While following the news, one can notice the same story can
have different impact depending on which news agent tells
it. One reason for this is how the facts are framed. Framing is described by communication sciences as an instrument
influencing on how people perceive, interpret and convey information. It can be obtained by use of specific word choice
and labeling that describe event or problem from a particular perspective, e.g. positive or negative. In order to derive a frame, social sciences usually perform a manual qualitative analysis, but recently a computer-assist quantitative
approaches commence to be an essential way of conducting
framing analysis. This work provides a literature review on
the existing frame derivation methods based on problem of
word choice and labeling.
Pests of mustard_Identification_Management_Dr.UPR.pdf
Putting News in a Perspective: Framing by Word Choice and Labeling
1. Putting News in a Perspective: Framing
by Word Choice and Labeling
Student: Anastasia Zhukova
Supervisor: Felix Hamborg
Examiner: Prof. Dr. Bela Gipp
Date: 2017-02-07
07/02/2017 Putting News in a Perspective: Framing by Word Choice and Labeling
2. Agenda
1. Introduction: What is a frame?
2. Framing analysis
3. Research questions
4. Inductive approaches
5. Deductive approaches
6. Discussion of results
7. Future work
8. Conclusion
07/02/2017 Putting News in a Perspective: Framing by Word Choice and Labeling 2
3. What is a frame?
07/02/2017 Putting News in a Perspective: Framing by Word Choice and Labeling 3
Framing is a conceptualization of an issue, the way how people organize, perceive,
and communicate information.
Example:
After this procedure 90% of patients are alive
After this procedure 10% of patients are deceased
The information in the two sentences is the same.
4. What is a frame?
07/02/2017 4
➢ A frame is central organizing idea, that selects and emphasizes information.
➢ Framing devices are salient indicators of a frame
• Word choice
• Metaphors
• Catchphrases
• Visual images
• Stereotypes
• Labels
• Etc.
➢ Framing is a type of media bias.
➢ It influences on how people interpret and use information.
Putting News in a Perspective: Framing by Word Choice and Labeling
Frame
5. Word choice and labeling
07/02/2017 Putting News in a Perspective: Framing by Word Choice and Labeling 5
Word choice:
• Influences on the message, changing its perception
Example:
“Heart-wrenching tales of hardship faced by people
whose care is dependent on Medicaid”
VS
“Information on the lifestyles of Medicaid
dependents”
Labeling:
• Describes someone or something in one word
• Refers to stereotypes
Example:
political party (democrat/republican), religion (Muslim/Christian), race
(black/white), etc.
6. 07/02/2017 Putting News in a Perspective: Framing by Word Choice and Labeling 6
Framing analysis
Qualitative
analysis
Quantitative
analysis
Text meaning Statistical features
Manual
analysis
Computer-assisted
analysis
To do framing analysis is to identify framing devices, which describe an issue.
7. 07/02/2017 Putting News in a Perspective: Framing by Word Choice and Labeling 7
Framing analysis
Agenda-setting
“What to think about”
Framing
“How to think about it”
Issue Frame 2
Frame 1
Frame 3
Framing device Reasoning device
8. 07/02/2017 Putting News in a Perspective: Framing by Word Choice and Labeling 8
Framing analysis
Inductive analysis Deductive analysis
Derive frames from texts Find given frames in texts
Frame
News
Frame
News Found
frames in
news
10. Inductive approaches
07/02/2017 Putting News in a Perspective: Framing by Word Choice and Labeling 10
How to find frames?
Most frequent
words
Correlation
between words
[1]
• Cosine similarity matrix
• PCA
• Hierarchical clustering
Word
co-occurrence
[4]
• Semantic network
• Cosine similarity matrix
• Value ≥ threshold
node in a network
[3]
• Word = neuron
• Self-organizing map
• Hierarchical
clustering
[2]
• Covariance matrix
• PCA
• Significant eigenvalues
frames
11. Inductive approaches
07/02/2017 Putting News in a Perspective: Framing by Word Choice and Labeling 11
Most frequent
words
…
Text structure
Center resonance
analysis
[5]
• Noun phrases = idea
centers
• Word co-occurrence
resonance of
words
• PCA
Keyword-weight
model
[6]
• Weight words according
to location of
occurrence
• Location = place
in news pyramid
•Aspect-based clustering
How to find frames?
