johannes kepler university linz...•space working package 1: building a conceptual foundation. ‣...
TRANSCRIPT
JOHANNES KEPLER
UNIVERSITY LINZ
JOHANNES KEPLER UNIVERSITY LINZ Altenberger Straße 69 4040 Linz, Austria jku.at
Draft paper version prepared for the Young Economists Conference 2019 1-2 October 2019, ViennaMatthias Aistleitner, [email protected] www.jku.at/icae
(Towards) exploring the genesis of competition in economic thought
Research background: SPACE project
!3
• Spatial Competition and Economic Policies (SPACE): Discourses,Institutions and Everyday Practices
Research background: SPACE project
!3
• Spatial Competition and Economic Policies (SPACE): Discourses,Institutions and Everyday Practices
‣ Investigating the impact of an increasingly strong reliance oncompetition as a prime mode of social organization
‣ Competition as a core concept for designing social institutions
Research background: SPACE project
!3
• Spatial Competition and Economic Policies (SPACE): Discourses,Institutions and Everyday Practices
‣ Investigating the impact of an increasingly strong reliance oncompetition as a prime mode of social organization
‣ Competition as a core concept for designing social institutions
• Six Working Packages
Research background: SPACE project
!3
Research background: SPACE project
!4
• SPACE Working Package 1: Building a conceptual foundation.
Research background: SPACE project
!4
• SPACE Working Package 1: Building a conceptual foundation.
‣ How is C defined and conceptualized in economics?
Research background: SPACE project
!4
• SPACE Working Package 1: Building a conceptual foundation.
‣ How is C defined and conceptualized in economics?
Research background: SPACE project
!4
• SPACE Working Package 1: Building a conceptual foundation.
‣ How is C defined and conceptualized in economics?
‣ Which temporal changes has the dominant meaning of Cundergone in the post WWII-period?
Research background: SPACE project
!4
• SPACE Working Package 1: Building a conceptual foundation.
‣ How is C defined and conceptualized in economics?
‣ Which temporal changes has the dominant meaning of Cundergone in the post WWII-period?
Research background: SPACE project
!4
• SPACE Working Package 1: Building a conceptual foundation.
‣ How is C defined and conceptualized in economics?
‣ Which temporal changes has the dominant meaning of Cundergone in the post WWII-period?
‣ How do major economic conceptions of C differ between economicparadigms?
Research background: SPACE project
!4
Aim of this paper
!5
• A contribution in identifying and deconstructing the conceptualfoundation/evolution of C in economics and its observed changes overtime.
Aim of this paper
!5
• A contribution in identifying and deconstructing the conceptualfoundation/evolution of C in economics and its observed changes overtime.
‣ Analysis of economic research literature via topic modeling
Aim of this paper
!5
• A contribution in identifying and deconstructing the conceptualfoundation/evolution of C in economics and its observed changes overtime.
‣ Analysis of economic research literature via topic modeling‣ Topic models = algorithms that discover the content of large
collections of documents via topics
Aim of this paper
!5
• A contribution in identifying and deconstructing the conceptualfoundation/evolution of C in economics and its observed changes overtime.
‣ Analysis of economic research literature via topic modeling‣ Topic models = algorithms that discover the content of large
collections of documents via topics‣ Latent Dirichlet Allocation (LDA) as a simple and widely used topic
model (Blei et al. 2003)
Aim of this paper
!5
Aim of this paper
!6
• Topic modeling has emerged as an established method in answeringspecific research questions in economics and economic history.
Aim of this paper
!6
• Topic modeling has emerged as an established method in answeringspecific research questions in economics and economic history.Focus on economic literature
Aim of this paper
!6
• Topic modeling has emerged as an established method in answeringspecific research questions in economics and economic history.Focus on economic literature
‣ located in a specific journal/s (e.g. Lüdering and Winker 2016;Wehrheim 2019, Ambrosino et al. 2018)
Aim of this paper
!6
• Topic modeling has emerged as an established method in answeringspecific research questions in economics and economic history.Focus on economic literature
‣ located in a specific journal/s (e.g. Lüdering and Winker 2016;Wehrheim 2019, Ambrosino et al. 2018)
‣ on a more comprehensive level (e.g. whole sub-fields for a givenperiod (Angrist et al. 2017; Jelveh et al. 2015)
Aim of this paper
!6
• Topic modeling has emerged as an established method in answeringspecific research questions in economics and economic history.Focus on economic literature
‣ located in a specific journal/s (e.g. Lüdering and Winker 2016;Wehrheim 2019, Ambrosino et al. 2018)
‣ on a more comprehensive level (e.g. whole sub-fields for a givenperiod (Angrist et al. 2017; Jelveh et al. 2015)
• In this paper: applying topic modeling to a part of the literature which isex ante constraint by a specific research topic: competition.
Aim of this paper
!6
Research background: SPACE project
!7
• SPACE Working Package 1: Building a conceptual foundation.
Research background: SPACE project
!7
• SPACE Working Package 1: Building a conceptual foundation.
‣ How is C defined and conceptualized in economics?
Research background: SPACE project
!7
• SPACE Working Package 1: Building a conceptual foundation.
‣ How is C defined and conceptualized in economics?
‣ Which latent topics emerge within the literature that relates tocompetition?
Research background: SPACE project
!7
• SPACE Working Package 1: Building a conceptual foundation.
‣ How is C defined and conceptualized in economics?
‣ Which latent topics emerge within the literature that relates tocompetition?
‣ Which temporal changes has the dominant meaning of Cundergone in the post WWII-period?
Research background: SPACE project
!7
• SPACE Working Package 1: Building a conceptual foundation.
‣ How is C defined and conceptualized in economics?
‣ Which latent topics emerge within the literature that relates tocompetition?
‣ Which temporal changes has the dominant meaning of Cundergone in the post WWII-period?
‣ Focusing on the dominant (mainstream) research literature ineach period: Are there differences in latent topics over time?
Research background: SPACE project
!7
• SPACE Working Package 1: Building a conceptual foundation.
‣ How is C defined and conceptualized in economics?
‣ Which latent topics emerge within the literature that relates tocompetition?
‣ Which temporal changes has the dominant meaning of Cundergone in the post WWII-period?
‣ Focusing on the dominant (mainstream) research literature ineach period: Are there differences in latent topics over time?
‣ How do major economic conceptions of C differ between economicparadigms?
Research background: SPACE project
!7
• SPACE Working Package 1: Building a conceptual foundation.
‣ How is C defined and conceptualized in economics?
‣ Which latent topics emerge within the literature that relates tocompetition?
‣ Which temporal changes has the dominant meaning of Cundergone in the post WWII-period?
‣ Focusing on the dominant (mainstream) research literature ineach period: Are there differences in latent topics over time?
‣ How do major economic conceptions of C differ between economicparadigms?
