CS6010 Social Network Analysis Unit II

CS6010 - Social Network Analysis
Unit II – Modelling, Aggregating and
Knowledge Representation
Kaviya.P
AP/IT
Kamaraj College of Engineering & Technology
1

Unit II – Modelling, Aggregating and
Knowledge Representation
Ontology and their role in the Semantic Web: Ontology-based
knowledge Representation - Ontology languages for the Semantic Web:
Resource Description Framework - Web Ontology Language -
Modelling and aggregating social network data: State-of-the-art in
network data representation - Ontological representation of social
individuals - Ontological representation of social relationships -
Aggregating and reasoning with social network data - Advanced
representations.
2

Ontology and their role in the
Semantic Web
3

Ontology-based knowledge Representation
• The idea of the Semantic Web is to extend unstructured information with
machine processable descriptions of the meaning (semantics) of information
and to provide missing background knowledge where required.
• An ontology, is a shared, formal conceptualization of a domain, i.e. a
description of concepts and their relationships.
• Ontologies are domain models with two special characteristics:
– Ontologies are expressed in formal languages with a well-defined
semantics.
– Ontologies build upon a shared understanding within a community.
4

• RDF and OWL are the languages most commonly used on the Semantic
Web .
• The most simple structures are glossaries or controlled vocabularies.
• Thesauri are richer structures in that they describe a hierarchy between
concepts and typically also allow to describe related terms and aliases.
• An ontology at least contains a hierarchy between the concepts (subclass
relationships).
• Weaker models with no explicit hierarchies, such as the folksonomies.
• Folksonomies contain additional information about the social context of
tags.
5

Types of Ontology
• Lightweight ontology: Make a distinction between classes, instances and
properties, but contain minimal descriptions of them.
• Heavyweight ontologies: allow to describe more precisely how classes are
composed of other classes, and provide a richer set of constructs to
constrain how properties can be applied to classes.
• Complex knowledge bases that use the full expressivity of first order logic
(FOL) to define to a great detail the kind of instances a concept may have
and in which cases two instances may be related using a certain
relationship.
6

7

Ontology Languages for the
Semantic Web
8

Resource Description Framework (RDF) & RDF Schema
• RDF is to describe resources on the World Wide Web.
• RDF is domain-independent and can be used to model both real world objects and
information resources.
• There are two kinds of primitives in RDF: resources and literals (character
sequences).
• Resources are either identified by a URI or left blank.
• Literals are strings (character literals) with optional language and data type
identifiers.
• Expressions are formed by making statements (triples) of the form (subject,
predicate, object). The subject of a statement must be a resource (blank or with a
URI), the predicate must be a URI and the object can be either kind of resource or
a literal.
9

• The following brief RDF document describes a person named Rembrandt
@prefix rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns> .
@prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#label> .
@prefix foaf: <http://xmlns.com/foaf/0.1/> .
@prefix example: <http://www.example.org/> .
example:Rembrandt rdf:type foaf:Person .
example:Saskia rdf:type foaf:Person .
example:Rembrandt foaf:name "Rembrandt" .
example:Rembrandt foaf:mbox <mailto:rembrandt@example.org> .
example:Rembrandt foaf:knows example:Saskia .
example:Saskia foaf:name "Saskia" .
10

• In this visualization the nodes are the subjects and objects of statements, labelled
with the URI or literal or left blank, and the edges connect the subjects and
objects of statements and are labelled with the URI of the property.
11

• The RDF language provides the basic term to assign a type to a resource (rdf :type)
and to declare a resource as a property.
• It also provides features to describe collections of instances and to make statements
about statements (reification).
• RDF Schema is a simple extension of RDF defining a modelling vocabulary with
notions of classes and subclasses. Classes and properties can be connected by
specifying the domain and range of properties.
12

FOAF Ontology
• The first group of statements describe the class Person.
• The type of the resource is specified as owl:Class, we give a label, name a superclass
and state that this class is disjoint from the class of documents.
• Then describe the foaf :knows and foaf :name properties we have used above, specifying
their type label, domain, range and a superproperty in the case of foaf :name.
13

SPARQL: querying RDF sources across the Web
• SPARQL, establishing a standard query language and a protocol for interacting
with RDF sources.
• The query language captures the common set of features of the existing RDF query
languages.
• The protocol prescribes the way a software application would query a remote
ontology store across the Internet.
14

