=Paper=
{{Paper
|id=Vol-3290/short_paper6153
|storemode=property
|title=Right-wing Mnemonics
|pdfUrl=https://ceur-ws.org/Vol-3290/short_paper6153.pdf
|volume=Vol-3290
|authors=Phillip Stenmann Baun,Kristoffer Nielbo
|dblpUrl=https://dblp.org/rec/conf/chr/BaunN22
}}
==Right-wing Mnemonics==
https://ceur-ws.org/Vol-3290/short_paper6153.pdf
Right-wing Mnemonics
Phillip Stenmann Baun1 , Kristo昀昀er Nielbo2,3
1
Department of Global Studies, Aarhus University, 8000 Aarhus C, Denmark
2
Center for Humanities Computing Aarhus, Aarhus University, 8000 Aarhus C, Denmark
3
Interacting Minds Centre, Aarhus University, 8000 Aarhus C, Denmark
Abstract
This paper presents a natural language processing technique for studying memory on the far-right
political discussion forum /pol/ on 4chan.org. Memory and the use of history play a pivotal role on
the far-right for temporally structuring beliefs about social life and order. However, due in part to
methodological limitations, there is a lack of knowledge regarding the speci昀椀c historical entities that
make up the far-right memory culture and wider historiography. To better grasp the structure of far-
right memory, this paper opts for a data-intensive methodology, using machine learning on a data set of
approximately 66 million posts from /pol/ from 2020. 19,821 random posts were manually annotated,
according to the presence of historical entities. A昀琀er evaluating interrater reliability, data were used
to train a naïve Bayes text classi昀椀er to learn the lexical features of so-called “posts of memory” (POMs).
A昀琀er parameter tuning, the model extracted from the dataset a total of 1.083.471 POMs with a precision
score of 98.43%. It is argued that this technique provides a novel way to automate the identi昀椀cation of
historical entities within the far-right authored text, of bene昀椀t for the 昀椀elds of memory studies and far-
right studies, two 昀椀elds that have traditionally relied on more qualitative close-reading approaches. By
investigating the mnemonic features of the /pol/ posts during steps in the methodological pipeline, the
paper contributes important insights into the challenges of identifying and classifying lexical features
in hyper-vernacular digital spaces like 4chan, where communication is highly de昀椀ned by intertextuality,
semantic ambiguity, and cacography.
Keywords
4chan, Text classi昀椀cation, Far-right memory, Media and memory, Right-wing extremism
Introduction
Memory plays a pivotal role for far-right groups as it does in most processes of collective iden-
tity formation. Previous work on the memory practices of far-right groups and actors has
emphasized, how elements from the past are strategically energized within this ideological en-
vironment as symbolic and cultural building blocks for forming identity, strengthening group
a昀케liations, systematizing ideological strands, and directing contemporary political objectives
[2, 3, 5, 8, 10, 16, 22, 25, 32, 38, 40, 41]. Collectively imagined beliefs about social life and order,
borne out of today’s far-right mnemonic practices, can function as key analytical entryways
to understanding those concomitant ideals that structure their particular worldview.
CHR 2022: Computational Humanities Research Conference
£ psb@cas.au.dk (P. S. Baun); kln@cas.au.dk (K. Nielbo)
ç https://psbaun.github.io/ (P. S. Baun); https://knielbo.github.io/ (K. Nielbo)
ȉ 0000-0002-2433-8212 (P. S. Baun); 0000-0002-5116-5070 (K. Nielbo)
© 2020 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR
Workshop
Proceedings
http://ceur-ws.org
ISSN 1613-0073
CEUR Workshop Proceedings (CEUR-WS.org)
152
No study has to our knowledge mapped the far-right memory-scape in its totality. This is
due in part to the limited ability of qualitative methods to make generalizable claims about
the wider far-right milieu. Instead, such methodological approaches are more o昀琀en employed
for the purpose of ethnographic studies of select subsections or case studies of the far-right,
such as the memory practices of single actors or groups [25, 38]. As a result, there is limited
knowledge about the general historical elements that bind together this overarching far-right
collective memory.
Large-scale data-driven approaches o昀昀er a partial solution to this issue of generalizability.
More speci昀椀cally, statistical techniques for supervised and unsupervised learning make it pos-
sible to analyze the lexical qualities of memory on the far-right in ways normally unfeasible
by traditional close-observational studies.
