Package {leadeR}

Title:	Profiling Leaders at a Distance
Version:	0.1.0
Description:	Profiles political leaders at a distance from text data such as speeches, interviews, press conferences, and other public statements. Computes Leadership Trait Analysis scores for seven personality traits – including need for power, conceptual complexity, and self-confidence – and classifies leaders into one of eight leadership styles. Also computes Operational Code Analysis scores summarising a leader's beliefs about politics and the use of power.
URL:	https://github.com/mmukaigawara/leadeR
License:	MIT + file LICENSE
LazyData:	true
Encoding:	UTF-8
Depends:	R (≥ 4.1.0)
Imports:	spacyr, dplyr, stringr, tidyr, stringi, data.table, tibble, countries, countrycode, purrr
Suggests:	knitr, metafor, rmarkdown, testthat (≥ 3.0.0)
VignetteBuilder:	knitr
Config/testthat/edition:	3
Config/roxygen2/version:	8.0.0
NeedsCompilation:	no
Packaged:	2026-05-13 00:09:01 UTC; mitsurumukaigawara
Author:	Mitsuru Mukaigawara [cre, aut], Joshua Kertzer [aut]
Maintainer:	Mitsuru Mukaigawara <mitsuru_mukaigawara@g.harvard.edu>
Repository:	CRAN
Date/Publication:	2026-05-18 18:20:02 UTC

leadeR: Profiling Leaders at a Distance

Description

The leadeR package profiles political leaders using text analysis, implementing Leadership Trait Analysis (LTA) and Operational Code Analysis (OCA). You provide text data and the package performs the analyses.

Author(s)

Maintainer: Mitsuru Mukaigawara mitsuru_mukaigawara@g.harvard.edu (ORCID)

Authors:

• Mitsuru Mukaigawara mitsuru_mukaigawara@g.harvard.edu (ORCID)

• Joshua Kertzer jkertzer@gov.harvard.edu (ORCID)

See Also

Useful links:

• https://github.com/mmukaigawara/leadeR

Add dots to acronyms

Description

Converts a short uppercase acronym into dotted forms (e.g., "US" becomes "U.S" and "U.S.").

Usage

acronym_dots(x)

Arguments

Value

A character vector of dotted variants, or character(0) if the input is not a 2-4 letter acronym.

Build country list for entity matching

Description

Constructs a vector of lowercased country names including aliases, without international organizations.

Usage

build_country_list()

Value

A character vector of country names.

Build entities corpus for entity matching

Description

Constructs a vector of lowercased international organization names and country names (including aliases) for use in entity detection.

Usage

build_entities_corpus()

Value

A character vector of entity names.

Clean text for analysis

Description

Removes editorial annotations in square brackets, parentheses, and curly braces from political speech transcripts. Also normalizes whitespace and collapses double em-dashes left behind by removals.

Usage

clean_text(x)

Arguments

Value

A character vector with annotations removed and whitespace normalized.

Examples

clean_text("We must act now [applause] for the future.")
clean_text("The president (speaking loudly) left the room.")

Count valid (non-negated) pattern matches

Description

Count valid (non-negated) pattern matches

Usage

count_valid(txt, pattern, window_chars = 40)

Arguments

Value

Integer count of valid matches.

Expand country name aliases

Description

Given a country name or code, returns all known aliases including official names, ISO codes, CLDR names, and dotted acronym variants.

Usage

expand_aliases_country(term)

Arguments

Value

A character vector of lowercased country name aliases.

Compute affiliation scores

Description

Measures the affiliation orientation of a leader's speech by classifying verb-level actions as affiliative (A) or other-affiliative (OA) based on the speaker's own-entity references and sentence-level context.

Usage

get_aff(own_entity, text, bootstrap = FALSE, B = 1000)

Arguments

Value

A one-row tibble. When bootstrap = FALSE, columns are A and OA. When bootstrap = TRUE, columns are meanA, meanOA, varA, varOA.

Compute conceptual complexity scores

Description

Counts high-complexity (HC) and low-complexity (LC) markers in the text, using word stems, phrases, and exact words with negation-aware counting.

Usage

get_complex(
  text,
  bootstrap = FALSE,
  B = 1000,
  quote_strip = TRUE,
  window_chars = 40
)

Arguments

Value

A one-row tibble. When bootstrap = FALSE, columns are HC and LC. When bootstrap = TRUE, columns are meanHC, varHC, meanLC, varLC.