‣ Are there latent topics which can be assigned to economicparadigms?
Research background: SPACE project
!7
The LDA model and its main assumptions
!8Source: Steyvers and Griffiths 2007
• LDA is best described by its “imaginary random process by which themodel assumes the documents arose” (Blei 2012, 78)
The LDA model and its main assumptions
!8Source: Steyvers and Griffiths 2007
• LDA is best described by its “imaginary random process by which themodel assumes the documents arose” (Blei 2012, 78)
• In the model, topics are give first and then the documents aregenerated.
The LDA model and its main assumptions
!8Source: Steyvers and Griffiths 2007
• LDA is best described by its “imaginary random process by which themodel assumes the documents arose” (Blei 2012, 78)
• In the model, topics are give first and then the documents aregenerated.
• Documents = bag of words
The LDA model and its main assumptions
!8Source: Steyvers and Griffiths 2007
• LDA is best described by its “imaginary random process by which themodel assumes the documents arose” (Blei 2012, 78)
• In the model, topics are give first and then the documents aregenerated.
• Documents = bag of words
• For every document in the corpus
The LDA model and its main assumptions
!8Source: Steyvers and Griffiths 2007
• LDA is best described by its “imaginary random process by which themodel assumes the documents arose” (Blei 2012, 78)
• In the model, topics are give first and then the documents aregenerated.
• Documents = bag of words
• For every document in the corpus1. Pick a topic
The LDA model and its main assumptions
!8Source: Steyvers and Griffiths 2007
• LDA is best described by its “imaginary random process by which themodel assumes the documents arose” (Blei 2012, 78)
• In the model, topics are give first and then the documents aregenerated.
• Documents = bag of words
• For every document in the corpus1. Pick a topic2. Pick a word
The LDA model and its main assumptions
!8Source: Steyvers and Griffiths 2007
• LDA is best described by its “imaginary random process by which themodel assumes the documents arose” (Blei 2012, 78)
• In the model, topics are give first and then the documents aregenerated.
• Documents = bag of words
• For every document in the corpus1. Pick a topic2. Pick a word3. Place it in the bag
until the document is complete
The LDA model and its main assumptions
!8Source: Steyvers and Griffiths 2007
The LDA model and its parameters
!9Source: Steyvers and Griffiths 2007
The LDA model and its parameters
!9# of words per document
word
Topic assignment for word
Source: Steyvers and Griffiths 2007
The LDA model and its parameters
!9
topic-document distribtution (which topics for which document?)
# of documents
# of words per document
word
Topic assignment for word
Source: Steyvers and Griffiths 2007
The LDA model and its parameters
!9
topic-document distribtution (which topics for which document?)
# of documents
word-topic distribution (which words for which topic?)
# of topics# of words per document
word
Topic assignment for word
Source: Steyvers and Griffiths 2007
The LDA model and its parameters
!9
topic-document distribtution (which topics for which document?)
# of documents
word-topic distribution (which words for which topic?)
# of topics# of words per document
word
Topic assignment for word
Hyperparameters (priors)
Source: Steyvers and Griffiths 2007
The LDA model and its parameters
!9
topic-document distribtution (which topics for which document?)
# of documents
word-topic distribution (which words for which topic?)
# of topics# of words per document
word
Topic assignment for word
Hyperparameters (priors)
Text preprocessing
Source: Steyvers and Griffiths 2007
The puzzle of statistical inference
!10
• The posterior (conditional) distribution is
The puzzle of statistical inference
!10
• The posterior (conditional) distribution is
The puzzle of statistical inference
!10
• The posterior (conditional) distribution is
The puzzle of statistical inference
!10
Joint distribution of all hidden and observed variables
• The posterior (conditional) distribution is
The puzzle of statistical inference
!10
Joint distribution of all hidden and observed variables
Marginal probability of the observations
• The posterior (conditional) distribution is
“[The marginal probability] is the probability of seeing the observed corpus under any topic model.” (Blei 2012, 81)
The puzzle of statistical inference
!10
Joint distribution of all hidden and observed variables
Marginal probability of the observations
• The posterior (conditional) distribution is
“[The marginal probability] is the probability of seeing the observed corpus under any topic model.” (Blei 2012, 81)➡ Summing up all possible ways of assigning each observed word
(usually in the order of millions!) to one of the topics
The puzzle of statistical inference
!10
Joint distribution of all hidden and observed variables
Marginal probability of the observations
• Collapsed Gibbs Sampling as inference algorithm
The puzzle of statistical inference
!11
• Collapsed Gibbs Sampling as inference algorithm
The puzzle of statistical inference
!11
The puzzle of statistical inference
!12
• Collapsed Gibbs sampling (CGS) as inference algorithm (Griffiths andSteyvers 2004)
The puzzle of statistical inference
!12
• Collapsed Gibbs sampling (CGS) as inference algorithm (Griffiths andSteyvers 2004)
• Markov Chain Monte Carlo (MCMC) based method
The puzzle of statistical inference
!12
• Collapsed Gibbs sampling (CGS) as inference algorithm (Griffiths andSteyvers 2004)
• Markov Chain Monte Carlo (MCMC) based method
• Inferring the conditional distribution via estimating z (assignments ofwords to topics)
The puzzle of statistical inference
!12
• Collapsed Gibbs sampling (CGS) as inference algorithm (Griffiths andSteyvers 2004)