To reconstruct the memory-scape of the far-right, this paper presents an observational study
that combines systematic annotation and text classi昀椀cation to the ‘politically incorrect’-board
(colloquially known as /pol/) of 4chan.org, a social media site and chat forum well known for
harboring far-right views and content [6, 9, 17, 26, 36, 37, 39].
Methods
Data
Posts from 4chan’s /pol/-board were randomly extracted from a dataset of threads from 2020
using 4chan’s API. In total 19, 821 random posts were manually annotated by one of the au-
thors in Doccano [27] for the presence of memory, sorting all posts into two non-overlapping
groups: posts of memory (POM) and posts not of memory (non-POM). Conceptually, memory
on /pol/ is de昀椀ned as reference to historical entities in posts targeting those narrative categories
in which a disorganized past becomes meaningful in discourse. Without assuming a priori the
precise contents of such entities, they are lexically and semantically speaking related to histor-
ical signi昀椀ers such as the names of past individuals, events, periods, etc. In order to delineate
between entities conceptually understood by users as belonging to either the past or present,
the concept only considers entities related to before the year 2000.
With this approach out of the 19, 821 posts, 1, 236 meet the criteria as being POMs, with the
other 18, 585 posts consequently being classi昀椀ed as non-POMs. In other words, 6.24% of posts
contained references to a historical entity before the year 2000.
Validation
To test annotator reliability, three independent raters were recruited and tasked with annotat-
ing a subset of the 19, 821 posts. A昀琀er being instructed in the coding procedure, they were each
provided with a dataset of 500 posts consisting of 50/50 randomly selected posts from each cat-
egory (POM/non-POM). Cohen’s �㔅 was calculated for the original annotations and each rater
independently: Rater 1 �㔅 = .94, rater 2 �㔅 = .92, and rater 3 �㔅 = 0.82. Landis and Koch’s
benchmark for interpreting kappa values suggests that a �㔅 of more than 0.81 is considered ‘al-
most perfect’ [24]. Fleiss’ �㔅 – recommended for multiple raters determining among nominal
153
categories [13, 15] – were also calculated for all raters �㔅 = 0.86. Statistics like Scott’s �㔋 and Krip-
pendor昀昀’s �㗼 coe昀케cient yielded similar results [23, 33]. We take the high level of annotation
agreement of POMs as strong evidence of the original annotations’ reliability.
Text Normalization and Representation
Before model training, a series of text normalization transformations were applied to the posts,
speci昀椀cally lemmatization, case-folding, and removal of punctuation. The posts were subse-
quently structured using a vector space model of lexical features, speci昀椀cally the empirical
probability of words and sequential combinations (i.e., n-grams). The n-gram range was treated
as a parameter value together with minimum and maximum document feature frequency.
Parameter Optimization & Model Training
For model training, data were balanced with under-sampling, that is, by randomly sampling
1, 236 non-POMs without replacement, corresponding to the total number of POMs, and remov-
ing the rest of the majority category, resulting in a balanced dataset of 50% posts from each
category. As opposed to over-sampling (i.e., multiplying data from the minority category to
match the level of the majority category), research suggests that under-sampling provides bet-
ter sensitivity at the cost of speci昀椀city [1]. Given that we are interested in accurately detecting
true positives of POMs, under-sampling was the more sensible approach. Following a similar
logic, precision was chosen as the performance measure, since precision is de昀椀ned as the num-
ber of true positives (ground truth POMs), divided by the number of predicted positives (model
classi昀椀ed POMs, either correctly or incorrectly), thereby giving a sense of the exactness of how
well the model is able to 昀椀nd actual POMs, whilst limiting the number of false positives.
The multinomial Naive Bayes algorithm was chosen as a classi昀椀cation technique due to its
computational e昀케ciency and level of explainability [35, 31]. The algorithm is based on Bayes’
Theorem and computes the posterior probability for the target and non-target class (POM vs.
non-POM) for each document using the Bayes rule and assigns the document to a class based
on the maximum posterior probability, or formally, the probability of a document �㕑 belonging
in class �㕐, �㕃(�㕐 ∣ �㕑) is:
�㕚
�㕃(�㕐 ∣ �㕑) ∝ �㕃(�㕐) ∏ �㕃(�㕡�㕖 ∣ �㕐) (1)
�㕖=1
and the class of a document �㕑 is then computed as:
�㕐�㕀�㔴�㕃 = �㕎�㕟�㕔 �㕚�㕎�㕥�㕐∈{�㕐1 ,�㕐2 } �㕃(�㕐 ∣ �㕑) (2)
Model parameters were optimized using a train/test split ratio of 75/25 with 149, 760 candi-
date parameter values for nine parameters using a 昀椀ve-fold cross-validation method, totaling
748, 800 昀椀ts. When comparing an unoptimized model without default parameters to the op-
timized model, precision increased from 63.49% to 76.19%. By shi昀琀ing the decision threshold
value, making the model more discriminant in only classifying posts with a predicted probabil-
ity over 0.9 as a POM, precision increased to 98.43%.