Compute self-confidence scores

Description

Classifies first-person pronoun occurrences in the text as self-confident (SC) or other-self-confident (OSC) based on sentence-level context patterns covering three conditions: self as instigator, self as authority, and self as recipient of positive recognition.

Usage

get_conf(text, bootstrap = FALSE, B = 1000)

Arguments

Value

A one-row tibble. When bootstrap = FALSE, columns are SC and OSC. When bootstrap = TRUE, columns are meanSC, meanOSC, varSC, varOSC.

Compute control (need for power) scores

Description

Classifies verbs associated with first-person subjects as instrumental-control (IC) or other-control (OC) using dependency parsing to link subjects to their governing verbs.

Usage

get_ctrl(own_entity, text, bootstrap = FALSE, B = 1000)

Arguments

Value

A one-row tibble. When bootstrap = FALSE, columns are IC and OC. When bootstrap = TRUE, columns are meanIC, meanOC, varIC, varOC.

Compute distrust scores

Description

Performs per-entity classification of references in the text as suspicious (S) or other-suspicious (OS) based on distrust verbs, harmful nominals, and contextual modifiers.

Usage

get_dist(own_entity, text, bootstrap = FALSE, B = 1000)

Arguments

Value

A one-row tibble. When bootstrap = FALSE, columns are S and OS. When bootstrap = TRUE, columns are meanS, varS, meanOS, varOS.

Leadership Trait Analysis (LTA)

Description

Runs all eight LTA trait functions at once and returns a single combined tibble. The eight traits are: power, affiliation, distrust, conceptual complexity, task orientation, self-confidence, nationalism, and control.

Usage

get_lta(own_entity, text, bootstrap = FALSE, B = 1000)

Arguments

Value

A one-row tibble::tibble.

When bootstrap = FALSE, columns include raw counts (P, OP, A, OA, S, OS, HC, LC, TI, IP, SC, OSC, N, ON, IC, OC) plus trait proportions (Pp, D, C, Ta, Ss, Na, B).

When bootstrap = TRUE, columns include bootstrap means and variances for the raw counts (meanP, varP, meanOP, varOP, ...) plus trait proportions and their delta-method variances (Pp, varPp, D, varD, C, varC, Ta, varTa, Ss, varSs, Na, varNa, B, varB).

Examples

# Requires spaCy to be installed; see spacyr::spacy_install().
spacyr::spacy_initialize()
get_lta(own_entity = "United States", text = "We will defend our nation.")
get_lta(own_entity = "United States", text = "We will defend our nation.",
        bootstrap = TRUE, B = 500)

Compute nationalism scores

Description

Classifies entity references in the text as nationalistic (N) or other-nationalistic (ON) based on own/other entity detection and contextual modifiers at the referent level.

Usage

get_nat(own_entity, text, bootstrap = FALSE, B = 1000)

Arguments

Value

A one-row tibble. When bootstrap = FALSE, columns are N and ON. When bootstrap = TRUE, columns are meanN, meanON, varN, varON.

Compute operational code (VICS) scores

Description

Performs full VICS (Verbs in Context System) operational code analysis on a speech text, classifying verb-entity pairs as Self/Other and scoring them on a -3 to +3 scale (Punish to Reward). Returns P1–P5 and I1–I5 indices.

Usage

get_oca(own_entity, text, bootstrap = FALSE, B = 1000)

Arguments

Value

A one-row tibble with P1–P5, I1–I5, and raw Self/Other counts by score category.

Compute Power Score (LTA)

Description

Classifies verbs in text as reflecting Power (P) or Other Power (OP) based on the Leadership Trait Analysis codebook.

Usage

get_power(own_entity, text, bootstrap = FALSE, B = 1000)

Arguments

Value

A tibble with columns P and OP (or meanP, meanOP, varP, varOP if bootstrap = TRUE).

Compute VICS operational code scores from verb-entity coding

Description

Internal function that takes a data frame of coded utterances (with opc_score and entity_type columns) and computes the full set of P and I indices.

Usage

get_scores(dat)

Arguments

Value

A one-row tibble with P1–P5, I1–I5, and raw counts.

Compute task orientation scores

Description

Classifies lemmas in the parsed text as task-instrumental (TI) or interpersonal (IP) based on codebook word lists, with token-level quote and negation handling.

