MPSA Annual Conference · April 2026
Constructing the Nation
Identity and historical narratives in South Korean history textbooks, 1895–2016.
Steven Denney  ·  Aron van de Pol
Leiden University
§ i  ·  The Question
Denney & van de Pol · 2026
Research question
How do the language and ideas of national identity in state-sanctioned educational material vary across regime types?
Constructing the Nation
Denney & van de Pol
§ i  ·  Motivation
Denney & van de Pol · 2026
Why national history textbooks
Institutions and Narrative.
The curricular core.
National history is the discourse least likely to change — where the state tells its most formal story of the nation.
Compressed history.
1974–2002: the state was not a certifier but the author. One textbook, across Park, Chun, and the democratic transition.
Constructing the Nation
Denney & van de Pol
§ i  ·  Expectations
Denney & van de Pol · 2026
Competing hypotheses
Regime type shapes
textbook language.
Expectation A
Authoritarian states compress the national vocabulary to a single official story.
Expectation B
Democratic institutions dissolve that story into plural, contested meanings.
Literature
Snyder (2000): democratization can intensify nationalism.
Jo (2025): elites may entrench rather than replace the inherited narrative.
Constructing the Nation
Denney & van de Pol
§ ii
Corpus
& periods.
§ ii  ·  The Corpus
Denney & van de Pol · 2026
Merged corpus · NIKH + Georg Eckert Institute
67 textbooks, 1895–2016.
67
textbooks
51 from NIKH digital archive · 16 OCR'd from existing library scans.
1895–
2016
coverage
Kabo reforms through the 7th national curriculum.
Part of a cross-national corpus digitizing 1,000s of history textbooks.
Constructing the Nation
Denney & van de Pol
§ ii  ·  Periodization
Denney & van de Pol · 2026
Our periodization · based on curriculum changes
Five periods of textbook production.
1895—1910
Late Chosŏn
N = 14
1910—1945
Colonial
N = 4
1945—1972
Postwar
N = 8
1973—1993
Authoritarian
N = 19
1993—2016
Democratic
N = 22
Pre-curr.
Colonial
1st
2nd
3rd
4th
5th
6th
7th +
Constructing the Nation
Denney & van de Pol
§ iii
How we
measure meaning.
§ iii  ·  Methods
Denney & van de Pol · 2026
Term discovery · LLM consensus pipeline
LLM-assisted term discovery.
01 · Extract
Three EXAONE models (2.4B / 7.8B / 32B) scan 4,000-char windows per textbook, returning identity-relevant nouns.
02 · Score
Candidates weighted 60 / 40 on semantic similarity and frequency; must appear ≥ 3 times.
03 · Consensus
A term enters only if at least 2 of 3 models select it. Twelve terms survive.
Validated against
BERTopic (recovers 12 / 12)  ·  Latent Dirichlet Allocation (recovers 7 / 12)
Constructing the Nation
Denney & van de Pol
§ iii  ·  Methods
Denney & van de Pol · 2026
Sentence embeddings · BM-K/KoSimCSE-roberta-multitask
Three axes of semantic change.
Every sentence containing an identity term is embedded. For each term and each transition, we measure change along three complementary axes.
Logistic Regression (LR)
Discriminability
How easily a classifier can tell which period a sentence belongs to. Higher = more distinctly period-specific usage.
Contrastive PCA (cPCA)
Contextual spread
Whether a term's usage expands into new contexts or contracts into fewer. Positive = wider range of meanings.
Procrustes Analysis (PA)
Geometric relocation
How far a term moves in the embedding space after optimal alignment. Large = repositioned relative to other terms.
Applied to the two post-liberation transitions: Postwar → Authoritarian and Authoritarian → Democratic.
Constructing the Nation
Denney & van de Pol
§ iii  ·  Vocabulary
Denney & van de Pol · 2026
The twelve identity terms
Twelve terms that define the nation.
nation / ethnos
minjok
민족
state
kukka
국가
citizens
kungmin
국민
independence
tongnip
독립
unification
t'ongil
통일
liberation
haebang
해방
history
yŏksa
역사
democracy
minjujuŭi
민주주의
politics
chŏngch'i
정치
society
sahoe
사회
territory
yŏngt'o
영토
community
kongdongch'e
공동체
Constructing the Nation
Denney & van de Pol
§ iii
What each word
is allowed to mean.
§ iv  ·  Findings
Denney & van de Pol · 2026
Frequency · post-liberation vocabulary
Nation-building vocabulary
appears after 1945.