The Web Ontology Language (OWL)
• OWL was designed to add the constructs of Description Logics (DL) to RDF,
significantly extending the expressiveness of RDF Schema both in characterizing
classes and properties.
• Description Logics are a set of Knowledge Representation languages with formal
semantics based on their mapping to First Order Logic (FOL).
• The Web Ontology Language is a set of three languages with increasing
expressiveness: OWL Lite, OWL DL and OWL Full.
• These languages are extensions of each other (OWLLite ⊆ OWLDL ⊆ OWLFull)
both syntactically and semantically.
• Every OWL Lite document is a valid OWL DL document and has the same
semantics when considered as an OWL DL document.
15

• Languages of the OWL family would be an extension of RDF(S) in the same
sense, this is only true for OWL Full, the most expressive of the family
(RDF(S) ⊆ OWLFull).
• OWL DL was the original target of standardization and it is a direct mapping to an
expressive Description Logic.
• OWL DL is not allowed to extend constructs of the language, i.e. the concepts in
the RDF, RDF Schema and OWL namespaces.
• OWL Full is a “limitless” OWL DL: every RDF ontology is also a valid OWL Full
ontology and has the same semantics when considered as an OWL Full document.
16

• Unique features of RDF/OWL
– OWL DL provides more primitives than UML such as the disjointness, union,
intersection and equivalence of classes.
– OWL allows to describe defined classes.
– RDF/OWL both treat properties as first class citizens of the language.
Properties are global: they do not belong to any class.
– Properties can be defined as subproperties of other properties.
– Classes can be treated as instances, allowing for meta-modelling.
– RDF reification is more flexible than the association class mechanism of UML.
– All non-blank RDF resources are identified with a URI, UML classes,
instances, attributes etc. do not have such an ID.
• Unique features of UML
– UML has the notion of relationship roles, which is not present in RDF/OWL.
– UML allows n-ary relations, which are not part of RDF.
– Aggregation and composition relations are available in UML.
– UML makes a distinction between attributes and associations.
17

Modelling and Aggregating Social
Network Data
18

State-of-the-art in Network Data Representation
• The most common kind of social network data can be modelled by a graph where
the nodes represent individuals and the edges represent binary social relationships.
• Matrices for attribute data.
• A number of formats exist for serializing such graphs and attribute data in machine-
processable electronic documents.
– Text-based formats: Pajek and UCINET
– The two formats are incompatible. (UCINET has the ability to read and write
the .net format of Pajek, but not vice versa).
– Social sciences often represent their data initially using Microsoft Excel
spreadsheets, which can be exported in the simple CSV (Comma Separated
Values) format.
– dot /GraphML (XML for graphs)
19

20

• GraphML files can be edited, stored, queried, transformed etc. using
generic XML tools.
• Graph representations focus on the graph structure, which is the primary
input to network analysis and visualization.
• Attribute data when entered electronic form is typically stored separately
from network data in Excel sheets, databases or SPSS tables.
• One of the common reasons to use multiple data sources is to perform
triangulation.
– use a variety of data sources and/or methods of analysis to verify the
same conclusion.
• We need to be able to recognize matching instances in the different data
sources and merge the records before we can proceed with the analysis.
• We need a representation that allows to capture and compare the identity of
instances and relationships.
• Maintaining the identity of individuals and relationships is also crucial.
21

Ontological Representation of
Social Individuals
22

Ontological Representation of Social Individuals
• The Friend-of-a-Friend (FOAF) ontology is an OWL based format for
representing personal information and an individual’s social network.
• FOAF started as an experimentation with Semantic Web technology.
• The idea of FOAF was to provide a machine processable format for
representing the kind of information that made the original Web
successful, namely the kind of personal information described in
homepages of individuals.
• FOAF has a vocabulary for describing personal attribute information
typically found on homepages such as name and email address of the
individual, projects, interests, links to work and school homepage etc.
23

• FOAF became the center point of interest in 2003 with the spread of Social Networking
Services such Friendster, Orkut, LinkedIn etc.
• Drawbacks to centralized social networking services
– Information is under the control of the database owner.
– Centralized systems do not allow users to control the information they provide on
their own terms.
• FOAF profiles are created and controlled by the individual user and shared in a
distributed fashion.
• FOAF profiles are typically posted on the personal website of the user and linked from
the user’s homepage with the HTML META tag.
• An advantage of FOAF in terms of sharing FOAF data is the relative stability of the
ontology. The number of FOAF users means that the maintainers of the ontology are
obliged to keep the vocabulary and its semantics stable.
24

25

• FOAF has a rich ontology for characterizing individuals especially with respect to their
online presence, but it is rather poor as a vocabulary for describing relationships.
• One may define in RDF a relationship called example:supervises between a
example:Teacher and a example:Student, where supervises is a subPropertyOf
“knows” and Teacher and Student are subclasses of foaf :Person.
• The SIOC (Semantically Enabled Online Communities) project aims at connecting
discussions across various types of fora (Usenet, discussion boards, blogs, mailing lists
etc.) by exposing the postings according to a shared ontology.
• The key concepts of this ontology are the sioc:User account that is used to create a
sioc:Post, which is part of a sioc:Forum at a certain sioc:Site. A sioc:User is not a
subclass of foaf :Person (as a person may have multiple accounts), but related to the
description of a person using the sioc:account of property.
26