154
A note on the interpretation of parameter optimization, the optimal n-gram range was [1, 5],
suggesting that historical entities can be expressed through a multitude of complex word se-
quences. The minimum document frequency of a feature was determined as two documents
(i.e., the model 昀椀lters out words that appear in less than two posts). The fact that the optimal
minimum document frequency is not higher, suggests that even very rare words contribute to
model learning, seeing as they might represent obscure historical entities. Conversely, optimal
maximum document frequency was determined to be 30%, a somewhat extensive upper bound-
ary, suggesting that the corpus is riddled with many common words that have little signi昀椀cance
in terms of learning.
Results
To understand the classi昀椀er’s behavior, we looked at the speci昀椀c n-grams that were most pre-
dictive of the two categories. By counting the number of times each word occurs in all posts
divided by the number of posts in each category, we calculated the percentage of times a word
has appeared in one or the other category. In other words, by dividing the number of times
a word appears in the POM category by the number of times it appears in the non-POM cate-
gory, we are able to calculate a ‘memory ratio’ for each word and rank them according to this
scale. To avoid division by zero in the instances where a word only occurs in one category, a
pseudo-count of 1 was added to all word counts, similar to the Laplace smoothing technique
used to regularize the naive Bayes algorithm to avoid a probability estimate of zero when a
feature value does not occur in a given category. Looking at the top and bottom of this sorted
list of n-grams in table 1, we see what words most strongly correlate with each category i.e.,
what features are most and least ‘memory-like’.
The results indicate that the model correctly picks up on key historical entities that to a
human interpreter is understandable as having mnemonic signi昀椀cance, for example, distinct
historical persons, events, or places like ‘Hitler’, ‘Rome’, ‘Holocaust’, ‘Weimar’ and ‘WW2’, but
also more general historical or temporal terms such as ‘history’, ‘century’, ‘ancient’, ‘civiliza-
tion’, and ‘ancestor’. There are also less historically signifying terms like ‘catholic’, ‘Germany’,
or ‘Europe’, implying that these have been used in mnemonic context across multiple posts for
the model to start associating these terms with the POM category. Such re-contextualization of
generally nondescript terms is also interesting in the cases of ‘build’ and ‘steal’, which indicate
that /pol/ users ascribe speci昀椀c sentiments to their mnemonic discussions, expressed through
not only historical entities themselves, but also with descriptive modi昀椀ers such as verbs.
Conversely, the features in table 1 with the lowest memory ratio expectedly signify little
mnemonic substance or contain less contextual meaning (such as words like ‘she’, ‘her’, ‘any-
way’, ‘no no’, and ‘ah’). The low scores of words like ‘Biden’, ‘virus’, ‘test’, and ‘death’ also
indicate that discussions about memory do not overlap with discussions about contemporary
events in 2020, such as the election of Joe Biden and the COVID-19 pandemic. Terminology
speci昀椀c to 4chan like ‘tfw’ (that face when), ‘kek’ (synonymous to LOL; laughing out loud),
‘image’, ‘post’, as well as ‘rare’ (referring to the ‘rare 昀氀ag’ meme about users from small or un-
familiar countries), are also unrelated to the POM category, suggesting that discussions about
memory are a distinct subtheme on /pol/, branched o昀昀 from more general 4chan topics.
155
Table 1
Ratio for top and bottom 20 words most and least related to the POM category.
