a solution to determine radicalization preceding violence...
TRANSCRIPT
A Solution to Determine Radicalization Preceding Violence or
Peace in Iranian Discourse
Presented by
Vicki Nisbett, M.A.
New Mexico State University
Las Cruces, New Mexico
1
A Solution to Determine Radicalization toward Violence or Peace in Iranian Discourse
This white paper discusses why using Latent Semantic Analysis (LSA) will help
determine if there is radicalization in specific Iranian discourse. LSA gives a left to right
analysis of discourse over time and provides you with the tools and expertise to determine if
there is an escalation in violent discourse or peaceful discourse. LSA analyzes discourse no
matter what language is used.
Methods Discussion
Manual Castells argues that the capability of effectively engaging in intimidation or
violence “requires the framing of individual and collective minds.” 1
From the representationalist
view of language, Hacker presents in his essay on political linguistic discourse analysis that
traditional communication-theory scholars argue that “people learn word meanings by
associating words with objects and experiences.” 2
Castells and Lakoff argue that narratives,
which are built into frames, get intertwined in complex frames or related words (semantic fields)
linked to language.1,3
The frames are mapped in our brains; thus, we act upon our past and
present experiences (see Appendix A).1,3
Through decision-making, involving feelings,
emotions, and reasoning, we take action. Therefore, experience, and emotions are used to frame
political behavior.
Radicalization. The first step that can lead to violence that is twofold is radicalization,
according to Countering International Terrorism.4 (1) If someone is alienated from society, they
may adopt extreme views and become radicalized. A small minority of these individuals become
terrorists. Coming from a range of “potential factors,” there is no “single factor that
predominates” to cause radicalization. (2) Within the process of becoming “a radical” is a
“function of being exposed more frequently to a radical political ideology than to a more
conservative one.” 5
A differential socialization model shows that “radicalization is basically a
communication process.” 5
Data showed that any individual who holds radical political views has
a “simple linear function to the extent which he or she has been exposed preponderance favoring
such a view.” 5
Sprinzak notes that right-wing terrorist groups can reach terrorism via a “split
delegitimazation,” implying they have a major conflict with an inferior community and a minor
conflict with the government. 6
He lists several types of groups, but notes their commonality
which is not against governments and is not dedicated in the name of widespread values. These
groups are usually right-wing “collectivities, vigilante groups, or racist organizations” that do not
speak for humanity. By nature, they are particularistic and respond to discernments of insecurity
and threats, and, therefore, they fight their private wars against aggressive ethnic communities,
illicit religious denominations, and classes of undesired classes of people or substandard races. 6
In nation radicalization, Baham notes that the end of the nineteenth-Century Central Europe,
and the end of the twentieth Century, historically had rages of nationalism. Ever since then,
scholars try to understand nationalism and the enrollment of national identity, engrained in
beginnings of current social and political crisis.7 Kenneth Burke’s study on Hitler suggests that
political groups or nations can use framed language leaning toward violence to encourage
hostility in their listeners, especially when escalated messages are given over time.8 It is possible
that a political leader’s language, messages or speeches, can be analyzed preceding violent action
2
by his country or group. Burke noted that Hitler used formulaic expressions to successfully
increase German hostility to the Jewish people. Burke notes how Hitler played on the Germans’
crumpled emotional state, economic distress, and weakened religious belief structure from the
World War I defeats, using a number of formulaic devices. 8
How do you then determine
radicalization over time? LSA is a part of multi-level methods used in determining if there is
radicalization over time, but its contribution is valid with human intersession.
Our Solution
Latent semantic analysis (LSA) is a statistical approach to language analysis using
Singular Value Decomposition (SVD). According to Landauer, Foltz, and Laham, LSA is not
only a method used for extracting as well as representing contextual-usage word meaning
through statistical calculations applied to a great amount of corpus of text, but it is also a theory
as well as a method.9 The theory of LSA suggests that factor analysis can effectively model the
human language (Simon & Xenos, 2004). LSA is deemed artificial intelligence by many
scholars.
When using LSA, words, phrases, and paragraphs are copied from the corpus (body) of
text and then analyzed.9 LSA is related to factor analysis and is also “related to neural net
models,” brain circuitry computer simulations, and is “based on singular value decomposition
(SVD).”9 SVD is a mathematical method producing a semantic space arrangement, which
provides reflection of major comparative patterns in the data,10
SVD is used for data reduction.11
The result reduces the dimension of the semantic space, according to Landauer, Foltz, and
Laham, LSA represents the meaning of a word, from a sort of average of passages in which it
appears and simulates the connections of the human mind.9
Halliday Schilling notes that LSA can simulate many “human cognitive phenomena,”
such as recognizing “vocabulary to word-categorization,” “discourse comprehension,” and
“semantic priming” (meaning that a word is more easily recognized “when it is preceded by a
related stimulus rather than an unrelated stimulus.” 12
LSA can also make judgments of essay-
quality text.9 When analyzing documents, if each word only meant one concept and each concept
was only described by one word, the task would entail a simplistic mapping from words to
concepts.13
The LSA process. LSA searches through the corpus for words and text (terms or frames)
placed in the LSA command line.14
When LSA finds these terms or frames, it performs a
mathematical analysis of the relationship of the terms or frames to the corpus. LSA also finds
other words or text that has similar mathematical relationships to the corpus as the terms or
frames placed in the command line and not just the words that it searches for.
