Theory and Method for Studying Meaning and Related Social Structures

8/2/2019 Theory and Method for Studying Meaning and Related Social Structures

1/12

Mark Thomas Kennedy, 2010. All Rights Reserved.

Theory and Method for

Studying Meaning andRelated Social Structures

Introducing Association Engine (),

A Practical Tool for Relational Content Analysis

2010 Academy of Management Meetings | MontrealContent Analysis in Organizational Research:

Techniques and Applications


2/12


Outline

Cool research questions

Meaning and social structure

New (research) realities

Key Inputs

Simple interface

Robust technologies

Community, standards matter!

Theory meetsMethod

Anatomyof a Project/Introducing

Relational ContentAnalysis Made Easy

Conclusion


3/12


Cool topics being addressed by linkingmacro entities to patterns ofassociation

Grass-fed meat, dairyactivists found a collectionof codes seen as groundfor new identity, practices.1

WorkstationsMedia co-mentions of rivals revealcategory, prototype.2

Greenseen by finding

overlap in antagonistsassociation of attributes.3

Nanotechrepeatedpatterns of associationamong patent classes.4

Social and politicalmovements: How dofringe interests becomemainstream issues?

Markets: how doerstwhile misfits becomenew categories?

Corporate reputation:

how do criteria changeand emerge?

Fields, Forms &Genres: how do cross-category blends getnaturalized?

Published

In process

1. Weber et al. 2008

(ASQ)2. Kennedy 2008 (ASR)

3. Kennedy & Chok


4/12


Repeated patterns of association: awindow to dynamics of social realities

Meaning,Structure

Networks,Structure

Process,NetworkDynamics

Because meaning and structure are both seen inrelations among entities/ideas, they are duallyconstituted (Mohr and Duquenne 1997, Mohr 1998)

Recognition of social realities (sometimes alsolegitimate) emerges from repeated associations thatlink actors, entities or ideas (Emirbayer 1997, Tilly2008, Philips et al. 2000, Powell et al. 2005, Kennedy

2008)

One can trace the process of field-level change (Hardy& Maguire 2008) by studying the dynamics of networksof changing patterns of association between entities,

ideas, etc., (Owen-Smith & Powell 2008) and vice

But this type of work is laborious and riskybecause:1. the data is hard to get and code;2. finding relations requires RDB and trickyqueries;3. theory and method are new to most reviewers


5/12


Still hard, but the flood of web content =big opportunities for content analysis

Opportunity

Computer-assisted

Hundreds of thousands+

Also Blogs, RSS feeds

Commenters, too

Make-your-own!

Months?

Months, weeks, days,

Semantic networks

Before

Human coding

Hundreds

Media, PR

Reporters, Analysts

Published catalogs

Year or more

Usually years

Qualitative themes

Coding

Scale

Sources

Lag

Periods

Product


6/12


Of the styles of content analysis,this calls for relationalcontent analysis*

Semiotic

Semantic

Relational1

Statistical

Interpretation Explanation

FocalTer

m

Setsof

Terms

Em

piricalFocus

Styles of Content Analysis

Study Intent1. Carley and Palmquist 1992


7/12


8/12 Mark Thomas Kennedy, 2010. All Rights Reserved.

Interface (netbook example)



Heres what a fuzzy ontology looks like

MTKcomputers

netbooka small, light-weight low cost laptop computer with long battery life; often used for Internet access and wireless

communicationnoun

lightweight, inexpensive, wireless,small, smaller than a laptoplong battery life

Dell, Sylvania, Acer, MSI, Lenovo, Samsung, HP, Toshiba, Asus|ASUSHP 2140 Mini-Note, Samsung NC10, Dell Inspiron Mini 9, MSI Wind, U100-279US, Asus Eee PC 1000HA, Lenovo

Ideapad S10, HP Mini 1000 XP Edition, Acer Aspire One AOA150-1126, Asus Eee PC 900HA, Sylvania GNET28001SO, Mesolaptop, mini-laptop, notebook, mini-notebook, sub-notebook, ultraportablemainframe, workstation, server

laptop, notebook, personal computer, PCsmartphone, tablet

personal computer

a small digital computer based on a microprocessor and designed to be used by one person at a timepersonal, small, microprocessor, desktop, speed, size, memory, storagePC, desktop



builds on robust open sourcelibraries for search, database, networkanalysis

Slicer& Dicer

Mention-Finder4

ConceptMapper

NetworkAnalyzer6

GUI(Front-End)

CorpusInput Files1

FuzzyOntology2

ConceptNetworks7

OutputCorpus6

LongitudinalDataset8

Inputs Outputs

isan association engine, a search utility that produces dynamic concept networks from 3 inputs:(1) a collection of relevant texts from LexisNexis, Factiva, or RSS/Atom feeds from blogs or websites (corpus);(2) a set of terms potentiallyrelevant to the concepts meaning (fuzzy ontology); and(3) rule(s) for when to treat co-mentions of these terms as links (association rules).

1. From any of the following:

LexisNexis (HTML), Factiva (XML), or the web (RSS/Atom feeds).

2. .fo file containing terms,definitions, synonyms, etc., orcommands for fetching themautomatically.

6. Corpus input files convertedand combined into one .c filefor later re-use as desired.

7. As Pajek .net file for analysis,visualization, exchange.

8. Output (.dat) file contains time-varying measures of conceptnetwork structure, position ofrelevant terms.

3. Builds on Lucene, an open source library (Java) used in a wide rangeof production search applications.

4. Builds on Lucene and MySQL, an open source implementation ofSQL. For table structure, see Functional Specification, section 2.4.

5. Builds on JUNG, an open source library (in Java) for social networkanalysis.

Notes

Modules

DB5



Anatomy of a Project:3 Key Outputs

Visualize

Analyze

Explore* planned *

produces dynamic concept networks (DCNs),

Graph models of meaning and structure

Visualize / animate to observe changing patterns of

association over time

Get flat-file datasets for quantitative analysis (one rowper for each node-period observation)

Use DCNs to browse the corpus

Clicking nodes fetches texts that mention that term

Clicking a link fetches texts that co-mention thelinked terms


12/12 Mark Thomas Kennedy 2010 All Rights Reserved

Conclusion: Sober Excitement

Vision

Fitness

Or JustDeviation?

Enable study of difficult questions (dynamics), but cuttime, expense, riskincrease research productivity

Environmental factors giving content analysis new zip

Explosion of web contentripe for analysis Open source, esp. for search engine development

Advances in theory that link macro-level entities torecognition of patterns in lower-level relations

Take-off will also require: theory, identity, community,mutual recognition, complementary roles

Hazard-rate reg.+ institutions / ecology = take-off!

New content analysis + ______________ = ???

Other ideas?

Documents

Theory and Method for Studying Meaning and Related Social Structures