12. Inductive approaches
07/02/2017 Putting News in a Perspective: Framing by Word Choice and Labeling 12
Most frequent
words
…
Text structure
…
Semantic
meaning
[7]
• Latent semantic analysis
• High cosine similarity
words with similar
meaning
Keywords
[8]
• Log-likelihood ratio of words in
a list w.r.t. reference word-list
• Word co-occurrence
• Qualitative analysis
Named entity
recognition
[9]
• Extract different entity
categories
• Find salient terms
• Qualitative analysis
How to find frames?
13. Deductive approaches
07/02/2017 Putting News in a Perspective: Framing by Word Choice and Labeling 13
Classification
How to find presence of frames?
Logistic regression
Ensemble of
logistic
regressions
Supervised
Hierarchical LDA
[10]
• Binary classifier
• 1 classifier = 1 frame
• Features: based on word’s
presence/absence in a
vocabulary
• Prediction = measure of frame
presence
[11]
• Collection of classifiers
• 1 classifier = 1 framing device
• Features: based on TF-IDF score
• Prediction = measure of frame
device presence
[12]
• Probabilistic model
• Forms frames around
assigned label
14. Deductive approaches
07/02/2017 Putting News in a Perspective: Framing by Word Choice and Labeling 14
How to find presence of frames?
Classification
…
Homogeneity analysis
Clustering Rule-based approach
[13]
• Clusters = frames
• Features = based on coding
agreement of framing
device
• Hierarchical clustering
[14]
• Homogeneity analysis
most influential devices in
each frame
• Final result for a text: 2
indices of 2 frames’
presence
[15]
• Semantic Network analysis
• Framing devices = sentences
• Texts = network of sentences
• Sentence is parsed into 3
roles
15. Discussion
RQ1: How do scholars approach computer-assisted framing analysis?
07/02/2017 Putting News in a Perspective: Framing by Word Choice and Labeling 15
Inductive Deductive
Word Frequency
• Correlation
• Co-occurrence
Classification
Text Structure
• Central Resonance
Analysis
• Keyword-weight model
Homogeneity analysis
Latent Semantic Analysis Clustering
Keywords Rule-based approach
Named Entity Recognition
16. Discussion
RQ2: What are the methods that focus analysis on constructing frames based on
word choice and labeling?
1. Keyword-weight model word choice is connected with the most prominent
parts on news
2. LSA semantic similarity of words could addressed word choice influence
3. LSA semantic similarity of labels w.r.t. the context
4. Named Entity Extraction salience topics, framing devices, and if extended,
labels
5. Neural Network concentration around salience concepts. Log-likelihood
instead of frequent words in an option.
6. Current methods don’t capture “between line” information
7. Frames representing same entity from several perspectives are not united to
one concept.
07/02/2017 Putting News in a Perspective: Framing by Word Choice and Labeling 16
17. Future work
1. LDA as topic modeling could be used for inductive analysis
2. Review other keyword or keyphrase extraction methods
3. Use named entity extraction for framing devices identification
07/02/2017 Putting News in a Perspective: Framing by Word Choice and Labeling 17
18. Conclusion
Performed a comprehensive literature review. Current approaches:
+ Focus on interpretation of frame definition from social sciences
+ Address both inductive and deductive analysis
+ Frames are based on semantic and structural properties of text
- Mostly semi-automated analyses
- Represent more agenda-setting analysis rather than framing
- Construction of framing devices is not solved
07/02/2017 Putting News in a Perspective: Framing by Word Choice and Labeling 18
19. References
[1] Miller, M. Mark. "Frame mapping and analysis of news coverage of contentious issues." Social Science
Computer Review 15.4 (1997): 367-378.
[2] Crawley, Catherine E. "Localized debates of agricultural biotechnology in community newspapers: A
quantitative content analysis of media frames and sources." Science Communication 28.3 (2007): 314-346.
[3] Tian, Yan, and Concetta M. Stewart. "Framing the SARS crisis: A computer-assisted text analysis of CNN
and BBC online news reports of SARS." Asian Journal of Communication 15.3 (2005): 289-301.