• Markov Chain Monte Carlo (MCMC) based method
• Inferring the conditional distribution via estimating z (assignments ofwords to topics)
The puzzle of statistical inference
!12
• Collapsed Gibbs sampling (CGS) as inference algorithm (Griffiths andSteyvers 2004)
• Markov Chain Monte Carlo (MCMC) based method
• Inferring the conditional distribution via estimating z (assignments ofwords to topics)
The puzzle of statistical inference
!12
• Collapsed Gibbs sampling (CGS) as inference algorithm (Griffiths andSteyvers 2004)
• Markov Chain Monte Carlo (MCMC) based method
• Inferring the conditional distribution via estimating z (assignments ofwords to topics)
The puzzle of statistical inference
!12
• Collapsed Gibbs sampling (CGS) as inference algorithm (Griffiths andSteyvers 2004)
• Markov Chain Monte Carlo (MCMC) based method
• Inferring the conditional distribution via estimating z (assignments ofwords to topics)
The puzzle of statistical inference
!12
• Collapsed Gibbs sampling (CGS) as inference algorithm (Griffiths andSteyvers 2004)
• Markov Chain Monte Carlo (MCMC) based method
• Inferring the conditional distribution via estimating z (assignments ofwords to topics)
• Calculating 𝜙 (word-topic distribution)
The puzzle of statistical inference
!12
• Collapsed Gibbs sampling (CGS) as inference algorithm (Griffiths andSteyvers 2004)
• Markov Chain Monte Carlo (MCMC) based method
• Inferring the conditional distribution via estimating z (assignments ofwords to topics)
• Calculating 𝜙 (word-topic distribution)
• Calculating 𝜃 (topic-document distribution)
The puzzle of statistical inference
!12
Data
!13
• Database: JSTOR Data for Research
Data
!13
• Database: JSTOR Data for Research
• Search term “competition“
Data
!13
• Database: JSTOR Data for Research
• Search term “competition“
• 124.749 items published in 260 journals between 1851-2017
Data
!13
• Database: JSTOR Data for Research
• Search term “competition“
• 124.749 items published in 260 journals between 1851-2017
• Full text ngram counts + bibliometric meta-data
Data
!13
• Database: JSTOR Data for Research
• Search term “competition“
• 124.749 items published in 260 journals between 1851-2017
• Full text ngram counts + bibliometric meta-data
• “Narrow” sample: items containing “competition” > 3
‣ 27.488 articles
‣ 227 journals (thereof 34 heterodox journals)
‣ Text preprocessing + “pseudo” abstracts as final input for the model
‣ 98.572 unique words (5.496.608 total words in the corpus)
Data
!13
(preliminary) results I: descriptives
!14
1850 1860 1870 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 20100
2000
4000
6000
8000
10000
no.ofpapers
total resultscompetition >1competition >3
(preliminary) results I: descriptives
!14
1850 1860 1870 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 20100
2000
4000
6000
8000
10000
no.ofpapers
total resultscompetition >1competition >3
(preliminary) results I: descriptives
!15
1850 1860 1870 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010
0
500
1000
1500
2000
2500
no.ofpapers
competition >1competition >3
(preliminary) results I: descriptives
!15
1850 1860 1870 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010
0
500
1000
1500
2000
2500
no.ofpapers
competition >1competition >3
(preliminary) results I: descriptives
!16
(preliminary) results I: descriptives
!16
(preliminary) results I: descriptives
!16
4 out of the ‘Big 5’
(preliminary) results I: descriptives
!16
4 out of the ‘Big 5’
HET journals
• CGS with overall “narrow” sample
‣ Number of topics ! =30
‣ Dirichlet hyperparameters ⍺ = 50/! , β = 0.01
‣ Number of iterations: 500
TT
(preliminary) results II: a bird’s eye’s view on competition in economic research
!17
(preliminary) results II: some examples
!18
Topic1
Topic2
Topic3
Topic4
Topic5
Topic6
Topic7
Topic8
Topic9
Topic10
Topic11
Topic12
Topic13
Topic14
Topic15
Topic16
Topic17
Topic18
Topic19
Topic20
Topic21
Topic22
Topic23
Topic24
Topic25
Topic26
Topic27
Topic28
Topic29
Topic30
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35document-topicdistribution(θ)
David P., Angel, James, Engstrom (1995):Manufacturing Systems and Technological Change: The U.S. Personal Comp
Economic Geography
(preliminary) results II: some examples
!18
Topic 1 Weight (ϕ)technology 0.0247industry 0.0172innovation 0.0164firms 0.0158product 0.0153technological 0.0151research 0.0147production 0.013process 0.0128firm 0.0119investment 0.0113development 0.0108industries 0.0104products 0.0104industrial 0.0103
Topic1
Topic2
Topic3
Topic4
Topic5
Topic6
Topic7
Topic8
Topic9
Topic10
Topic11
Topic12
Topic13
Topic14
Topic15
Topic16
Topic17
Topic18
Topic19
Topic20
Topic21
Topic22
Topic23
Topic24
Topic25
Topic26
Topic27
Topic28
Topic29
Topic30
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35document-topicdistribution(θ)
David P., Angel, James, Engstrom (1995):Manufacturing Systems and Technological Change: The U.S. Personal Comp
Economic Geography
• Topic1: technological innovation and industrial development
(preliminary) results II: some examples
!18
Topic 1 Weight (ϕ)technology 0.0247industry 0.0172innovation 0.0164firms 0.0158product 0.0153technological 0.0151research 0.0147production 0.013process 0.0128firm 0.0119investment 0.0113development 0.0108industries 0.0104products 0.0104industrial 0.0103
Topic1
Topic2
Topic3
Topic4
Topic5
Topic6
Topic7
Topic8
Topic9
Topic10
Topic11
Topic12
Topic13
Topic14
Topic15
Topic16
Topic17
Topic18
Topic19
Topic20
Topic21
Topic22
Topic23
Topic24
Topic25
Topic26
Topic27
Topic28
Topic29
Topic30
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35document-topicdistribution(θ)
David P., Angel, James, Engstrom (1995):Manufacturing Systems and Technological Change: The U.S. Personal Comp
Economic Geography
• Topic1: technological innovation and industrial development
