Tutorial: Federalist Papers

Reproducing Mosteller & Wallace's (1964) classical authorship attribution of the 85 Federalist Papers.

Background

The Federalist Papers (1787–1788) were published under the pseudonym Publius to argue for ratification of the US Constitution. Authorship of 73 papers is known (Hamilton, Madison, Jay); 12 are disputed between Hamilton and Madison. Mosteller & Wallace (1964) used word-frequency Bayesian inference to attribute all 12 disputed papers to Madison — a result confirmed by every subsequent stylometric analysis.

This tutorial uses tamga to reproduce the essentials of their result: training Burrows Delta on the known Hamilton / Madison papers, projecting the disputed papers onto the learned space, and visualising the separation with PCA and a Ward dendrogram.

What you'll build

By the end you will have:

• A project skeleton with the 85 Federalist Papers ingested.
• A study.yaml declaring four analyses: Burrows Delta, PCA, Ward cluster, Craig's Zeta contrast between Hamilton and Madison.
• A results/demo/ directory containing per-method Result JSONs plus rendered figures.
• A single HTML report stitching everything together.

1. Initialise the project

tamga init federalist
cd federalist

This scaffolds a project directory with an empty corpus/ and a starter study.yaml.

2. Drop in the papers

The repo's examples/federalist/ directory has all 85 papers as individual .txt files plus a ready-made metadata.tsv. Copy corpus/ and metadata.tsv over, or follow the example's own README.md to build it from Project Gutenberg.

metadata.tsv has one row per paper with: filename, author, number, role (train for known-author papers, test for disputed).

3. Edit study.yaml

name: federalist
seed: 42
output:
  dir: results
  timestamp: false

corpus:
  path: corpus
  metadata: corpus/metadata.tsv
  filter:
    role: [train]            # hold out the disputed papers from training

features:
  - id: mfw200
    type: mfw
    n: 200
    scale: zscore
    lowercase: true

methods:
  - id: burrows
    kind: delta
    method: burrows
    features: mfw200
    group_by: author

  - id: pca
    kind: reduce
    features: mfw200
    params: { n_components: 2 }

  - id: ward
    kind: cluster
    features: mfw200
    params: { n_clusters: 3, linkage: ward }

  - id: zeta_hamilton_madison
    kind: zeta
    group_by: author
    params:
      top_k: 50
      group_a: Hamilton
      group_b: Madison

The filter: role: [train] line hides the disputed papers during training so Delta has a clean Hamilton / Madison centroid; we project the disputed set back in at the analysis step.

4. Run the study

tamga run study.yaml --name demo

Expect per-method directories under results/demo/:

results/demo/
├── resolved_config.json
├── burrows/
│   └── result.json
├── pca/
│   └── result.json
├── ward/
│   └── result.json
└── zeta_hamilton_madison/
    ├── result.json
    ├── table_0.parquet     # Hamilton-preferred vocabulary
    └── table_1.parquet     # Madison-preferred vocabulary

5. Render figures

Matplotlib rendering is a thin post-processing step (full integration lands in a later phase). The example ships a render_figures.py you can invoke:

python examples/federalist/render_figures.py results/demo metadata.tsv

This produces pca.png, ward.png, and zeta.png inside each method directory.

6. Report

tamga report results/demo --output results/demo/report.html

Open the HTML in a browser — you get a single-page report with the method sections, embedded figures, and the full provenance JSON.

Expected outcome

On PCA the Hamilton and Madison papers form two tight clusters along the first two components (together ~35 % variance), with Jay's five essays at the margin. Burrows Delta at MFW=200 attributes every disputed paper to Madison — matching Mosteller & Wallace's 1964 result.

The quickstart mini-version of this tutorial is at examples/quickstart/ if you want to run through the pipeline on just 9 papers first.