Before 1945, most of the twelve terms are absent or marginal in school history.
After 1945, the full set arrives together — a coordinated language of nation, state, politics, and citizenship.
National history installed a shared vocabulary for nation, state, and political community across all schools.
Constructing the Nation
Denney & van de Pol
§ iv  ·  Frequency
Denney & van de Pol · 2026
Fig. 1 · Four representative terms · per 1,000 tokens
Minjok peaks under authoritarianism.
Mean frequency per 1,000 tokens across five periods.
Constructing the Nation
Denney & van de Pol
§ iv  ·  Aggregate
Denney & van de Pol · 2026
Mean absolute shift magnitudes · 12 terms × 3 axes
Variance dominates both transitions.
Axis
Postwar → Auth.
Auth. → Dem.
Discriminative (LR)
0.46
0.41
Variance (cPCA)
1.37
1.81
Alignment (PA)
0.31
0.40
The dominant mode of change is not wholesale replacement of one vocabulary by another. It is expansion or contraction of the range of contexts in which terms appear. cPCA shifts are 3–5× larger than discriminative or alignment shifts.
Constructing the Nation
Denney & van de Pol
§ iv  ·  Authoritarian
Denney & van de Pol · 2026
Postwar → Authoritarian · cPCA (contextual spread)
Selective appropriation, not
uniform closure.
Expanding — usable for legitimation
Independence 독립
+2.83
Nation 민족
+1.74
Citizens 국민
+1.59
History 역사
+1.49
Democracy 민주주의
+0.52
Unification 통일
+0.32
Compressing — harder to control
Territory 영토
−0.18
State 국가
−0.85
Politics 정치
−0.89
Society 사회
−1.18
Liberation 해방
−1.98
Community 공동체
−2.89
LR is uniformly negative across all terms (all become more period-specific). The split is on cPCA: widen what serves legitimation, compress what opens onto political or social analysis.
Constructing the Nation
Denney & van de Pol
§ iv  ·  Authoritarian
Denney & van de Pol · 2026
What the split reveals
The regime widened the meanings of concepts useful for legitimation and narrowed the rest to a few repeated phrases.
Selective appropriation · not uniform closure
Expansion
Ethnic belonging, colonial grievance, disciplined citizenship.
Compression
Society, politics, community -- pulled toward formulaic use.
Freezing
Unification and democracy -- present but held in safe historical registers.
Constructing the Nation
Denney & van de Pol
§ iv  ·  Case: Minjok 민족
Denney & van de Pol · 2026
Minjok 민족 · Postwar → Authoritarian
Same word, wider range,
new ideological load.
Postwar
국사 교육은 우리 민족의 과거 생활을 정확하게 인식시켜 자아를 반성하고…
"Recognizing our nation's past life, fostering self-reflection and patriotism…"
Postwar textbook
Authoritarian
일제의 식민지 지배는 우리 민족의 민족 운동을 탄압하고, 경제적 착취를 강행하며…
"Colonial rule suppressed our nation's movement, enforced exploitation…"
1981 textbook
+1.74 cPCA — usage contexts widen
0.63 PA — changing neighbouring words
Constructing the Nation
Denney & van de Pol
§ iv  ·  Case: Minjok 민족
Denney & van de Pol · 2026
Minjok 민족 · Authoritarian → Democratic
Keeps expanding — adds reconciliation
to the inherited frame.
Authoritarian · colonial grievance
일제의 식민지 지배는 우리 민족의 민족 운동을 탄압하고, 경제적 착취를 강행하며…
"Colonial rule suppressed our nation's movement, enforced exploitation…"
1981 textbook
Democratic · inter-Korean reconciliation
7·4 남북 공동 성명에서 남북한은 자주 통일, 평화 통일, 민족 대단결의 3대 원칙을 내세웠다.
"North and South Korea put forward three principles: self-reliant reunification, peaceful reunification, and great national unity."
2006 textbook
+1.22 cPCA — still expanding, less dramatically
Constructing the Nation
Denney & van de Pol
§ iv  ·  Case: T'ongil 통일
Denney & van de Pol · 2026
T'ongil 통일 · under authoritarianism
Frequent but frozen — confined to
anti-communist ritual.
Historical referent
신라는 삼국을 통일하여 민족 문화 발전의 기틀을 마련하였다.
"Silla unified the Three Kingdoms and laid the foundation for national culture."