(preliminary) results II: some examples
!18
Topic 1 Weight (ϕ)technology 0.0247industry 0.0172innovation 0.0164firms 0.0158product 0.0153technological 0.0151research 0.0147production 0.013process 0.0128firm 0.0119investment 0.0113development 0.0108industries 0.0104products 0.0104industrial 0.0103
Topic1
Topic2
Topic3
Topic4
Topic5
Topic6
Topic7
Topic8
Topic9
Topic10
Topic11
Topic12
Topic13
Topic14
Topic15
Topic16
Topic17
Topic18
Topic19
Topic20
Topic21
Topic22
Topic23
Topic24
Topic25
Topic26
Topic27
Topic28
Topic29
Topic30
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35document-topicdistribution(θ)
David P., Angel, James, Engstrom (1995):Manufacturing Systems and Technological Change: The U.S. Personal Comp
Economic Geography
Topic 16 Weight (ϕ)countries 0.0178international 0.0161country 0.0142foreign 0.0139trade 0.0128world 0.0122domestic 0.012policy 0.0081export 0.008national 0.0079european 0.0077development 0.0077growth 0.0073economy 0.0073sector 0.0072
• Topic1: technological innovation and industrial development
(preliminary) results II: some examples
!18
Topic 1 Weight (ϕ)technology 0.0247industry 0.0172innovation 0.0164firms 0.0158product 0.0153technological 0.0151research 0.0147production 0.013process 0.0128firm 0.0119investment 0.0113development 0.0108industries 0.0104products 0.0104industrial 0.0103
Topic1
Topic2
Topic3
Topic4
Topic5
Topic6
Topic7
Topic8
Topic9
Topic10
Topic11
Topic12
Topic13
Topic14
Topic15
Topic16
Topic17
Topic18
Topic19
Topic20
Topic21
Topic22
Topic23
Topic24
Topic25
Topic26
Topic27
Topic28
Topic29
Topic30
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35document-topicdistribution(θ)
David P., Angel, James, Engstrom (1995):Manufacturing Systems and Technological Change: The U.S. Personal Comp
Economic Geography
Topic 16 Weight (ϕ)countries 0.0178international 0.0161country 0.0142foreign 0.0139trade 0.0128world 0.0122domestic 0.012policy 0.0081export 0.008national 0.0079european 0.0077development 0.0077growth 0.0073economy 0.0073sector 0.0072
Topic 21 Weight (ϕ)management 0.0102based 0.0101research 0.0097different 0.0093role 0.009university 0.0085information 0.0082important 0.008process 0.0075performance 0.0074structure 0.0072specific 0.0072analysis 0.0071approach 0.007business 0.007
!19
Topic 1 Weight (ϕ)technology 0.0247industry 0.0172innovation 0.0164firms 0.0158product 0.0153technological 0.0151research 0.0147production 0.013process 0.0128firm 0.0119investment 0.0113development 0.0108industries 0.0104products 0.0104industrial 0.0103
Topic1
Topic2
Topic3
Topic4
Topic5
Topic6
Topic7
Topic8
Topic9
Topic10
Topic11
Topic12
Topic13
Topic14
Topic15
Topic16
Topic17
Topic18
Topic19
Topic20
Topic21
Topic22
Topic23
Topic24
Topic25
Topic26
Topic27
Topic28
Topic29
Topic30
0.00
0.05
0.10
0.15
0.20
0.25document-topicdistribution(θ)
David J., Teece (1989):Inter-Organizational Requirements of the Innovation Process
Managerial and Decision Economics
• Topic1: technological innovation and industrial development
(preliminary) results II: some examples
!19
Topic 1 Weight (ϕ)technology 0.0247industry 0.0172innovation 0.0164firms 0.0158product 0.0153technological 0.0151research 0.0147production 0.013process 0.0128firm 0.0119investment 0.0113development 0.0108industries 0.0104products 0.0104industrial 0.0103
Topic1
Topic2
Topic3
Topic4
Topic5
Topic6
Topic7
Topic8
Topic9
Topic10
Topic11
Topic12
Topic13
Topic14
Topic15
Topic16
Topic17
Topic18
Topic19
Topic20
Topic21
Topic22
Topic23
Topic24
Topic25
Topic26
Topic27
Topic28
Topic29
Topic30
0.00
0.05
0.10
0.15
0.20
0.25document-topicdistribution(θ)
David J., Teece (1989):Inter-Organizational Requirements of the Innovation Process
Managerial and Decision Economics
• Topic1: technological innovation and industrial development
(preliminary) results II: some examples
Topic 5 Weight (ϕ)law 0.0133act 0.0119public 0.0099control 0.0082government 0.0078states 0.0075commission 0.0074policy 0.0072federal 0.0067com 0.0067legal 0.0063cases 0.0061power 0.006case 0.0059business 0.0059
!19
Topic 1 Weight (ϕ)technology 0.0247industry 0.0172innovation 0.0164firms 0.0158product 0.0153technological 0.0151research 0.0147production 0.013process 0.0128firm 0.0119investment 0.0113development 0.0108industries 0.0104products 0.0104industrial 0.0103
Topic1
Topic2
Topic3
Topic4
Topic5
Topic6
Topic7
Topic8
Topic9
Topic10
Topic11
Topic12
Topic13
Topic14
Topic15
Topic16
Topic17
Topic18
Topic19
Topic20
Topic21
Topic22
Topic23
Topic24
Topic25
Topic26
Topic27
Topic28
Topic29
Topic30
0.00
0.05
0.10
0.15
0.20
0.25document-topicdistribution(θ)
David J., Teece (1989):Inter-Organizational Requirements of the Innovation Process
Managerial and Decision EconomicsTopic 12 Weight (ϕ)price 0.0213prices 0.0198product 0.0191sales 0.0172market 0.0172consumer 0.0171costs 0.0161consumers 0.0157cost 0.0142competitive 0.0135demand 0.0135firms 0.0134markets 0.0131firm 0.013products 0.013
• Topic1: technological innovation and industrial development
(preliminary) results II: some examples
Topic 5 Weight (ϕ)law 0.0133act 0.0119public 0.0099control 0.0082government 0.0078states 0.0075commission 0.0074policy 0.0072federal 0.0067com 0.0067legal 0.0063cases 0.0061power 0.006case 0.0059business 0.0059
!19
Topic 1 Weight (ϕ)technology 0.0247industry 0.0172innovation 0.0164firms 0.0158product 0.0153technological 0.0151research 0.0147production 0.013process 0.0128firm 0.0119investment 0.0113development 0.0108industries 0.0104products 0.0104industrial 0.0103
Topic1
Topic2
Topic3
Topic4
Topic5
Topic6
Topic7
Topic8
Topic9
Topic10
Topic11
Topic12
Topic13
Topic14
Topic15
Topic16
Topic17
Topic18
Topic19
Topic20
Topic21
Topic22
Topic23
Topic24
Topic25
Topic26
Topic27
Topic28
Topic29
Topic30
0.00
0.05
0.10
0.15
0.20
0.25document-topicdistribution(θ)