Ontological Representation of
Social Relationships
27

Ontological Representation of Social Relationships
• Ontological representations of social networks such as FOAF need to be extended with a
framework for modelling and characterizing social relationships for two principle reasons:
– To support the automated integration of social information on a semantical basis
– To capture established concepts in Social Network Analysis.
• Characteristics of social relationships
– Sign: A relationship can represent both positive and negative attitudes such as like or hate.
– Strength: Tie strength itself is a complex construct of several characteristics of social
relations. Tie strength lists the following: Frequency/frequent contact, Reciprocity,
Trust/enforceabletrust, Complementarity.
– Provenance: A social relationship may be viewed differently by the individual participants
of the relationship, sometimes even to the degree that the tie is unreciprocated.
– Relationship history: Social relationships come into existence by some event involving
two individuals.
– Relationship roles: A social relationship may have a number of social roles associated with
it, which we call relationship roles. For example, in a student/professor relationship. 28

• For example, Orkut allows to describe the strength of friendship relations on a 5-point
scale from “haven’t met” to “best friend”, while other sites may choose other scales or
terms.
• Ideally, all users of all these services would agree to a single shared typology of social
relations and shared characterizations of relations.
Conceptual Model
• Social relations could be represented as n-ary predicates; however, n-ary relations are
not supported directly by the RDF/OWL languages.
• To deal n-ary relations we employ the technique that is known as reification. we
represent the relation as a class, whose instances are concrete relations of that type.
• RDF has a reified representation of statements: the rdf :Statement object represents the
class of statements.
29

• This class has three properties that correspond to the components of a statement, namely rdf
:subject, rdf :predicate, rdf :object.
• Relationships become subclasses of the rdf :Statement class.
• Relationship class is related to a general Parameter class by the hasParameter relationship.
Relationship types such as Friendship are subclasses of the Relationship class, while their
parameters (such as strength or frequency) are subtypes of the Parameter class.
• The two alternatives differ in the representation of parameters.
– The first scheme borrows from the design of OWL-S for representing service parameters.
– The second alternative differs in that the “native” method of RDF is used for representing
parameters.
• Cognitive structuring, works by applying the generic pattern we associate with such a
relationship to the actual state-of-affairs we observe.
• For example, a student/professor relationship at the Free University of Amsterdam (and the
attached role of student and professor) is defined by the social context of the university and this
kind of relationship may not be recognizable outside of the university. 30

31

• The Descriptions and Situations ontology design pattern that provides a model of context and
allows to clearly delineate these two layers of representation (Figure 5.5).
32

• D&S is a generic pattern for modelling non-physical objects whose intended meaning results
from statements.
• D&S is an ontology-design pattern used as a template for creating domain ontologies in
complex areas.
• D & S has been successfully applied in a wide range of real-life ontology engineering
projects from representing Service Level Agreements (SLAs) to the descriptions of Web
Services.
33

Aggregating and Reasoning with
Social Network Data
34

Aggregating and Reasoning with Social Network Data
• Our first step is to convert some data sets in traditional formats (relational databases,
Excel sheets, XML files etc.) into an RDF-based syntax, which allows to store the
data in an ontology store and manipulate it with ontology-based tools.
• Apply ontology mapping to unify our data on the schema level by mapping classes
(types) and properties from different schemas to a shared ontology such as FOAF.
• It includes set of process
– Assign identifiers to resources
– On the notation of equality
– Determining equality
– Reasoning with instance
– Evaluating Smushing
35

Representing Identity
• The primary mechanism is the assignment of URIs to resources. Every resource,
except blank nodes is identified by a URI.
• For example, in the publishing world a number of identifier schemes are in use.
Standard schemes such as ISBN and ISSN numbers for books and periodicals as well
as DOIs (Digital Object Identifiers) are widely adopted, while each publisher and
online repository also maintains its own identifier scheme.
• Further, publications that are accessible online can be represented by their URL. All of
these identifiers are unique (an identifier identifies a single resource), but mostly not
single (a resource may be identified by multiple identifiers).
• DOIs whose registration is centralized that the same publication is assigned two or
more identifiers.
36