Memory words Memory ratio Non-memory words Memory ratio
hitler 13.75 biden 0.24
history 6.87 virus 0.31
rome 6.21 twf 0.32
century 5.61 kek 0.32
catholic 4.83 she 0.34
pyramid 4.63 her 0.36
holocaust 4.55 death 0.38
ancient 4.50 test 0.38
germany 4.35 make you 0.40
soviet 4.26 anyway 0.40
weimar 4.18 image 0.40
civilization 4.17 love 0.40
ancestor 4.09 no no 0.42
build 4.05 fit 0.43
ww2 3.88 post 0.43
the pyramid 3.80 lel 0.43
steal 3.76 ah 0.44
nazi 3.69 rare 0.45
europe 3.67 community 0.45
slave 3.63 google 0.46
Looking at the predicted probability of POMs provided by the model, many posts with high
predicted probably are also generally lengthier and mention several historical entities. For
example, one post with a word length of 133 (decidedly higher than the corpus average of 44
number of words per post), and with a 99.23% probability of being a POM, repeats the word
‘USSR’ four times as well as consisting of other signi昀椀cant historical entities, such as ‘Soviet
Union’ and the names of former Russian leaders. In contrast, scanning through posts with
low POM probability, these are much shorter (posts with only 10% POM probability contain on
average 23 words), and use many unspeci昀椀c words.
Discussion
Examples from the data set can provide further insights into the model’s learning behavior.
Consider a post from the dataset that reads: ‘enjoy your bat meat, you fucking barbarians.’
While the term ‘barbarian’ was originally used in ancient Greece to refer to non-Greek peoples
based on cultural-linguistic di昀昀erences, and would therefore adhere to the target class criteria,
the context within which it is used here – involving the zoonotic origin of the coronavirus
– is detracted from its original historical context. This is despite the remnant of its histori-
cally rhetorical function still undergirding its usage in the posts (i.e., to ‘uncivilize’ someone
by stereotyping them as barbarian). The semantic ambiguity – that there is not a direct correla-
tion between the use of a term and that term always represents a historical entity – necessarily
156
complicates the machine learning procedure, because the model relies on the assumption that
there exists an unambiguous lexical distinction between the POM and non-POM category’s tex-
tual content. Semantic ambiguity at a lexical level tends to be the rule rather than the exception
in natural language, and hence an inevitable condition of any machine learning process dealing
with unstructured text from a real-life environment. Some terms will inherently 昀氀uctuate in
their potential for expressing memory.
Depending on the circumstance, the composition of entities can be more or less metonymi-
cally representative of that particular conceptual category to which they are assumed to belong.
By exploring the lexical features of an abstract category such as memory, we are not only tak-
ing the necessary, precautionary steps of transparently revealing the data that goes into the
machine learning model but are also shedding light on the complex ways that memory is ex-
pressed through language in decidedly ambiguous ways. This can be demonstrated from an
example in the dataset: ‘we wuz vikangz.’ Brie昀氀y put, the post is a satirical rehashing of the
meme colloquially known as ‘we wuz kangz’, which was originally directed towards a type of
Afrocentric memory concerning the disputed and anachronistic claim that ancient Egypt was
a black civilization. Consequently, the rehashing is now being used to satirize the pretense of
people claiming to have Viking ancestry. While certainly interesting on its own as a case for
how memory can be embedded in multilayered intertextual contexts, the post also directs our
attention towards a speci昀椀c characteristic of the data. That is, historical entities can be rep-
resented in text by lexical symbols that may be synonymous in their meaning but which are
orthographically heterogeneous in their written format. The words ‘Vikings’, and ‘Vikangz’
both connote the same historical entity, but since they are spelled di昀昀erently it blocks the
machine learning model from disambiguating their semiotic synonymity.
The presence of homonyms also needs to be taken into consideration as these also distort
the informational quality of the data. As an example, the word ‘Rome’ was o昀琀en used to refer
to the historical Mediterranean civilization but is also used to denote the capital of present-day
Italy. Likewise, ‘Romans’ would generally refer to the people of ancient Rome, but would also
昀椀gure in discussions of biblical scripture, referring to the Letter to the Romans, part of the
Pauline epistles. While rare, these homonymic cases exemplify some of the linguistic features
of the data that need to be considered when constructing the machine learning model.
The subtle variations of meaning and semantics explored in the dataset present di昀케culty for
accurately extracting memory on /pol/. Nevertheless, by highlighting these cases, the article
has identi昀椀ed some of the preliminary linguistic characteristics that undergird discourses about
memory on /pol/, which ultimately become important to consider when building and evaluat-
ing machine learning models for 4chan-speci昀椀c and/or history-related classi昀椀cation problems.
Consequently, a major contribution of this study lies in its impelling of a still rather nascent
and sparse 昀椀eld of research that employs various natural language processing techniques in
the study of memory in the digital realm speci昀椀cally, e.g. [42, 34, 14, 19, 12, 18].