[4] Hellsten, Iina, James Dawson, and Loet Leydesdorff. "Implicit media frames: Automated analysis of public
debate on artificial sweeteners." Public Understanding of Science 19.5 (2010): 590-608.
[5] Papacharissi, Zizi, and Maria de Fatima Oliveira. "News frames terrorism: A comparative analysis of
frames employed in terrorism coverage in US and UK newspapers." The International Journal of
Press/Politics 13.1 (2008): 52-74.
[6] Park, Souneil, et al. "NewsCube: delivering multiple aspects of news to mitigate media bias." Proceedings
of the SIGCHI Conference on Human Factors in Computing Systems. ACM, 2009.
[7] Sendén, Marie Gustafsson, Sverker Sikström, and Torun Lindholm. "“She” and “He” in news media
messages: pronoun use reflects gender biases in semantic contexts." Sex Roles 72.1-2 (2015): 40-49.
[8] Touri, Maria, and Nelya Koteyko. "Using corpus linguistic software in the extraction of news frames:
towards a dynamic process of frame analysis in journalistic texts." International Journal of Social Research
Methodology 18.6 (2015): 601-616.
[9] Ananiadou, S., et al. "Supporting frame analysis using text mining." 5 th International Conference on e-
Social Science. 2009.
07/02/2017 Putting News in a Perspective: Framing by Word Choice and Labeling 19
20. References
[10] Boydstun, Amber E., et al. "Tracking the Development of Media Frames within and across Policy Issues."
(2014).
[11] Odijk, Daan, et al. "Automatic thematic content analysis: Finding frames in news." International
Conference on Social Informatics. Springer International Publishing, 2013.
[12] Nguyen, Viet-An, Jordan L. Boyd-Graber, and Philip Resnik. "Lexical and hierarchical topic regression."
Advances in neural information processing systems. 2013.
[13] Matthes, Jörg, and Matthias Kohring. "The content analysis of media frames: Toward improving
reliability and validity." Journal of communication 58.2 (2008): 258-279.
[14] Van Gorp, Baldwin. "Where is the frame? Victims and intruders in the Belgian press coverage of the
asylum issue." European Journal of Communication 20.4 (2005): 484-507.
[15] van Atteveldt, Wouter, Tamir Sheafer, and Shaul Shenhav. "Automatically extracting frames from media
content using syntacting analysis." Proceedings of the 5th Annual ACM Web Science Conference. ACM, 2013.
Picture credits:
Slide 3: http-//www.anmbadiary.com/2015/04/framing-effect-and-marketing.html
Slide 3: http-//www.thesleuthjournal.com/wp-content/uploads/2014/02/the-western-mainstream-media-
at-work.jpg
Slide 5: https-//tomakeaprairie.wordpress.com/2014/05/11/
07/02/2017 Putting News in a Perspective: Framing by Word Choice and Labeling 20
21. Thank you for attention!
Questions?
07/02/2017 Putting News in a Perspective: Framing by Word Choice and Labeling 21
23. Inductive approaches
07/02/2017 Putting News in a Perspective: Framing by Word Choice and Labeling 23
№ Main methods Pre-processing Post-processing Summary
1
[1]
• Cosine similarity matrix of
co-occurring frequent terms
• PCA
• Hierarchical clustering on 3
eigenvector values
• stop words and ambiguous
words manually removed
• Number of most frequent
words chosen by authors
• 1 document = 1 list
• Cluster name = the
top most term in a
cluster
•Semi-A.
•Agenda s.
2
[2]
• PCA of most frequent
words with varimax
rotation, 8 most meaningful
eigenvalues selected
--||-- • Terms with loading
≥ 0.3 form a frame
• Frames are named
manually
•Semi-A.
•Agenda s.
3
[3]
• Self-organizing map of most
frequent words based on
neural network
• Hierarchical clustering
based on Ward’s method
• Stop words and verbs
removed
• 1 document = 1 list
• Top 40 words are manually
ranked
• Cluster name is
given manually
based on clustering
results
•Semi-A.
•Agenda s.