Representing Identity
• Multiple identifiers can be represented in RDF in two separate ways:
• One can introduce a separate resource and use the identifiers as URIs for these
resources
• The other alternative is to chose one of the identifiers and use it as a URI.
• The first criterion means that good URIs should be unambiguous.
• The second criterium is no way to rename resources (to reassign URIs). Once a URI
changes the only solution is to introduce a new resource and assert its equivalence
with the old resource.
• The use of http:// URIs has been somewhat complicated by the long outstanding issue
of whether such URIs could be used for only web resources or also for abstract
concepts.
• The response code used by web servers to indicate success should be used only for
web resources.
37

On the notation of equality
• RDF and OWL (starting with OWL Lite) allow us to identify resources and to
represent their (in)equality using the owl:sameAs and owl:differentFrom properties.
• In particular, the meaning of equality with respect to the objects of the domain
depends on the domain itself and the goals of the modelling.
• Identity of Indiscernibles or Leibniz-law: the principle of all things being identical
unless we are able to distinguish (discern) them or put it differently: no two distinct
objects have exactly the same properties.
• The Leibniz-law is formalized using the logical formula given in Formula 5.1.
• The converse of the Leibniz-law is called Indiscernibility of Identicals and written as
Formula 5.2.
38

On the notation of equality
• The reflexive, symmetric and transitive properties of equality follow from these
definitions.
• Example
39

Determining equality
• In OWL there are a limited set of constructs that can lead to (in)equality statements. Functional
and inverse functional properties (IFPs) and maximum cardinality restrictions in general can lead
to conclude that two symbols must denote the same resource when otherwise the cardinality
restriction could not be fulfilled.
• For example, the foaf :mbox property denoting the email address of a person is inverse-
functional as a mailbox can only belong to a single person.
• As another example, consider a hypothetical ex:hasParent property, which has a maximum
cardinality of two.
• If we state that a single person has three parents (which we are allowed to state) an OWL
reasoner should conclude that at least two of them has to be the same. Once inferred, the equality
of instances can be stated by using the owl:sameAs property. There are more ways to conclude
that two symbols do not denote the same object, which is expressed by the owl:differentFrom
property. For example, instances of disjoint classes must be different from each other.
40

Reasoning with instance equality
• Reasoning is the inference of new statements (facts) that necessarily follow from the set of
known statements.
• Every piece of additional knowledge excludes some unintended interpretation of our
knowledge base.
• Description Logic versus rule-based reasoners
– Description Logic reasoners are designed to support primary tasks of classification and
the consistency checking of ontologies.
– Rule-based reasoning: semantics of RDF(S) can be completely expressed using a set of
inference rules.
41

Reasoning with instance equality
• Forward versus backward chaining
– Forward chaining means that all consequences of the rules are computed to obtain what is
called a complete materialization of the knowledge base.
– Advantage: Queries are fast to execute since all true statements are readily available.
– Disadvantage: Storing the complete materialization often takes exorbitant amounts of space
as even the simplest RDF(S) ontologies result in a large amount of inferred statements.
– Backward-chaining, rules are executed “backwards” and on demand, i.e. when a query
needs to be answered. While with forward chaining we check the prerequisites of rules and
add all their conclusions, here we are given a conclusion and check whether it is explicitly
stated or whether it could inferred from some rule, either because the prerequisites of the
rule are explicitly stated to be true or again because they too could be inferred from some
other rule(s).
– Drawback: Backward chaining is longer query execution times.
42

Evaluating smushing
• Smushing can be considered as either a retrieval problem or a clustering problem.
• In terms of retrieval, we try to achieve a maximum precision and recall of the theoretically
correct set of mappings.
• In ontology mapping where there is a one-to-one mapping between elements of two
ontologies.
• In the instance unification problem a single resource may be mapped to a number of
other resources.
• Instance unification or smushing can be considered as an optimization task where we try to
optimize an information retrieval or clustering-based measure.
43

Advanced Representation
• Some of the knowledge we wanted to express could not be captured in OWL DL.
• More expressive power in representation is expected from the Rule Interchange Format
(RIF) currently under development by the W3C.
• Additional expressive power can be brought by using logics that go beyond predicate calculus.
• For example, temporal logic extends assertions with a temporal dimension: using temporal logic
we can express that a statement holds true at a certain time or for a time interval.
• Extending logic with probabilities is also a natural step in representing our problem more
accurately.
• When combining similarities we can use weights to express the relevance of certain properties.
• Such weights can be determined ad hoc or can be learned from training data using machine
learning. The resulting probability could be represented in a probabilistic logic.
45

CS6010 Social Network Analysis Unit II

More Related Content

What's hot (20)

Similar to CS6010 Social Network Analysis Unit II (20)

More from pkaviya (20)

Recently uploaded (20)

CS6010 Social Network Analysis Unit II