It also speaks directly to the di昀케culty of operationalizing a socially situated memory con-
cept in contrast to more individual-oriented studies attempted by neurobiologists and cognitive
psychologists, most recently pointed out by [29] and [28]. Similar critiques of the supposed
conceptual unclarity of “memory”, regarding the concept’s supposed over-extension and se-
mantic overloading (for example in the metaphorical misuse of psychological terms such as
“trauma” in supra-individual social contexts), resulting in redundant and unsophisticated uses
157
of the concept as a rhetorical signal, rather than as a clearly de昀椀ned analytical tool, have also
been pointed out by [20, 4, 11, 30, 21] and [7]. Recognizing this criticism within the 昀椀eld
of memory studies, this article has attempted to o昀昀er a more systematized and reproducible
way of identifying and understanding memory, speci昀椀cally in the context of memory’s lexical
manifestation.
However, we do not argue to have presented anything approaching a sui generis methodol-
ogy, completely separate and distinct from previous endeavors. If anything, the detailing of
many of the elements of note involved in applying machine learning, from evaluating statis-
tical biases to the exploration of linguistic variations, shows how a project like this requires
an inclusive, interdisciplinary outlook capable of combining and balancing the potential from
multiple perspectives: including both theory and method, the close reading spotlight and the
distant reading bird’s eye view, human awareness and computer industriousness.
There are also certain limitations involved with this method. As has been pointed out pre-
viously, there is a general disconnect on both a practical and theoretical level between the
lexical data that the models were trained on and those conceptual de昀椀nitions that structured
the qualitative interpretation of that data. While ‘historical entities” as a concept, allegorically
symbolizing objects of some abstract and intangible past, may be comprehensible on a theo-
retical level as the basic constituents of mnemonic communication, there is no guarantee that
such a concept, once operationalized, is bijectional translated into exactly replicable lexical
items making up a conversation on /pol/. In other words, the word tokens that are the fun-
damental components of the machine learning models are dimensionally, linguistically, and
conceptually speaking di昀昀erent from the theoretical de昀椀nition of memory, even when such
inferential parallelism is assumed in the study’s design. The polysemantic and cacography of
natural language thus impedes machine learning, for example when it has to determine the
mnemonic signi昀椀cance of ambivalent features, or when it needs to learn from quirky spelling
and terminology.
Moreover, given the multifarious representations that conceptually fall under the category of
historical entities, there is also a certain bias related to the initial, manually coded dataset. Even
though the dataset was constructed from a random sampling of /pol/ posts, there are quite
likely numerous historical entities that did not make it into this relatively small subset of the
general /pol/ conversation, even if they might have appeared relatively frequently and would
have been important to the board’s collective memory. Such biases would then be replicated
in the model’s algorithm, resulting in favoritism of historical entities that statistically might
describe very well the core memory on /pol/, represented by historical entities and other
“memory lingo” so common as to appear distinctly in the random sampling. However, it would
be blind to the “edges” of this collective memory, to the speckling of historical entities that the
model never got a chance to see, and which were subsequently not included in the extraction
process and therefore most likely wouldn’t feature as some of the top words in the 昀椀nal topic
modeling. This is, of course, but an unavoidable condition of necessarily decomplexifying the
exceedingly amorphous cultural concept of memory into quanti昀椀able entities suitable for study.
With any perspective striving for the macroscopic, there is bound to be a corresponding loss
in detail.
158
References
[1] S. Alsaif and A. Hidri. “Impact of Data Balancing During Training for Best Predictions”.
In: Informatica 45 (2021). doi: 10.31449/inf.v45i2.3479.
[2] A. J. Bauer. “The Alternate Historiography of the Alt-Right: Conservative Historical Sub-
jectivity from the Tea Party to Trump.” In: Far-Right Revisionism and the End of History:
Alt/Histories. Ed. by L. D. Valencia-Garcı́a. Routledge, 2020, pp. 120–137.
[3] P. S. Baun. “Memory and far-right historiography: The case of the Christchurch shooter”.
In: Memory Studies 15.4 (2022), pp. 650–665. doi: 10.1177/17506980211044701. url: http
s://journals.sagepub.com/doi/abs/10.1177/17506980211044701.
[4] D. Berliner. “The Abuses of Memory: Re昀氀ections on the Memory Boom in Anthropology”.
In: Anthropological Quarterly 78.1 (2005), pp. 197–211.