According to Landauer and Dumais, and LSA theory, words and text, which have similar
mathematical relationships within the corpus, also have similar meaning, even if they are totally
different terms or frames.15
LSA outputs numerical cosine values in the range of 0.0 to 1.0 that
reflect the similarity of the corpus to the provided terms or frames as well as other terms or
frames that have a similar mathematical relationship within the corpus.14
LSA output cosine values near 1.0 are the most similar and those near 0.0 are the least
similar. The corpus that is searched by LSA may contain many separate documents or source
information(s), which are separated by line spaces. This allows many documents to be analyzed
simultaneously.15
3
Figure 1. Example of highest 10% of sentence similarity for advocating violence
terms of Iranian politician’s speeches.
2005-2006, 14
2009, 9
2010-2011, 13
2005-2006, 10 2007-2008, 8
2010, 12
0
2
4
6
8
10
12
14
16
2005-2006 2007-2008 2010
Pe
rce
nta
ge o
f Se
nte
nce
Sim
ilari
tie
s
Year of Speeches
Percentage of Highest Similarity Sentences for Advocating Violence Terms
for Iranian Politicians: 2005-2006, 2009-2011
AhmadinejadKhamenei
Research for a left to right analysis. In 2012, our research shows by first determining
violent or peaceful discourse for comparison to other discourse, with minimal human
interpretation, LSA will find those terms as well as similar terms just as correctly as humans,
and, at times even better, analyzing the discourse in record time. The similarities can be sorted
into highest percentages and graphed easily showing outcomes over time from terms, sentences,
phrases, or documents. When these graphs are presented, the results are obvious and easy to
understand, as shown in figure 1.
Figure 1 clearly shows that sentence similarity increased in 2007 for the Supreme Leader
Khamenei and 2009 for President Ahmadinejad.16
Noting the last similarity increase peak in
2010 and 2011 for both politicians, research showed this is when Iran experienced their fourth
imposed Western sanction and the speeches were given after protesters had rioted because of the
presidential re-election of President Ahmadinejad.
What the iTeam can do for you. As a research team having an array of experience in
communication studies (including political, national security, intelligence, social network
systems, and cultural), linguistics, systems analysis, and computational analysis, we can put an
analysis plan into action showing the results you need now.
Note: Violent discourse frames were compared to two Iranian politician’s speeches.
There was escalation in violent discourse beginning in 2007 through 2011.16
4
Framed Metaphors
(Narratives). Pictures
in our Heads*
War is good
Real-world experience
Word association (War is good) *
Repeated real-world experience
Word association (War is good)*
Repeated real-world experience
Word association (War is good)*
Repeated real-world experience
Word association (War is good)*
and so on …
Real-world experiences with word association of positive words with violent frames lead to complex frames with the
possibility of preceding to violent action
Complex
Frames*
* Emotions and feelings: core values/belief system.
Castells (2009), Communication power. New York: Oxford
University Press.
Lakoff, G. (2008). The political mind. Why you can’t
understand 21st-century politics with an 18th-century brain.
The Penguin Group: New York, N.Y.
(We must have
war)
Possible
Violent
Action
5
Works Cited
1. Castells, M. (2009). Communication power. New York: Oxford University Press.
2. Hacker, K. (1996). Political linguistic discourse analysis. Analyzing the relationship of power
and language. Stuckey, M. (Ed.). Albany, NY: State University of New York Press.
3. Lakoff, G. (2008). The political mind. Why you can’t understand 21st-century politics
with an 18th
-century brain. The Penguin Group: New York, N.Y.
4. Countering international terrorism: The United Kingdom’s strategy (2006, July),
presented to the prime minister and the secretary of state for the home department by
command of her majesty. HM Government. (Cm 6888), retrieved from
http://www.cabinetoffice.gov.uk/media/136036/countering.pdf
5. Woelfel, Joseph, Woefel, John, Gillham, J., & McPhail, T. (1974). Political radicalization
as a communication process. Communication Research, 1(3). Thousand Oaks, CA:
Sage Publications, Inc.
6. Sprinzak, E. (1995). Right‐wing terrorism in a comparative perspective: The case of split
delegitimization. Terrorism and Political Violence. 7(1).
7. Baham, K. F. (1998). Beyond the Bourgeoisie: Rethinking Nation, Culture, and Modernity
in Nineteenth-Century Central Europe. Austrian History Yearbook, 29, 19-35
doi:10.1017/S0067237800014788
8. Burke, K. (1941). The Rhetoric of Hitler’s Battle. The philosophy in literary form: Studies
in symbolic action. 191-220.
9. Landauer, T. K., Foltz, P. W., & Laham, D. (1998). Introduction to latent semantic
analysis. Retrieved from
http://lsa.colorado.edu/
10. Serafin R., & Di Eugenio, B. (2004). FLSA: Extending Latent Semantic Analysis with
features for dialogue act classification. Retrieved from
http://dl.acm.org/citation.cfm?id=1219043
11. Baker, K. (2005). Singular value decomposition tutorial. Rough draft. Retrieved from
www.cs.wits.ac.za/~michael/SVDTut.pdf
12. Halliday Schilling, H.E. (1998). Semantic priming by words and pictures in lexical
decision and pronunciation tasks. Electronic doctoral dissertations for U Mass
Amherst. Paper AAI9909217. Abstract retrieved from
http://scholarworks.umass.edu/dissertations/AAI9909217
13. Latent semantic analysis tutorial (n.d.). Meta Search. Retrieved from
http://www.puffinwarellc.com/index.php/news-and-articles/articles/33-latent-semantic-
analysis-tutorial.html
14. Laham, D. (1997). The LSA at CU Guidebook. (2nd
draft version). Not openly
distributed.
15. Landauer, T. K., & Dumais, S. T. (1997). A solution to Plato's problem: The latent
semantic analysis theory of the acquisition, induction, and representation of
knowledge. Psychological Review, 104, 211-140.
16. Nisbett, V. (2012). The study of Iranian discourse preceding violent action or peace. Thesis,
Communication Studies, New Mexico State University.