Knowledge Diversity in Media Content Analysis An Analytico-synthetic - - PowerPoint PPT Presentation

Devika vika P. Madall dalli and A.R.D. D.Pr Prasad asad

DRTC, Indian Statistical Institute Bangalore, INDIA

Knowledge Diversity in Media Content Analysis – An Analytico-synthetic Approach

UDCC Classification and Ontologies Seminar, The Hague, Sept 2011

Premise

• Knowledge is increasingly characterized by diversity that

results in complexities

• Socio-cultural background and purposes attribute to

knowledge diversity.

• Analytico-synthetic approach for topical diversity
• Facet analysis (FA) and synthesis for use in annotations for

Media Content Analysis within the EC funded 'Living Knowledge' Project.

Introduction

As an asset, diversity of knowledge could lead to projections of semantic correlations, bias and

• pinion mining.

Background – LivingKnowledge Project

• Living Knowledge’ (LK) [EU FET project n0 231126]

considers diversity as an asset and aims to make it traceable, understandable and exploitable, with the goal of improving navigation and searching in very large datasets (Maltese, etal, 2009).

Aims of the project

• study the effects of diversity and time on opinions and

bias in socio-economic relevance, especially for seamless representation and exchange of information.

• Intuitive search and navigation tools (e.g. search

engines) need produce more insightful, better

• rganized, aggregated and easier-to-understand output.

Scope of LivingKnowledge Project

• LK employs interdisciplinary competence from philosophy
• f science, cognitive science, library science and semiotics.
• The proposed solution is based on the foundational notions
• f context and its ability to localize meaning, and the notion
• f facet, as from library science, and its ability to organize

knowledge as a set of interoperable components (i.e.facets).

• Diversity of knowledge can be attributed to socio-cultural

interactions and dynamics that casts different aspects of same or similar concepts in the same and similar domains.

Facet Analysis and Faceted Classification

• ‘Facet’ has been used in different connotations.
• Facet is synonymous to category, attribute, class,

group, concept etc (La Barre, 2006).

Faceted Analysis and Classification (Contd.)

• Facet is a distinct division of domain
• Facet is the set of classes got by applying

characteristics successively

• Many facets together make a domain (different

domains have different combinations of facets)

Facet Analysis

• FA is an intellectual process leading to analysis of

a subject into its facets according to postulates and principles (S.R.Ranganathan)

• It results in sorting of terms in a given domain

into homogeneous, mutually exclusive facets, each derived from the parent universe (CIU) by a single characteristic of division at each level in a hierarchy.

Media Content Analysis (MCA)

• SORA (Institute for Social Research and

Consulting Ogris & Hofinger GmbH) carries out Analysis of Media Content.

• MCA carried out in 2 phases:
• Phase 1: Facts are entered in a ‘codebook’ that

formalises the representation of the media resources

• Phase 2: Media analysis is based on annotations and

coding resources in order to answer some typical questions.

Example of Facetization in MCA

• To reconstruct the mediated discourse we have to

describe the relevant actors, patterns of interpretation, differences between actors, topics and different degrees of

• diversity. Example: ‘European elections and integration.’
• The basic research questions are:
• What (sub-)topics occur to what extent in the mediated

discourse on ‘integration’?

• What actors in what roles are present in the mediated

discourse on ‘integration’?

• What patterns of interpretation occur in the mediated

discourse on ‘integration’?

Example of Facetization in MCA

• Integration refers to the two chosen sub-topics (labour

market and religion) and raised these questions:

• What are the main [topics, actors and countries, arguments,

frames] related to integration?

• Which of these [topics, actors and countries, arguments,

frames] are the most [controversial, accepted, subjective, biased, etc.].

• What are the main [politicians, parties, organisations etc.]

discussing integration in a [negative, positive, neutral] context?

• Which [politicians, parties, organisations etc.] have changed

their discourse on integration (i.e. from positive to negative)?

• What time periods are most important for integration, and

what other events are correlated to these periods?

• How developed is the discursive character of statements from

different [politicians, parties, organisations etc.]?

Analytico-synthetic Approach -Topical Annotations of Media Content

• Library based faceted knowledge organization system build

representations by division of domains into distinct facets based on principles of shared unambiguous characteristics between the member concepts (Bhattacharya, 1979).

• The same approach has been adopted for MCA, identifying

actors indicators - 'Who' -- 'What' and 'To Whom'-- 'What'.