[5] H. Betz. “Revisiting Lepanto: the political mobilization against Islam in contemporary
Western Europe.” In: Patterns of Prejudice 43.3-4 (2009), pp. 313–334.
[6] K. Bezio. “Ctrl-Alt-Del: gamergate as a precurser to the rise of the alt-right.” In: Leader-
ship 14.5 (2018), pp. 556–566.
[7] A. Con昀椀no. “Collective Memory and Cultural History: Problems of Method”. In: The
American Historical Review 102.5 (1997), pp. 1386–1404.
[8] J. Dozier. “Hate Groups and Greco-Roman Antiquity Online: To Rehabilitate or Recon-
sider?” In: Far-Right Revisionism and the End of History: Alt/Histories. Ed. by L. D. Valencia-
Garcı́a. Routledge, 2020, pp. 251–296.
[9] B. Elley. ““The rebirth of the West begins with you!”–Self-improvement as radicalisation
on 4chan”. In: Humanities and Social Sciences Communications 8.1 (2021), p. 67. doi: 10.1
057/s41599-021-00732-x. url: https://doi.org/10.1057/s41599-021-00732-x.
[10] A. B. Elliott. “Internet medievalism and the White Middle Ages”. In: History Compass
16.3 (2018), e12441. doi: https://doi.org/10.1111/hic3.12441. url: https://compass.online
library.wiley.com/doi/abs/10.1111/hic3.12441.
[11] J. Fabian. “Remembering the Other: Knowledge and Recognition in the Exploration of
Central Africa”. In: Critical Inquiry 26 (1999), pp. 49–69.
[12] M. Ferron and P. Massa. “Beyond the encyclopedia: Collective memories in Wikipedia”.
In: Memory Studies 7 (2013), pp. 22–45. doi: 10.1177/1750698013490590.
[13] J. L. Fleiss. “Measuring nominal scale agreement among many raters”. In: Psychological
Bulletin 76.5 (1971), pp. 378–382. doi: 10.1037/h0031619.
[14] R. Garcı́a-Gavilanes, A. Mollgaard, M. Tsvetkova, and T. Yasseri. “The memory remains:
Understanding collective memory in the digital age”. In: Science Advances 3.4 (2017),
e1602368. doi: doi : 10 . 1126 / sciadv . 1602368. url: https : / / www . science . org / doi / abs
/10.1126/sciadv.1602368.
[15] N. Gisev, J. S. Bell, and T. F. Chen. “Interrater agreement and interrater reliability: key
concepts, approaches, and applications”. In: Res Social Adm Pharm 9.3 (2013), pp. 330–8.
doi: 10.1016/j.sapharm.2012.04.004.
159
[16] R. Gri昀케n. “Fixing Solutions: Fascist Temporalities as Remediesfor Liquid Modernity”. In:
Journal of Modern European History 13.1 (2015), pp. 5–23. doi: 10.17104/1611-8944\_201
5\_1\_5. url: https://journals.sagepub.com/doi/abs/10.17104/1611-8944%5C%5F2015%5
C%5F1%5C%5F5.
[17] G. Hawley. The Alt-Right: What Everyone Needs to Know. New York: Oxford University
Press, 2019.
[18] A. Jatowt, D. Kawai, and K. Tanaka. Digital History Meets Wikipedia: Analyzing Historical
Persons in Wikipedia. Conference Paper. 2016. doi: 10.1145/2910896.2910911. url: https:
//doi.org/10.1145/2910896.2910911.
[19] N. Kanhabua, T. N. Nguyen, and C. Niederée. “What triggers human remembering of
events? A large-scale analysis of catalysts for collective memory in Wikipedia”. In: IEEE/ACM
Joint Conference on Digital Libraries (2014), pp. 341–350.
[20] W. Kansteiner. “Finding Meaning in Memory: A Methodological Critique of Collective
Memory Studies”. In: History and Theory 41.2 (2002), pp. 179–197.
[21] K. L. Klein. “On the Emergence of Memory in Historical Discourse”. In: Representations
69 (2000), pp. 127–150.
[22] C. Kølvraa. “Embodying ‘the Nordic race’: imaginaries of Viking heritage in the online
communications of the Nordic Resistance Movement”. In: Patterns of Prejudice 53.3 (2019),
pp. 270–284. doi: 10.1080/0031322x.2019.1592304. url: https://doi.org/10.1080/0031322
X.2019.1592304.