Usage

get_task(text, bootstrap = FALSE, B = 1000)

Arguments

Value

A one-row tibble. When bootstrap = FALSE, columns are TI and IP. When bootstrap = TRUE, columns are meanTI, meanIP, varTI, varIP.

International organizations list

Description

A character vector of international organization names and acronyms used for entity detection across the package.

Usage

io

JFK Speech: Inaugural Address (January 20, 1961)

Description

Full text of JFK's Inaugural Address on January 20, 1961.

Usage

jfk19610120

Format

A character string containing the full text of the speech.

Source

https://www.govinfo.gov/app/details/PPP-1961-book1. These original source texts are U.S. government works and are not subject to copyright protection in the United States under 17 U.S.C. § 105. The package authors do not claim copyright in the original presidential speech texts.

JFK Speech: Address Before the UN General Assembly (September 25, 1961)

Description

Full text of JFK's address before the UN General Assembly on September 25, 1961.

Usage

jfk19610925

Format

A character string containing the full text of the speech.

Source

https://www.govinfo.gov/app/details/PPP-1961-book1. These original source texts are U.S. government works and are not subject to copyright protection in the United States under 17 U.S.C. § 105. The package authors do not claim copyright in the original presidential speech texts.

JFK Speech: Commencement Address at American University (June 10, 1963)

Description

Full text of Senator John F. Kennedy's commencement address at American University on June 10, 1963.

Usage

jfk19630610

Format

A character string containing the full text of the speech.

Source

https://www.govinfo.gov/app/details/PPP-1963-book1. These original source texts are U.S. government works and are not subject to copyright protection in the United States under 17 U.S.C. § 105. The package authors do not claim copyright in the original presidential speech texts.

Prepare exact-word regex pattern

Description

Prepare exact-word regex pattern

Usage

prep_exact_regex(words)

Arguments

Value

A single regex string.

Prepare low-complexity regex pattern

Description

Prepare low-complexity regex pattern

Usage

prep_lc_regex(items)

Arguments

Value

A single regex string.

Prepare phrase regex pattern

Description

Prepare phrase regex pattern

Usage

prep_phrase_regex(phrases)

Arguments

Value

A single regex string.

Prepare word-stem regex pattern

Description

Prepare word-stem regex pattern

Usage

prep_wordstem_regex(stems)

Arguments

Value

A single regex string.

Strip quoted text from a string

Description

Strip quoted text from a string

Usage

strip_quoted(x)

Arguments

Value

The string with quoted passages removed.

Classify LTA Traits into Hermann's Leadership Typology

Description

Takes per-speech LTA output (from get_lta()) and aggregates trait scores across speeches, then classifies the leader along three dimensions (constraint, openness, motivation toward world) and maps the first two plus task orientation to one of eight leadership styles.

Usage

type_lta(
  lta,
  precision_weighted = FALSE,
  need_for_power = 0.5,
  control = 0.44,
  complex_high = 0.56,
  confidence_high = 0.81,
  task = 0.59,
  distrust = 0.41,
  ingroup = 0.42
)

Arguments

Value

A one-row tibble::tibble with aggregated trait values, standard errors (when precision_weighted = TRUE), and classification columns: constraint, openness, motivation_toward_world, task_orientation, typology, and method.

Examples

# Simple-mean aggregation works on any data frame with the seven trait
# columns; here we use a small illustrative input.
example_lta <- data.frame(
  Pp = c(0.45, 0.52, 0.48),
  B  = c(0.40, 0.48, 0.43),
  C  = c(0.60, 0.55, 0.58),
  Ss = c(0.75, 0.82, 0.79),
  Ta = c(0.62, 0.58, 0.65),
  D  = c(0.35, 0.42, 0.38),
  Na = c(0.38, 0.45, 0.41)
)
type_lta(example_lta)


# Precision-weighted aggregation needs per-trait variances, which come
# from get_lta(..., bootstrap = TRUE). Requires spaCy to be installed.
spacyr::spacy_initialize()
res <- data.table::rbindlist(
  lapply(c(jfk19610120, jfk19610925, jfk19630610), function(x)
    get_lta(own_entity = "United States", text = clean_text(x),
            bootstrap = TRUE, B = 1000))
)
type_lta(res, precision_weighted = TRUE)