# README – Dataset for “Project 2025 as a Technocratic Blueprint: A Corpus-Based Linguistic Analysis of Conservative Governance Discourse”

This file was generated on 2025-11-08 by Julia Schilling

---

## A. GENERAL INFORMATION

### 1. Title of the dataset
**Project 2025 Corpus Analysis Dataset (Text, Collocations, Keyness, and LIWC Outputs)**

### 2. Brief description of the research project and its aims
This dataset accompanies the manuscript *“Project 2025 as a technocratic blueprint: A corpus-based linguistic analysis of conservative governance discourse”* (Schilling & Fuchs, submitted to *PLOS ONE*).  
It contains all data and code used to analyze the conservative policy document *Mandate for Leadership: The Conservative Promise 2025* (Heritage Foundation, 2023) in comparison with Democratic and Republican party platforms.  
The dataset includes:
- preprocessed text corpora (Project 2025, Democratic, and Republican platforms),  
- results from keyness and collocation analyses,  
- LIWC-22 psycholinguistic scores, and  
- reproducible analysis scripts in R and Python.

The project investigates how Project 2025 differs linguistically from party platforms in its lexical, collocational, and psycholinguistic patterns, highlighting its technocratic orientation within conservative discourse.

### 3. Author Information

**A. Investigator Contact Information**  
Name: Julia Schilling  
Institution: University of Bonn, Bonn Applied English Linguistics (BAEL)  
Address: Rabinstraße 8, 53113 Bonn, Germany  
Email: julia.schilling@uni-bonn.de  

Name: Prof. Dr. Robert Fuchs  
Institution: University of Bonn, Bonn Applied English Linguistics (BAEL)  
Address: Rabinstraße 8, 53113 Bonn, Germany  
Email: rfuchs@uni-bonn.de  

**C. Contact for dataset inquiries**  
Name: Julia Schilling  
Institution: University of Bonn  
Email: julia.schilling@uni-bonn.de  

### 4. Date of data collection
2025-09  

### 5. Funding sources
No external funding. Work conducted within the Bonn Applied English Linguistics (BAEL) group.

### 6. Language of the dataset
English  

### 7. Geographic location of data collection
Bonn, Germany  

---

## B. DATA & FILE OVERVIEW

### 1. File List

**Code folder**
- `analysis_project2025.Rmd` – R Markdown for corpus comparison, statistical analysis, and visualization  
- `ollocation_analysis.py` – Python script for collocation extraction (spaCy-based)  
- `keyness_analysis.py` – Python script for keyness calculation (log-likelihood, log-ratio, BIC)

**Raw data folder**
- `Project2025.csv` – Full Project 2025 text corpus  
- `Project2025_lemmaPOS.csv` – Full Project 2025 text corpus lemmatized and tagged for parts of speech 
- `Platforms_Republicans.csv`, `Platforms_Democrats.csv` – U.S. Republican and Democratic party platforms  

**LIWC folder**
- `Project2025_LIWC_subset.csv` – LIWC-22 scores for Project 2025  
- `Platforms_Republicans_LIWC_subset.csv`, `Platforms_Democrats_LIWC_subset.csv` – LIWC-22 scores for platforms  

**Keyness folder**
- `keyness_1gram_pos_manifesto_dem_overuse.csv`, `keyness_1gram_pos_manifesto_rep_overuse.csv` – unigram keyness results  
- `keyness_2gram_pos_manifesto_dem_overuse.csv`, `keyness_2gram_pos_manifesto_rep_overuse.csv` – bigram keyness results  

**Collocations folder**
- `P2025_colloc_ADJ_w7_top10_per_node.csv` – adjective collocates (±7 words, sentence-bounded)  
- `P2025_colloc_NOUN_w7_top10_per_node.csv` – noun collocates  
- `P2025_colloc_VERB_w7_top10_per_node.csv` – verb collocates  

### 2. Multiple versions of the dataset
No.

### 3. Relationship between files
- LIWC files contain psycholinguistic scores for the text segments in the raw data files.  
- Keyness and collocation results were derived from the raw text data using the included Python scripts.  
- All analyses were visualized and reported through the R Markdown file `analysis_project2025.Rmd`.

### 4. Additional related data
None.

---

## C. SHARING / ACCESS INFORMATION

### 1. Was data derived from another source?  
Yes.  
All texts are publicly available:

- Dans, P., Groves, S., & Roberts, K. D. (2023). *Mandate for leadership: The conservative promise 2025.* The Heritage Foundation.  
- Woolley, J., & Peters, G. (n.d.). *The American Presidency Project.* UC Santa Barbara. Retrieved January 21, 2025, from https://www.presidency.ucsb.edu  

### 2. License
Creative Commons **CC-BY 4.0 International License**

### 3. Related publication
Schilling, J., & Fuchs, R. (2025). *Project 2025 as a technocratic blueprint: A corpus-based linguistic analysis of conservative governance discourse.* *PLOS ONE* (submitted).

