exchanged
by the trading partners. EBDItemsCollection and EBDItem represent the structures
of the documents. e products and their features are represented by the following
properties of the EBDItem term: PartNumber, ItemName, and ItemDescription.
3.6.3 Process 3: Identify Features That Must Be Explicit
As noted, a relevant feature for a trading partner is its address (represented by the
Address term). According to Smith et al. [26], this term represents a quality exist-
ing in reality and cites a mailing address—not an e-mail address, for example. e
features of a mailing address are street, number assigned to building or entity on
street, oor, apartment, city, postal code, province or state, and country. In the
ontology, some items (oor or apartment, for example) need not be explicit if they
Personnel
Position String*
String*
Integer*
Integer*
Float*
String*
String*
String*
Title
Agent
HomePage
Phone
Email
Fax
hasAddress
hasAddress*
OrganizationName
contactPerson Personnel
Organization
Person
LastName
Address
Country
isa
isa
isa isa isa
hasItems
Address
String*
String*
EBD
EBDHorizon
Comments
Currency
EBDDate
EBDNumber
EBDItemsCollection
EBDItem
ItemDescription
ItemName
Quantity
Price
EBDItem
PartNumber
Period
hasItem*
EBDItemsCollection
hasItem
Name
String*
String*
String*
String*
Integer*
Organization
Supplier
Shipper
hasItems*
shipTo*
String*
String*
String*
String*
suppliedBy* shippedBy*
String*
Zip String*
String*
FirstName String*
City String*
Street String*
State String*
Instance*
Instance*
Instance*
Instance*
Instance*
Instance*
contactPerson*
Supplier Shipper
suppliedBy
shipTo
shippedBy Instance*
Figure 3.9 Original EBD ontology.
84 ◾ Mariela Rico et al.
do not prevent a correct interpretation of the entity in any context in which the
entity is considered. A reason for making these features explicit could be to achieve
a more complete representation of the entity.
e time period during which the replenishment plan is valid is represented by
the EBDHorizon property, whose data type is string. is representation does not
reveal whether the horizon is expressed as an amount of time or an interval, and
thus does not satisfy the minimal encoding bias criterion [14]. Based on Table3.2,
the representation should make the feature explicit as a calendar interval since the
horizon is a quantity of time and also a location on the time line. A similar analysis
can be made for periods within the horizon (represented by the Period property
of EBDItem).
Another feature of a replenishment plan is the quantity ordered of each product.
is feature is represented by the Quantity property (integer data type). At this point,
some questions arise. What unit of measurement expresses this quantity? e answer
could cite units of products or units of product packs, for example. Will any informa-
tion system that deals with such information make a correct interpretation of it? If
any misunderstanding is possible, the unit of measurement should be made explicit.
Products are represented by the following properties of the EBDItem term:
PartNumber (integer data type), ItemName, and ItemDescription (both of
string data type). PartNumber represents the product identication in Table3.2.
ItemName and ItemDescription do not represent, at least at rst glance, any of the
other product features. ItemDescription can appear in the natural language of the
product. However, the ItemName property should be replaced by three terms that
represent trademark, type, and size features.
3.6.4 Process 4: Make Features of Each Entity Explicit
To make a feature explicit, it is important to distinguish when that feature is an
entity and when it is not. For example, the Country feature of an address may be
considered an entity and be represented by a Country term instead of a property. In
this case, a formal relation [15] that joins the Address and Country terms is needed.
By contrast, the feature Floor is existentially dependent on the address in the context
under consideration, in which case a property is more appropriate to represent it.
Taking the OWL–Time ontology [16] into account, the feature horizon of
the Replenishment Plan entity should be represented by means of a dierent term
derived from the CalendarClockInterval term and linked to the EBD term by a
formal relation. e same treatment applies to the periods within the horizon.
Since the quantity ordered of each product is a simple feature, four elements
should be added to the ontology: (1) a QuantityDimension term denoting the
representation dimension, (2) a UnitOfMeasure term representing the measure-
ment unit of the dimension, (3) a relation between these two terms, and (4) a
relation between QuantityDimension and the term that represents the simple fea-
ture. However, quantity is represented by a property, not by a term. Although it
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.