• ‘Content’ in media is analysed following the faceted

approach for topical representation. ‘Topic' is used to inter- relate the other indicators in a subject based representation.

Topic Facets [D]iscipline [E]ntity [P]roperty [A]ction [M]odifier

Can be represented y using

Common Isolates Space Time Persons Form …etc

With

Who To Whom What Channel Can be represented using Actor Roles Affiliation Scope Form Topic Claims Frames Opinion Addressee Object Actor Mass Media Forum Blog Who What To Whom What Channel

Indicators (Facets)

Faceted representation of codebook indicators

Multidimensional Faceted Representation for Diversity

• Faceted representation for Political Science

through POPSI technique is represented.

Topic Nationalism as an issue in forthcoming European Elections 2012 M Space Modifiers Time Modifiers Asia 20th Century Australia 21st Century Europe 2012 D Economics Law Political Science Sociology E Socialism Communism Fascism Nationalism

A

Nomination Administration Representation Election

Examples

• Example : Nationalism as an issue in forthcoming European Elections

2012”. Analyzing into facets: the focus is on European Elections, Nationalism, location and time. Concept Identification: Political Science [Discipline], Nationalism [Entity], Election [Action], Forthcoming [Modifier for Action], European [Space], 2009 [Time]

Coordination

• Strings from Political Science have the subject

index alphabetic file:

• Political Science [Discipline], Election [Action],

Free [Modifier for Action], Egypt [Space]

• Political Science [Discipline], Election [Action],

Europe, Georgia [Space]

Application in MCA

• In manual annotations annotators pick a few key terms to

assign to the resources being annotated.

• A faceted annotator is equipped with subject strings that provide

context to the content of media resource.

• The advantages are:
• annotator can pick the relevant context from a list of

suggestions in the form of fully qualified modulated strings

• f descriptors.
• Such a process of subject index is based on principles where

context is intrinsically provided.

• For example – if the query is ‘Elections in Egypt’ then

retrieved set certainly has the strings with ‘Elections’ but since Egypt is space element the concept behind is disambiguated using ‘Egypt’ as context.

EU Corpus and FA

• For the purpose of the LK project web resources

in domain 'EU Elections' serves as corpus maintained by EA.

• Media content is analysed into facets and subject

strings are derived in a principled approach.

Example

POPSI categories and entries are made into a classaurus:

Annotations presented for codebook application.

Advantages – Domain-specific Annotations

• Domain-specific annotations can be deployed to

provide subject or topical approach to content.

• The context is driven by topical concepts at broader or

narrower levels to a given term or by space or temporal modifiers.

• In LK project facetization enables the MCA to retrieve

media discourse by certain concepts that are represented by facets and by intrinsic relations between such facets.

• In the SORA model keywords were assigned by

annotators, but in faceted model a faceted domain based

• ntology as shown is deployed to provide the semantic

context.

23

09/11/2001

Today, our fellow citizens, our way of life, our very freedom came under attack in a series of deliberate and deadly terrorist acts. Immediately following the first attack, I implemented our government's emergency response plans. Our military is powerful, and it's prepared. Our emergency teams are working in New York City and Washington, D.C., to help with local rescue efforts – 9/11/2001

09/11/2011

24

"These past 10 years underscore the bonds between all Americans. We have not succumbed to suspicion and mistrust. After 9/11, President (George) Bush made clear what we reaffirm today: the United States will never wage war against Islam or any

• religion. Immigrants come here from all

parts of the globe," The progeny of Muslim asylum seekers mar a moment of silence for 9/11 victims by burning an American flag at a London memorial

The time and topical change/evolution

25 09/11/2001 Invasion of Iraq- 2003 Afghanistan War - 2002 Mumbai Attack- 2008 Operation New Dawn - 2010 Ground Zero- 2011

The faceted approach

• The facets are

26

Topic Time Space

09/11/2001 2001, 2011

U.S, India, Afghanistan Iraq ….

• The diversity in context can be expressed by the combinations of the

above facets for particular instances

Conclusion

• Analytico synthetic approach basically facilitates

identifying the distinct facets of a domain and synthesizing the facets to represent distinct contexts.

• The method helps to deal with diversity, as background

knowledge is modelled in a principled approach that leads to a flexible (meccano) arrangement of divisions of domains that can combined and reused as per context.

THANK YOU!

devika@drtc.isibang.ac.in