1979 textbook
Anti-communist formula
민족적 숙원인 국토 통일을 위하여 공산주의와 대결할 수 있는 실력 배양에 전력을 집중한다.
"Concentrate all efforts on cultivating strength to confront communism for territorial unification."
1973 · Yushin era
Variance shift of only +0.32 cPCA. The term was retained but constrained -- confined to Unified Silla and ritualized slogans, never substantive policy.
Constructing the Nation
Denney & van de Pol
§ iv  ·  Case: T'ongil 통일
Denney & van de Pol · 2026
T'ongil 통일 · Authoritarian → Democratic
Under democracy, unification enters
contemporary policy.
Authoritarian · frozen
민족적 숙원인 국토 통일을 위하여 공산주의와 대결할 수 있는…
"…cultivating strength to confront communism for unification…"
1973 · Yushin era
Democratic · policy discourse
정상 회담 결과 통일 문제와 남북 관계의 기본 방향을 담은 6·15 남북 공동 선언이 발표되었다.
"The June 15 Joint Declaration… the basic direction for unification and inter-Korean relations."
2016 textbook
The +2.86 cPCA shift is the second-largest in the analysis. Unification had not been absent from authoritarian textbooks -- it had been constrained. Institutional change, not new referents, unlocked the range.
Constructing the Nation
Denney & van de Pol
§ iv  ·  Democratic
Denney & van de Pol · 2026
Authoritarian → Democratic · cPCA
All terms expand
under democracy.
Constructing the Nation
Denney & van de Pol
§ iv  ·  Case: Haebang 해방
Denney & van de Pol · 2026
Haebang 해방 · largest PA (alignment) shift
Haebang reactivates its
liberation sense.
Authoritarian · secondary sense
양민으로서 억울하게 노비가 된 자들을 해방시켜…
"emancipating those unjustly enslaved" — premodern royal policy.
1981 textbook
Democratic · secondary sense
독립 운동계는 민족 해방을 부르짖는 민족주의 계열과 계급 해방을 목표로 하는 사회주의 세력으로 양분되었다.
"…national liberation vs. class liberation" — political-movement framing.
2007 textbook
+1.91 PA — largest shift in the analysis. Geometric relocation: dormant senses activate.
Constructing the Nation
Denney & van de Pol
§ iv  ·  Durability
Denney & van de Pol · 2026
Durability · two terms never fully recover
Two terms never fully recover.
Sahoe 사회 · society
−1.18
cPCA · Post. → Auth.
→
+0.76
cPCA · Auth. → Dem.
Only ⅔ recovered. The evaluative edge -- society as critical category -- does not fully return.
Haebang 해방 · liberation
−1.98
cPCA · Post. → Auth.
→
+1.08
cPCA · Auth. → Dem.
Only ½ recovered. Three decades of democracy did not erase the narrowing.
A form of banal constraint: a narrowing of conceptual tools that persists precisely because it is unremarkable.
Constructing the Nation
Denney & van de Pol
§ v  ·  Synthesis
Denney & van de Pol · 2026
Summary · three phases
Three phases of state-driven
semantic change.
Phase I
Establish
Liberation installs a stable vocabulary. Every successor regime inherits it.
Phase II
Modulate
Selective appropriation: stretch what serves legitimation, flatten what doesn't.
Phase III
Pluralize
Democracy does not de-nationalize the textbook. It de-solidifies it: every term widens; the frame loosens.
The core persists. What varies is how tightly the surrounding language is gathered into one frame.
Constructing the Nation
Denney & van de Pol
§ v  ·  Contributions
Denney & van de Pol · 2026
What this paper offers
Contributions.
Method
Replicable pipeline for corpus-driven concept discovery using LLMs, validated against BERTopic and LDA.
Multi-axis semantic design (LR, cPCA, PA) that separates different mechanisms of meaning change.
Substantive & theoretical
Regime type shapes the grand narrative states transmit through education.
Authoritarian schooling selectively appropriates — expanding some terms while compressing others.
Democratization does not per se replace the inherited narrative — it widens inside the frame authoritarianism built.
Constructing the Nation
Denney & van de Pol
The claim, in one sentence
A national vocabulary can be shared across regimes.
The meanings those words carry can differ substantially with the institutions that produce them.
s.c.denney@hum.leidenuniv.nl  ·  a.m.van.de.pol@hum.leidenuniv.nl
Constructing the Nation
Denney & van de Pol