David J., Teece (1989):Inter-Organizational Requirements of the Innovation Process
Managerial and Decision EconomicsTopic 12 Weight (ϕ)price 0.0213prices 0.0198product 0.0191sales 0.0172market 0.0172consumer 0.0171costs 0.0161consumers 0.0157cost 0.0142competitive 0.0135demand 0.0135firms 0.0134markets 0.0131firm 0.013products 0.013
• Topic1: technological innovation and industrial development
(preliminary) results II: some examples
Topic 5 Weight (ϕ)law 0.0133act 0.0119public 0.0099control 0.0082government 0.0078states 0.0075commission 0.0074policy 0.0072federal 0.0067com 0.0067legal 0.0063cases 0.0061power 0.006case 0.0059business 0.0059
Topic 18 Weight (ϕ)theory 0.0204economics 0.0182analysis 0.0126press 0.0118university 0.0118journal 0.0112economists 0.0096economy 0.0094york 0.0089cambridge 0.0088new 0.0084american 0.0074general 0.0074view 0.0074approach 0.0073
!19
Topic 1 Weight (ϕ)technology 0.0247industry 0.0172innovation 0.0164firms 0.0158product 0.0153technological 0.0151research 0.0147production 0.013process 0.0128firm 0.0119investment 0.0113development 0.0108industries 0.0104products 0.0104industrial 0.0103
Topic1
Topic2
Topic3
Topic4
Topic5
Topic6
Topic7
Topic8
Topic9
Topic10
Topic11
Topic12
Topic13
Topic14
Topic15
Topic16
Topic17
Topic18
Topic19
Topic20
Topic21
Topic22
Topic23
Topic24
Topic25
Topic26
Topic27
Topic28
Topic29
Topic30
0.00
0.05
0.10
0.15
0.20
0.25document-topicdistribution(θ)
David J., Teece (1989):Inter-Organizational Requirements of the Innovation Process
Managerial and Decision EconomicsTopic 12 Weight (ϕ)price 0.0213prices 0.0198product 0.0191sales 0.0172market 0.0172consumer 0.0171costs 0.0161consumers 0.0157cost 0.0142competitive 0.0135demand 0.0135firms 0.0134markets 0.0131firm 0.013products 0.013
• Topic1: technological innovation and industrial development
(preliminary) results II: some examples
Topic 21 Weight (ϕ)management 0.0102based 0.0101research 0.0097different 0.0093role 0.009university 0.0085information 0.0082important 0.008process 0.0075performance 0.0074structure 0.0072specific 0.0072analysis 0.0071approach 0.007business 0.007
Topic 5 Weight (ϕ)law 0.0133act 0.0119public 0.0099control 0.0082government 0.0078states 0.0075commission 0.0074policy 0.0072federal 0.0067com 0.0067legal 0.0063cases 0.0061power 0.006case 0.0059business 0.0059
Topic 18 Weight (ϕ)theory 0.0204economics 0.0182analysis 0.0126press 0.0118university 0.0118journal 0.0112economists 0.0096economy 0.0094york 0.0089cambridge 0.0088new 0.0084american 0.0074general 0.0074view 0.0074approach 0.0073
!20
Topic1
Topic2
Topic3
Topic4
Topic5
Topic6
Topic7
Topic8
Topic9
Topic10
Topic11
Topic12
Topic13
Topic14
Topic15
Topic16
Topic17
Topic18
Topic19
Topic20
Topic21
Topic22
Topic23
Topic24
Topic25
Topic26
Topic27
Topic28
Topic29
Topic30
0.00
0.05
0.10
0.15
0.20
0.25
document-topicdistribution(θ)
Robert L., Ohsfeldt (2003):If the "Business Model" of Medicine Is Sick, What's the Diagnosis, and
The Independent Review
(preliminary) results II: some examples
!20
Topic 15 Weight (ϕ)health 0.0199insurance 0.0166care 0.0164services 0.0146service 0.0109cost 0.0108quality 0.0101medical 0.01coverage 0.0096premium 0.0093risk 0.0085airline 0.0077insurers 0.0077airlines 0.0074plans 0.0068
Topic1
Topic2
Topic3
Topic4
Topic5
Topic6
Topic7
Topic8
Topic9
Topic10
Topic11
Topic12
Topic13
Topic14
Topic15
Topic16
Topic17
Topic18
Topic19
Topic20
Topic21
Topic22
Topic23
Topic24
Topic25
Topic26
Topic27
Topic28
Topic29
Topic30
0.00
0.05
0.10
0.15
0.20
0.25
document-topicdistribution(θ)
Robert L., Ohsfeldt (2003):If the "Business Model" of Medicine Is Sick, What's the Diagnosis, and
The Independent Review
(preliminary) results II: some examples
!20
Topic 15 Weight (ϕ)health 0.0199insurance 0.0166care 0.0164services 0.0146service 0.0109cost 0.0108quality 0.0101medical 0.01coverage 0.0096premium 0.0093risk 0.0085airline 0.0077insurers 0.0077airlines 0.0074plans 0.0068
Topic1
Topic2
Topic3
Topic4
Topic5
Topic6
Topic7
Topic8
Topic9
Topic10
Topic11
Topic12
Topic13
Topic14
Topic15
Topic16
Topic17
Topic18
Topic19
Topic20
Topic21
Topic22
Topic23
Topic24
Topic25
Topic26
Topic27
Topic28
Topic29
Topic30
0.00
0.05
0.10
0.15
0.20
0.25
document-topicdistribution(θ)
Robert L., Ohsfeldt (2003):If the "Business Model" of Medicine Is Sick, What's the Diagnosis, and
The Independent Review
(preliminary) results II: some examples• Topic15: health care and insurance
!20
Topic 15 Weight (ϕ)health 0.0199insurance 0.0166care 0.0164services 0.0146service 0.0109cost 0.0108quality 0.0101medical 0.01coverage 0.0096premium 0.0093risk 0.0085airline 0.0077insurers 0.0077airlines 0.0074plans 0.0068
Topic1
Topic2
Topic3
Topic4
Topic5
Topic6
Topic7
Topic8
Topic9
Topic10
Topic11
Topic12
Topic13
Topic14
Topic15
Topic16
Topic17
Topic18
Topic19
Topic20
Topic21
Topic22
Topic23
Topic24
Topic25
Topic26
Topic27
Topic28
Topic29
Topic30
0.00
0.05
0.10
0.15
0.20
0.25
document-topicdistribution(θ)
Robert L., Ohsfeldt (2003):If the "Business Model" of Medicine Is Sick, What's the Diagnosis, and
The Independent Review
(preliminary) results II: some examples• Topic15: health care and insurance
Topic 12 Weight (ϕ)price 0.0213prices 0.0198product 0.0191sales 0.0172market 0.0172consumer 0.0171costs 0.0161consumers 0.0157cost 0.0142competitive 0.0135demand 0.0135firms 0.0134markets 0.0131firm 0.013products 0.013
!20
Topic 15 Weight (ϕ)health 0.0199insurance 0.0166care 0.0164services 0.0146service 0.0109cost 0.0108quality 0.0101medical 0.01coverage 0.0096premium 0.0093risk 0.0085airline 0.0077insurers 0.0077airlines 0.0074plans 0.0068
Topic1
Topic2
Topic3
Topic4
Topic5
Topic6
Topic7
Topic8
Topic9
Topic10
Topic11
Topic12
Topic13
Topic14
Topic15
Topic16
Topic17
Topic18
Topic19
Topic20
Topic21
Topic22
Topic23
Topic24
Topic25
Topic26
Topic27
Topic28
Topic29
Topic30
0.00
0.05
0.10
0.15
