SSQ-OWL Ontology Specification

advertisement
SSQ-OWL Ontology Specification
For the Morpheus project
Revision History
Serial No
Date
Description
By
1
Sept 09, 2009
Initial draft
Clint P. George
1. Introduction
The purpose of this document is to provide an overview of the OWL representation of the Semi
Structured Queries (SSQ) generated using the Query Resolution Recorder (QRR) [1].
The basic idea is to represent the real world objects using the OWL ontology classes and the objects’
relationships by the OWL ontology properties. We assume that we have a SSQ store, which contains a
rich data set of fully qualified SSQs i.e. SSQs with user input instances (based on the contexts) are
correctly mapped to the corresponding classes, and it is pointed to the desired query result. We can
form the basic building blocks of the ontology i.e. specific classes and properties from these fully
qualified SSQs. Our primary focus is on reusing the defined ontological structures, if they are available at
present. In addition, derive new ontological structures from these defined classes and properties, if we
cannot find the exact match. In fact, we may do some matching for those SSQ features to the existing
classes or properties (or we may define new classes on the fly).
2. Proposed SSQ representation in OWL structure
The whole SSQ can be looked as a collection of OWL ontology instances (individuals) of the OWL classes
and the ontology properties are as the relationships between these instances. A graphical
representation of the generated SSQ from the QRR is given below. The rectangles represents the
contexts in the SSQ, the classes are represented by rounded rectangles, and the oval shapes represents
the corresponding instances of the classes i.e. user defined inputs. In the case of a fully qualified SSQ, it
will be pointed to the specific classes [1].
SSQ
What_IN
Class 1
(E.g. Band)
Rolling Stone
When_IN
Where_IN
Class 2
(E.g. Date)
Friday, Sept 04, 2009
Where_OUT
Class 3
(E.g. Place)
Chicago
A sample SSQ tree
Class 4
(E.g. Building)
Stadium
The basic design of the OWL SSQ representation is shown below. The rounded rectangles represent
OWL classes; the rectangles represent the OWL properties; and the oval shapes are for the OWL
individuals.
SSQ Query Class
What_IN
When_IN
Band
Date
Rolling Stone
Friday, Sept 04, 2009
Where_IN
Where_OUT
Location
Chicago
Building
Stadium
A sample SSQ-OWL representation
The following OWL classes and properties represent the basic stubs of an OWL ontology, which can be
used for representing a complete SSQ. Our aim is to form an OWL ontology prototype for the SSQ at
first, and then try to generalize the SSQ structure.
2.1.
SSQ Query Class
This OWL ontology class definition represents the OWL SSQ ontology tree’s root i.e. the node to
represent the user defined query and its properties. The ontological properties can be defined as follows



rdfs:label (mandatory): class label which has the ‘SSQ Query Class’
hasRealm (mandatory): The realm of the user defined query, which is a user input (mandatory).
It can be represented as an object property [2] in the ontological structure.
o Domain: the most basic ontology class ‘Thing’
o Range (mandatory): the specific Realm classes such as Sports , Music etc
hasQuery (mandatory): The original user defined search query in simple English. It can be
defined as a data property [2].
2.2.
Contexts as properties
This represents the different contexts in the structure SSQ such as what, where, when, and who.
Currently, we assume that these contexts are static for all the SSQs generated by the QRR, to make it
simple in design. One way of thinking is that consider these contexts are as OWL ontological properties.
Its basic attributes can be defined as follows



rdfs:label(mandatory): This has values as the contexts such as who, where, when, and what.
Domain (mandatory): This is always fixed as the QueryClass.
Range (mandatory): We can start with most basic class definition ‘Thing’, then we can further
narrow down to specific classes (these classes can be chosen by the user during the creation of
SSQ query) e.g. Event, band, building, etc
2.3.
OWL Ontology property and class definitions
Data properties
A sample set of data properties for the OWL ontology prototype is described below. (I haven’t gone
much into the OWL / RDF syntax). A brief description of the data format also mentioned here.
hasDate: this date property which can be formed from different date formats (date)
hasAuthor, hasCreator, hasPresenter, hasSponsor: these properties shares same features (from string)
hasCounty – specifies the county (string)
hasCountry – specifies the country (string)
hasStreet – specifies the street (string)
hasState – specifies the state (string)
hasZipCode – specifies the zip code (integer)
hasName – name property (string )
hasBuildingNumber – building number (string)
hasQuery – represents the user defined query (string)
hasTime – represents the time
Class definitions
The OWL ontology class represents the real world objects. All classes are derived from the basis class
‘Thing’. It should have generic properties like ‘label’ and ‘inherited’ type information. The specific class
properties are defined using the ontological data properties and object properties. A sample set of the
useful class definitions are shown below.
Location
Building
EventDate
Place (equivalent to Location class, we can use the equivalentClass constraint in OWL for this)
Performance
Person
SportsTeam
SSQ (entire SSQ can be defined as a class combining its classes and properties)
Ontology Object properties:
hasLocation: specifies the class has a Location (a class defined separately)



rdfs:label(mandatory)
domain : thing
range : Location (class)
hasPerformer: specifies the relationship between a performance and the person who performs



rdfs:label(mandatory)
domain Performance (class)
range: Person (class)
3. References
1. Generalizing web pages for deep web information extraction by Christan Earl Grant, Joseph N
Wilson, Peter Dobbins University of Florida
2. http://www.w3.org/TR/owl-guide
Download