4
[4]
• Cosine similarity matrix of
co-occurring most frequent
terms
• Elements ≥ threshold form a
network
• Stop words removed
• All document of 1 media = 1
list
• Normalize
similarity
• Frames are
obtained by
manual
interpretation
•Semi-A.
•Agenda s.
24. Inductive approaches
07/02/2017 Putting News in a Perspective: Framing by Word Choice and Labeling 24
№ Main
methods
Pre-processing Post-processing Summary
5
[5]
Central resonance
analysis
•Only noun-phrases left,
other words form
connections
•Pronouns are dropped
•Stemming
Manually derived frame
names
- Semi-A.
- Agenda-s.
- Word Ch.
6
[7]
Latent Semantic
Analysis
•Stop-words removed Frame names are pregiven - A.
- Framing ±
- Word.Ch + possible
labeling
7
[6]
Keyword-weight
model
•Bag-of-words
•Stop-words removed
•Structure of text employed
No information about frame
names
- A.
- Framing
- Word choice
8
[8]
Keywords •Bag-of-words of an article
•Bag-of-words of all articles
Framing devices are
constructed fully manually
- Semi-A.
- Framing
- Word Ch. + labeling
9
[9]
Named entity
recognition
? ? - A.
- Framing ±
- Word.Ch + possible
labeling
25. Deductive approaches
07/02/2017 Putting News in a Perspective: Framing by Word Choice and Labeling 25
№ Main methods Pre-processing Post-processing Summary
1
[10]
Logistic regression •Manual frame coding
•Features: presence or
absence of each word
compared to code-
vocabulary
Obtained prediction is
used as a measure of
frame presence
- Semi-A.*
- Word-ch. + possibly
labeling**
2
[11]
Ensemble of logistic
regressions
•Manual frame coding
•Bag-of-words + TF-IDF score
•Sub linear term frequency
scaling + normalization
Obtained prediction is
used as a measure of
frame presence
- Semi-A.*
- Word-ch. + possibly
labeling**
3
[12]
Supervised
Hierarchical Latent
Dirichlet Allocation
•Obtain labels ? - A./Semi-A.*
- Word-ch. +
labeling**
4
[14]
Homogeneity
analysis
•Manual frame coding -- - Semi-A.*
- Word-ch. +
labeling**
5
[13]
Hierarchical
clustering
•Manual framing devices
coding
Obtained clusters should
be interpreted
- Semi-A.
- Word-ch. +
labeling**
6
[15]
Semantic network
analysis
•Parsed sentences -- - Semi-A.
- Word-ch. + labeling
*due to manual coding
** if labels have additional weight or are used in framing devices
26. Correlation between words
07/02/2017 Putting News in a Perspective: Framing by Word Choice and Labeling 26
• Meaningful frames found
• Agenda-setting
[1] [2]
27. Word co-occurrence
07/02/2017 Putting News in a Perspective: Framing by Word Choice and Labeling 27
• Agenda-setting
• Hard to distinguish and name frames in [4]
[3] [4]
28. Latent Semantic Analysis
07/02/2017 Putting News in a Perspective: Framing by Word Choice and Labeling 28
• Gendered word choice of “she” neighbors
• Possible use for labeling detection:
Label surrounding doesn’t correspond to it
Surrounding a label is implied
[7]
29. Central Resonance Analysis
07/02/2017 Putting News in a Perspective: Framing by Word Choice and Labeling 29
• Method reveals similarities on basic word choice
• Agenda-setting due to restriction to noun phrases
• Attention to word choice comparison
Washington Post’s
New York Times Post’s attack, September, terrorist
Al-Qaeda
Administration, President Bush
homeland, national, security
Afghanistan, Iraq, terrorism, war
Combatant, enemy
Islamic, radical
Al-Qaeda, leader, member, Iraq
policy, troop, United States, war
Political personas and characters =>
«elements that indicate
a more dramatic approach to
coverage»
[5]
30. Classification
07/02/2017 Putting News in a Perspective: Framing by Word Choice and Labeling 30
0.5
0.6
0.7
0.8
0.9
1
C1 C1 E1 E2 H1 H2 H3 H4 M1 M2 M3
Human
Ensemble
Single
classifier
[10]
[11]
[12]
Logistic regression
Ensemble of logistic regressions
SHLDA