0.20
0.25
document-topicdistribution(θ)
Robert L., Ohsfeldt (2003):If the "Business Model" of Medicine Is Sick, What's the Diagnosis, and
The Independent Review
Topic 16 Weight (ϕ)countries 0.0178international 0.0161country 0.0142foreign 0.0139trade 0.0128world 0.0122domestic 0.012policy 0.0081export 0.008national 0.0079european 0.0077development 0.0077growth 0.0073economy 0.0073sector 0.0072
(preliminary) results II: some examples• Topic15: health care and insurance
Topic 12 Weight (ϕ)price 0.0213prices 0.0198product 0.0191sales 0.0172market 0.0172consumer 0.0171costs 0.0161consumers 0.0157cost 0.0142competitive 0.0135demand 0.0135firms 0.0134markets 0.0131firm 0.013products 0.013
!20
Topic 15 Weight (ϕ)health 0.0199insurance 0.0166care 0.0164services 0.0146service 0.0109cost 0.0108quality 0.0101medical 0.01coverage 0.0096premium 0.0093risk 0.0085airline 0.0077insurers 0.0077airlines 0.0074plans 0.0068
Topic1
Topic2
Topic3
Topic4
Topic5
Topic6
Topic7
Topic8
Topic9
Topic10
Topic11
Topic12
Topic13
Topic14
Topic15
Topic16
Topic17
Topic18
Topic19
Topic20
Topic21
Topic22
Topic23
Topic24
Topic25
Topic26
Topic27
Topic28
Topic29
Topic30
0.00
0.05
0.10
0.15
0.20
0.25
document-topicdistribution(θ)
Robert L., Ohsfeldt (2003):If the "Business Model" of Medicine Is Sick, What's the Diagnosis, and
The Independent Review
Topic 16 Weight (ϕ)countries 0.0178international 0.0161country 0.0142foreign 0.0139trade 0.0128world 0.0122domestic 0.012policy 0.0081export 0.008national 0.0079european 0.0077development 0.0077growth 0.0073economy 0.0073sector 0.0072
(preliminary) results II: some examples• Topic15: health care and insurance
Topic 12 Weight (ϕ)price 0.0213prices 0.0198product 0.0191sales 0.0172market 0.0172consumer 0.0171costs 0.0161consumers 0.0157cost 0.0142competitive 0.0135demand 0.0135firms 0.0134markets 0.0131firm 0.013products 0.013
Topic 25 Weight (ϕ)public 0.0219government 0.0209tax 0.019policy 0.0164private 0.016income 0.0148benefits 0.0142costs 0.0108taxes 0.0103policies 0.0102social 0.01level 0.0099efficiency 0.0099budget 0.0091economics 0.0088
!20
Topic 15 Weight (ϕ)health 0.0199insurance 0.0166care 0.0164services 0.0146service 0.0109cost 0.0108quality 0.0101medical 0.01coverage 0.0096premium 0.0093risk 0.0085airline 0.0077insurers 0.0077airlines 0.0074plans 0.0068
Topic 29 Weight (ϕ)data 0.0151table 0.0131results 0.0126significant 0.0098variables 0.0094variable 0.0092average 0.0092using 0.0089empirical 0.0087evidence 0.0085level 0.0084effects 0.008used 0.0079effect 0.0079model 0.0078
Topic1
Topic2
Topic3
Topic4
Topic5
Topic6
Topic7
Topic8
Topic9
Topic10
Topic11
Topic12
Topic13
Topic14
Topic15
Topic16
Topic17
Topic18
Topic19
Topic20
Topic21
Topic22
Topic23
Topic24
Topic25
Topic26
Topic27
Topic28
Topic29
Topic30
0.00
0.05
0.10
0.15
0.20
0.25
document-topicdistribution(θ)
Robert L., Ohsfeldt (2003):If the "Business Model" of Medicine Is Sick, What's the Diagnosis, and
The Independent Review
Topic 16 Weight (ϕ)countries 0.0178international 0.0161country 0.0142foreign 0.0139trade 0.0128world 0.0122domestic 0.012policy 0.0081export 0.008national 0.0079european 0.0077development 0.0077growth 0.0073economy 0.0073sector 0.0072
(preliminary) results II: some examples• Topic15: health care and insurance
Topic 12 Weight (ϕ)price 0.0213prices 0.0198product 0.0191sales 0.0172market 0.0172consumer 0.0171costs 0.0161consumers 0.0157cost 0.0142competitive 0.0135demand 0.0135firms 0.0134markets 0.0131firm 0.013products 0.013
Topic 25 Weight (ϕ)public 0.0219government 0.0209tax 0.019policy 0.0164private 0.016income 0.0148benefits 0.0142costs 0.0108taxes 0.0103policies 0.0102social 0.01level 0.0099efficiency 0.0099budget 0.0091economics 0.0088
!21
Topic 15 Weight (ϕ)health 0.0199insurance 0.0166care 0.0164services 0.0146service 0.0109cost 0.0108quality 0.0101medical 0.01coverage 0.0096premium 0.0093risk 0.0085airline 0.0077insurers 0.0077airlines 0.0074plans 0.0068
Topic1
Topic2
Topic3
Topic4
Topic5
Topic6
Topic7
Topic8
Topic9
Topic10
Topic11
Topic12
Topic13
Topic14
Topic15
Topic16
Topic17
Topic18
Topic19
Topic20
Topic21
Topic22
Topic23
Topic24
Topic25
Topic26
Topic27
Topic28
Topic29
Topic30
0.00
0.05
0.10
0.15
0.20
0.25document-topicdistribution(θ)
Karen, Eggleston (2000):Risk Selection and Optimal Health Insurance-Provider Payment Systems
The Journal of Risk and Insurance
(preliminary) results II: some examples• Topic15: health care and insurance
!21
Topic 15 Weight (ϕ)health 0.0199insurance 0.0166care 0.0164services 0.0146service 0.0109cost 0.0108quality 0.0101medical 0.01coverage 0.0096premium 0.0093risk 0.0085airline 0.0077insurers 0.0077airlines 0.0074plans 0.0068
Topic1
Topic2
Topic3
Topic4
Topic5
Topic6
Topic7
Topic8
Topic9
Topic10
Topic11
Topic12
Topic13
Topic14
Topic15
Topic16
Topic17
Topic18
Topic19
Topic20
Topic21
Topic22
Topic23
Topic24
Topic25
Topic26
Topic27
Topic28
Topic29
Topic30
0.00
0.05
0.10
0.15
0.20
0.25document-topicdistribution(θ)
Karen, Eggleston (2000):Risk Selection and Optimal Health Insurance-Provider Payment Systems
The Journal of Risk and Insurance
Topic 3 Weight (ϕ)model 0.0078consider 0.0077equilibrium 0.0077section 0.0068given 0.0068second 0.0064case 0.0064assume 0.0063proposition 0.0062following 0.0061function 0.0057results 0.0057optimal 0.0056result 0.0056game 0.0055
(preliminary) results II: some examples• Topic15: health care and insurance
!21
Topic 15 Weight (ϕ)health 0.0199insurance 0.0166care 0.0164services 0.0146service 0.0109cost 0.0108quality 0.0101medical 0.01coverage 0.0096premium 0.0093risk 0.0085airline 0.0077insurers 0.0077airlines 0.0074plans 0.0068
Topic1
Topic2
Topic3
Topic4
Topic5
Topic6
Topic7
Topic8
Topic9
Topic10
Topic11
Topic12
Topic13
Topic14
Topic15
Topic16
Topic17
Topic18
Topic19
Topic20
Topic21
Topic22
Topic23
Topic24
Topic25
Topic26
Topic27
