Data analytics / Tableau

Book of Mormon — Authorship & Linguistic Analysis

A personal project connecting my interest in data analytics with my faith. Using 6,585 verses, I explored how word choice, vocabulary, and verse length vary across the text’s traditionally attributed authors with Python, natural language processing, and Tableau.

Interactive dashboard

Use the dashboard controls to explore the data. On smaller screens, scroll horizontally if needed. If the dashboard does not load, use “Open in Tableau” above.

Key findings

Distinct language patterns emerge across the assigned author groups, alongside shared religious themes.

Key Takeaway

To go beyond visual comparisons, I ran a one-way ANOVA on verse length across all primary author groups. I excluded “Various” from the book of Omni because it combines five authors with very small samples, making it unsuitable as a single author group for this comparison.

Results:

  • F-statistic: 7.1096
  • P-value: approximately 0.000001

If the author groups had equal mean verse lengths, the chance of observing an F-statistic this large or larger would be about 1 in 1,000,000, assuming the ANOVA assumptions hold. The differences in mean verse length are statistically significant, providing evidence beyond visual comparisons that the groups differ.

Or, in other words, if the author groups had equal mean verse lengths, a result this extreme or more extreme would occur about 1 in 1,000,000 times under the test’s assumptions.

Visualizations

Compare word use, vocabulary diversity, verse length, and similarity. Select an available chart to view the full image.

Which words occur most by author?

Authorship fingerprints — word clouds
Shared religious themes — word clouds

Methods & interpretation

Statistical results: The project reports a one-way ANOVA on verse length (F = 7.1096, p ≈ 0.000001) and a chi-square test on word usage (χ² = 20,919.81, reported p ≈ 0). These results indicate differences across the assigned author groups under the tests’ assumptions; they do not establish authorship or explain the cause of those differences.

Scope & limitations: The dataset contains 6,585 verses, with book, chapter, verse, assigned author, word count, and text. Attribution follows the project’s traditional author and abridger assignments. The combined “Various” category in Omni was excluded from ANOVA. Text length, subject matter, and grouping decisions matter when interpreting the results.

Tools: Python, Pandas, NLTK, scikit-learn, SciPy, Matplotlib, WordCloud, and Tableau.