### 4. Other publicly accessible data
None.

### 5. Links to ancillary datasets
None.

---

## D. METHODOLOGICAL INFORMATION

### 1. Methods for data generation
- Texts were extracted from *Mandate for Leadership* (Heritage Foundation, 2023) and party platforms from the American Presidency Project.  
- Preprocessing included tokenization, lemmatization, and POS tagging via **spaCy 3.7.2**.  
- Keyness analysis used **log-likelihood** and **log-ratio** (Hardie, 2014).  
- Collocates were extracted using a ±7-word, sentence-bounded window and part-of-speech restrictions.  
- LIWC-22 (Boyd et al., 2022) was used to compute psycholinguistic features.

### 2. Data processing
Data were cleaned, normalized to lowercase, and stripped of punctuation.  
Empty cells were replaced with “NA” to comply with repository formatting standards.

### 3. Software used
**R (4.4.1)**  
Packages: `tidyverse`, `tidytext`, `ggplot2`, `readr`, `dplyr`, `tidyr`, `stringr`, `scales`, `quanteda`, `quanteda.textplots`, `Matrix`, `rstatix`, `effsize`  

**Python (3.11)**  
Libraries: `spacy`, `pandas`, `numpy`, `matplotlib`

### 4. People involved

| Name | Role(s) |
|------|----------|
| **Julia Schilling** | Conceptualization, Data curation, Formal analysis, Methodology, Software, Visualization, Writing – original draft, Writing – review & editing |
| **Robert Fuchs** | Conceptualization, Methodology, Supervision, Writing – original draft, Writing – review & editing |

### 5. Quality assurance
Manual verification of text import, corpus statistics, and visualization outputs.  
No personal or sensitive data are included.

---

## E. DATA-SPECIFIC INFORMATION

### E.1 – Keyness Files (`keyness_1gram_*.csv`, `keyness_2gram_*.csv`)

| Variable | Description |
|-----------|-------------|
| `term` | Keyword identified as distinctive for the target corpus. |
| `freq_target` | Raw frequency of the term in Project 2025. |
| `freq_reference` | Raw frequency in the reference corpus (Democratic or Republican platforms). |
| `log_likelihood` | Log-likelihood (G²) statistic. |
| `bic` | Bayesian Information Criterion score. |
| `perc_diff` | Percentage difference in relative frequency. |
| `log_ratio` | Log-ratio measure of effect size. |
| `word_use` | Indicates whether the word is overused or underused. |

---

### E.2 – Collocation Files (`P2025_colloc_*.csv`)

| Variable | Description |
|-----------|-------------|
| `Collocation` | Node–collocate pair. |
| `Frequency` | Observed co-occurrence frequency (±7 words). |
| `PMI` | Pointwise Mutual Information. |
| `T-Score` | Measure of collocational strength emphasizing frequent co-occurrences. |
| `Z-Score` | Standardized deviation measure. |
| `Log-Likelihood` | Significance measure for association. |
| `Phi` | Effect size measure. |
| `Word_1`, `POS_1` | Node word and its POS tag. |
| `Word_2`, `POS_2` | Collocate word and its POS tag. |
| `LL_bin` | Significance category based on Log-Likelihood threshold: "LL≥3.84 (p<.05)", "LL≥6.63 (p<.01)", or "LL≥10.83 (p<.001). |

---

### E.3 – LIWC Files (`Project2025_LIWC_subset.csv`, `Platforms_Republicans_LIWC_subset.csv`, `Platforms_Democrats_LIWC_subset.csv`)

| Variable | Description |
|-----------|-------------|
| `id` | Unique segment ID. |
| `Headline`, `Subheadline` | Titles and section headings. |
| `text` | Analyzed text segment. |
| `year` | Year of document publication. |
| `party` | Political affiliation (Democratic, Republican, or Heritage Foundation). |
| `Segment` | Text segment identifier. |
| `WC`, `Dic`, `Analytic`, `Clout`, `Authentic`, `Tone` | Core LIWC summary metrics. |
| `focuspast`, `focuspresent`, `focusfuture` | Temporal focus indicators. |
| `power`, `achieve`, `affiliation` | Motivational drives. |

---

### E.4 – Project 2025 Raw Text File (`Project2025.csv`)

| Variable | Description |
|-----------|-------------|
| `Section`, `Chapter` | Section and chapter identifiers of *Mandate for Leadership*. |
| `Author` | Chapter author(s). |
| `Title`, `Headline` | Section title and heading. |
| `text` | Paragraph text. |
| `id`, `Segment` | Internal identifiers. |
| `WC`, `Dic`, `Analytic`, `Clout`, `Authentic`, `Tone`, `focuspast`, `focuspresent`, `focusfuture`, `power`, `achieve`, `affiliation` | LIWC-22 metrics as above. |



---

## CITATION

Schilling, J., & Fuchs, R. (2025). *Project 2025 as a technocratic blueprint: A corpus-based linguistic analysis of conservative governance discourse* [Dataset]. University of Bonn.  
License: CC-BY 4.0