Topic28
Topic29
Topic30
0.00
0.05
0.10
0.15
0.20
0.25document-topicdistribution(θ)
Karen, Eggleston (2000):Risk Selection and Optimal Health Insurance-Provider Payment Systems
The Journal of Risk and Insurance
Topic 3 Weight (ϕ)model 0.0078consider 0.0077equilibrium 0.0077section 0.0068given 0.0068second 0.0064case 0.0064assume 0.0063proposition 0.0062following 0.0061function 0.0057results 0.0057optimal 0.0056result 0.0056game 0.0055
(preliminary) results II: some examples• Topic15: health care and insurance
Topic 28 Weight (ϕ)demand 0.0101marginal 0.0099function 0.009output 0.0088price 0.0088equilibrium 0.0081cost 0.0081given 0.0076equal 0.0072constant 0.0071case 0.0071assumption 0.007assumed 0.007production 0.0069analysis 0.0066
!21
Topic 15 Weight (ϕ)health 0.0199insurance 0.0166care 0.0164services 0.0146service 0.0109cost 0.0108quality 0.0101medical 0.01coverage 0.0096premium 0.0093risk 0.0085airline 0.0077insurers 0.0077airlines 0.0074plans 0.0068
Topic1
Topic2
Topic3
Topic4
Topic5
Topic6
Topic7
Topic8
Topic9
Topic10
Topic11
Topic12
Topic13
Topic14
Topic15
Topic16
Topic17
Topic18
Topic19
Topic20
Topic21
Topic22
Topic23
Topic24
Topic25
Topic26
Topic27
Topic28
Topic29
Topic30
0.00
0.05
0.10
0.15
0.20
0.25document-topicdistribution(θ)
Karen, Eggleston (2000):Risk Selection and Optimal Health Insurance-Provider Payment Systems
The Journal of Risk and Insurance
Topic 3 Weight (ϕ)model 0.0078consider 0.0077equilibrium 0.0077section 0.0068given 0.0068second 0.0064case 0.0064assume 0.0063proposition 0.0062following 0.0061function 0.0057results 0.0057optimal 0.0056result 0.0056game 0.0055
(preliminary) results II: some examples• Topic15: health care and insurance
Question to the audience: Whats the difference between these two topics?
Some suggestions?
Topic 28 Weight (ϕ)demand 0.0101marginal 0.0099function 0.009output 0.0088price 0.0088equilibrium 0.0081cost 0.0081given 0.0076equal 0.0072constant 0.0071case 0.0071assumption 0.007assumed 0.007production 0.0069analysis 0.0066
!22
(preliminary) results II: some examples
!22
Topic 22 Weight (ϕ)anthony 0.0087morris 0.0066occurring 0.0064klaus 0.0053capturing 0.004observers 0.0036tighter 0.0036chowdhury 0.0034tai 0.003phillip 0.0025stayed 0.0025quiet 0.0023vic 0.0023fication 0.0023stacking 0.0021
(preliminary) results II: some examples
!22
Topic 22 Weight (ϕ)anthony 0.0087morris 0.0066occurring 0.0064klaus 0.0053capturing 0.004observers 0.0036tighter 0.0036chowdhury 0.0034tai 0.003phillip 0.0025stayed 0.0025quiet 0.0023vic 0.0023fication 0.0023stacking 0.0021
(preliminary) results II: some examples• Topic22:???
!22
Topic 22 Weight (ϕ)anthony 0.0087morris 0.0066occurring 0.0064klaus 0.0053capturing 0.004observers 0.0036tighter 0.0036chowdhury 0.0034tai 0.003phillip 0.0025stayed 0.0025quiet 0.0023vic 0.0023fication 0.0023stacking 0.0021
Topic1
Topic2
Topic3
Topic4
Topic5
Topic6
Topic7
Topic8
Topic9
Topic10
Topic11
Topic12
Topic13
Topic14
Topic15
Topic16
Topic17
Topic18
Topic19
Topic20
Topic21
Topic22
Topic23
Topic24
Topic25
Topic26
Topic27
Topic28
Topic29
Topic30
0.00
0.05
0.10
0.15
0.20
document-topicdistribution(θ)
(1995):Bibliography of Reinhard Selten's PublicationsThe Scandinavian Journal of Economics
(preliminary) results II: some examples• Topic22:???
!22
Topic 22 Weight (ϕ)anthony 0.0087morris 0.0066occurring 0.0064klaus 0.0053capturing 0.004observers 0.0036tighter 0.0036chowdhury 0.0034tai 0.003phillip 0.0025stayed 0.0025quiet 0.0023vic 0.0023fication 0.0023stacking 0.0021
Topic1
Topic2
Topic3
Topic4
Topic5
Topic6
Topic7
Topic8
Topic9
Topic10
Topic11
Topic12
Topic13
Topic14
Topic15
Topic16
Topic17
Topic18
Topic19
Topic20
Topic21
Topic22
Topic23
Topic24
Topic25
Topic26
Topic27
Topic28
Topic29
Topic30
0.00
0.05
0.10
0.15
0.20
document-topicdistribution(θ)
(1995):Bibliography of Reinhard Selten's PublicationsThe Scandinavian Journal of Economics
(preliminary) results II: some examples• Topic22:???
Topic 18 Weight (ϕ)theory 0.0204economics 0.0182analysis 0.0126press 0.0118university 0.0118journal 0.0112economists 0.0096economy 0.0094york 0.0089cambridge 0.0088new 0.0084american 0.0074general 0.0074view 0.0074approach 0.0073
Summary & Outlook
!23
Summary & Outlook
!23
• (Preliminary) results of the descriptive statistics show that the discourseon competition
Summary & Outlook
!23
• (Preliminary) results of the descriptive statistics show that the discourseon competition
‣ Starts with a time lag of about 10-15 years to intensify compared tothe overall development of economic research output
Summary & Outlook
!23
• (Preliminary) results of the descriptive statistics show that the discourseon competition
‣ Starts with a time lag of about 10-15 years to intensify compared tothe overall development of economic research output
‣ Is highly concentrated in terms of publication outlets: a third of allarticles in the sample are published in 10 journals
Summary & Outlook
!23
• (Preliminary) results of the descriptive statistics show that the discourseon competition
‣ Starts with a time lag of about 10-15 years to intensify compared tothe overall development of economic research output
‣ Is highly concentrated in terms of publication outlets: a third of allarticles in the sample are published in 10 journals
‣ Is – given the marginalization of heterodox approaches withineconomics – somewhat balanced: a substantial part of the literaturerelated to competition is published in (a few) heterodox journals
Summary & Outlook
!24
Summary & Outlook
!24
• Results of the first topic modeling exercise are promising
Summary & Outlook
!24
• Results of the first topic modeling exercise are promising
‣ A bird’s eye view on the overall corpus of research articles delivers atopic structure which links competition to various meaningful topicsthat emerge within the literature.