[23] K. Krippendor昀昀. Content analysis: An introduction to its methodology. 3rd edition. Thou-
sand Oaks, CA: Sage, 2013.
[24] J. R. Landis and G. G. Koch. “The Measurement of Observer Agreement for Categorical
Data”. In: Biometrics 33.1 (1977), pp. 159–174. doi: 10.2307/2529310. url: http://www.jst
or.org/stable/2529310.
[25] C. Miller-Idris. The extreme gone mainstream: commercialization and far right youth cul-
ture in Germany. Princeton: Princeton University Press, 2017.
[26] A. Nagle. Kill All Normies: Online Culture Wars from 4chan and Tumblr to Trump and the
Alt-Right. Zero Books, 2017.
[27] H. Nakayama, T. Kubo, J. Kamura, Y. Taniguchi, and X. Liang. doccano: Text Annotation
Tool for Human. 2018. url: https://github.com/doccano/doccano.
[28] J. Olick. The Politics of Regret: On Collective Memory and Historical Responsibility. Rout-
ledge, 2007.
[29] J. K. Olick, Vinitzsky-Seroussi, V., and D. Levy. “Introduction”. In: The Collective Memory
Reader. Ed. by D. L. Je昀昀rey K. Olick Vered Vinitzky-Seroussi. Oxford University Press,
2011, pp. 3–62.
[30] G. J. R. “Memory and Identity: The History of a Relationship.” In: Commemorations. Ed.
by J. R. Gillis. Princeton: Princeton University Press, 1994, pp. 3–24.
160
[31] S. Raschka. “Naive Bayes and Text Classi昀椀cation I - Introduction and Theory”. In: arXiv:1410.5329
[cs] (2017). url: http://arxiv.org/abs/1410.5329.
[32] I. Richards. “A Philosophical and Historical Analysis of “Generation Identity”: Fascism,
Online Media, and the European New Right”. In: Terrorism and Political Violence 34.1
(2022), pp. 28–47. doi: 10.1080/09546553.2019.1662403. url: https://doi.org/10.1080/0954
6553.2019.1662403.
[33] W. A. Scott. “Reliability of Content Analysis: The Case of Nominal Scale Coding”. In: The
Public Opinion Quarterly 19.3 (1955), pp. 321–325. url: http://www.jstor.org/stable/274
6450.
[34] Y. Sumikawa, A. Jatowt, and M. Düring. “Digital History meets Microblogging: Ana-
lyzing Collective Memories in Twitter”. In: Proceedings of the 18th ACM/IEEE on Joint
Conference on Digital Libraries (2018).
[35] S. L. Ting, W. H. Ip, and A. H. C. Tsang. “Is Naı̈ve Bayes a Good Classi昀椀er for Document
Classi昀椀cation?” In: International Journal of So昀琀ware Engineering and Its Applications 5.3
(2011), p. 11.
[36] M. Tuters. “Esoteric Fascism Online: 4chan and the Kali Yuga.Far-Right Revisionism and
the End of History.” In: Far-Right Revisionism and the End of History: Alt/Histories. Ed. by
L. D. Valencia-Garcı́a. Routledge, 2020, pp. 287–303.
[37] M. Tuters and S. Hagen. “(((They))) rule: Memetic antagonism and nebulous othering on
4chan”. In: New Media & Society 22.12 (2019), pp. 2218–2237. doi: 10.1177/146144481988
8746. url: https://doi.org/10.1177/1461444819888746.
[38] L. D. Valencia-Garcı́a. Far-Right Revisionism and the End of History: Alt/Histories. Rout-
ledge., 2020.
[39] M. Wendling. Alt-Right: From 4chan to the White House. London: Pluto Press, 2018.
[40] R. Wodak and B. Forchtner. “Embattled Vienna 1683/2010: right-wing populism, collec-
tive memory and the 昀椀ctionalisation of politics”. In: Visual Communication 13.2 (2014),
pp. 231–255. doi: 10.1177/1470357213516720. url: https://doi.org/10.1177/147035721351
6720.
[41] D. Wollenberg. “The new knighthood: Terrorism and the medieval.” In: Postmedieval 5.1
(2014), pp. 21–33.
[42] C.-m. A. Yeung and A. Jatowt. Studying how the past is remembered: towards computa-
tional history through large scale text mining. Conference Paper. 2011. doi: 10.1145/2063
576.2063755. url: https://doi.org/10.1145/2063576.2063755.
161