Summary & Outlook
!24
• Results of the first topic modeling exercise are promising
‣ A bird’s eye view on the overall corpus of research articles delivers atopic structure which links competition to various meaningful topicsthat emerge within the literature.
‣ Analysis of the topic-document distributions enables
Summary & Outlook
!24
• Results of the first topic modeling exercise are promising
‣ A bird’s eye view on the overall corpus of research articles delivers atopic structure which links competition to various meaningful topicsthat emerge within the literature.
‣ Analysis of the topic-document distributions enables
‣ a reconstruction of the genesis of specific topics (and itsdisappearance)
Summary & Outlook
!24
• Results of the first topic modeling exercise are promising
‣ A bird’s eye view on the overall corpus of research articles delivers atopic structure which links competition to various meaningful topicsthat emerge within the literature.
‣ Analysis of the topic-document distributions enables
‣ a reconstruction of the genesis of specific topics (and itsdisappearance)
‣ an analysis of co-occurrences between specific topics that relate tocompetition
Summary & Outlook
!25
Summary & Outlook
!25
• However, model parameters still need to be optimized
Summary & Outlook
!25
• However, model parameters still need to be optimized
‣ Number of Topics (more vs. less)
Summary & Outlook
!25
• However, model parameters still need to be optimized
‣ Number of Topics (more vs. less)
‣ Different Dirichlet priors (see also Tang et al. 2014, Wallach et al. 2009)
Summary & Outlook
!25
• However, model parameters still need to be optimized
‣ Number of Topics (more vs. less)
‣ Different Dirichlet priors (see also Tang et al. 2014, Wallach et al. 2009)
‣ ⍺ = small: each document is associated with few topics
Summary & Outlook
!25
• However, model parameters still need to be optimized
‣ Number of Topics (more vs. less)
‣ Different Dirichlet priors (see also Tang et al. 2014, Wallach et al. 2009)
‣ ⍺ = small: each document is associated with few topics
‣ β = small: topics are word sparse (large β means more word-diffused and similar topics)
Summary & Outlook
!25
• However, model parameters still need to be optimized
‣ Number of Topics (more vs. less)
‣ Different Dirichlet priors (see also Tang et al. 2014, Wallach et al. 2009)
‣ ⍺ = small: each document is associated with few topics
‣ β = small: topics are word sparse (large β means more word-diffused and similar topics)
‣ Symmetrical vs. asymmetrical priors
Summary & Outlook
!25
• However, model parameters still need to be optimized
‣ Number of Topics (more vs. less)
‣ Different Dirichlet priors (see also Tang et al. 2014, Wallach et al. 2009)
‣ ⍺ = small: each document is associated with few topics
‣ β = small: topics are word sparse (large β means more word-diffused and similar topics)
‣ Symmetrical vs. asymmetrical priors
‣ Text preprocessing (see e.g. Schofield et al. 2017)
Summary & Outlook
!25
• However, model parameters still need to be optimized
‣ Number of Topics (more vs. less)
‣ Different Dirichlet priors (see also Tang et al. 2014, Wallach et al. 2009)
‣ ⍺ = small: each document is associated with few topics
‣ β = small: topics are word sparse (large β means more word-diffused and similar topics)
‣ Symmetrical vs. asymmetrical priors
‣ Text preprocessing (see e.g. Schofield et al. 2017)
‣ “Pseudo-abstracts” vs. full-text analysis?
Summary & Outlook
!25
• However, model parameters still need to be optimized
‣ Number of Topics (more vs. less)
‣ Different Dirichlet priors (see also Tang et al. 2014, Wallach et al. 2009)
‣ ⍺ = small: each document is associated with few topics
‣ β = small: topics are word sparse (large β means more word-diffused and similar topics)
‣ Symmetrical vs. asymmetrical priors
‣ Text preprocessing (see e.g. Schofield et al. 2017)
‣ “Pseudo-abstracts” vs. full-text analysis?
‣ Remove standard stop-words vs. topic-specific stop-words?
Summary & Outlook
!25
• However, model parameters still need to be optimized
‣ Number of Topics (more vs. less)
‣ Different Dirichlet priors (see also Tang et al. 2014, Wallach et al. 2009)
‣ ⍺ = small: each document is associated with few topics
‣ β = small: topics are word sparse (large β means more word-diffused and similar topics)
‣ Symmetrical vs. asymmetrical priors
‣ Text preprocessing (see e.g. Schofield et al. 2017)
‣ “Pseudo-abstracts” vs. full-text analysis?
‣ Remove standard stop-words vs. topic-specific stop-words?
‣ Word stemming or not?
Summary & Outlook
!26
Summary & Outlook
!26
• Possible next steps are
Summary & Outlook
!26
• Possible next steps are
‣ Analyzing the corresponding bibliometric meta-data
Summary & Outlook
!26
• Possible next steps are
‣ Analyzing the corresponding bibliometric meta-data
‣ via publication year: when do these topics emerge/disappear?
Summary & Outlook
!26
• Possible next steps are
‣ Analyzing the corresponding bibliometric meta-data
‣ via publication year: when do these topics emerge/disappear?
‣ via publication outlet: where do these topics emerge/disappear?
Summary & Outlook
!26
• Possible next steps are
‣ Analyzing the corresponding bibliometric meta-data
‣ via publication year: when do these topics emerge/disappear?
‣ via publication outlet: where do these topics emerge/disappear?
‣ via cited references: what are the conceptual foundations of atopic?
Summary & Outlook
!26
• Possible next steps are
‣ Analyzing the corresponding bibliometric meta-data
‣ via publication year: when do these topics emerge/disappear?
‣ via publication outlet: where do these topics emerge/disappear?
‣ via cited references: what are the conceptual foundations of atopic?
‣ A more nuanced analysis / Extensions
Summary & Outlook
!26
• Possible next steps are
‣ Analyzing the corresponding bibliometric meta-data
‣ via publication year: when do these topics emerge/disappear?
‣ via publication outlet: where do these topics emerge/disappear?
‣ via cited references: what are the conceptual foundations of atopic?
‣ A more nuanced analysis / Extensions
‣ Extracting topics from period/decade subsamples (e.g. 1960s,1970s,… see also Ambrosino et al. 2018)
Summary & Outlook
!26
• Possible next steps are
‣ Analyzing the corresponding bibliometric meta-data
‣ via publication year: when do these topics emerge/disappear?
‣ via publication outlet: where do these topics emerge/disappear?
‣ via cited references: what are the conceptual foundations of atopic?
‣ A more nuanced analysis / Extensions
‣ Extracting topics from period/decade subsamples (e.g. 1960s,1970s,… see also Ambrosino et al. 2018)
‣ Replication for other fields (e.g. Sociology, Political Science,Anthropology)
!27
Many thanks for your attention