No theology. No human labels. We gave the text to three unsupervised algorithms and asked: what patterns do you find? The Bible organized its own story.
The approach: Embedding clusters, semantic field analysis, and transformer attention. Three independent methods, zero supervision, same conclusion.
Section A
The Bible Organizes Itself
We converted every passage to a mathematical vector and let UMAP + HDBSCAN find natural groupings. No one told the algorithm what the Bible is about. Here is what it found.
843
Passages
29
Clusters
12
Cross-Testament
219
Noise Points
How to read this chart: Each dot is one of the Bible's 843 passages, positioned by semantic similarity. Passages that say similar things cluster together. Colors represent clusters the algorithm found on its own. Hover over any point to see the passage reference and book. Notice how books from different centuries and authors land in the same clusters.
UMAP Embedding Space
12 Clusters That Bridge the Testaments
These clusters contain passages from both the Old and New Testaments. The algorithm found these connections with zero guidance. Books written 400-1,500 years apart, by different authors, in different languages, land in the same semantic neighborhood.
Surprise Placements
These passages ended up in unexpected clusters. The transformer sees connections that are not obvious on a surface reading.
Hebrews 1:1-2:18 ("The Son Superior to Angels") clusters with the nativity stories
Cluster: Child / Mary / Angel. The algorithm sees the Incarnation theme bridging angelic announcements and the theological argument of Hebrews.
Romans 11:1-36 ("The Remnant, the Olive Tree") clusters with the David narrative
Cluster: Samuel / Eli / David. Paul's argument about Israel's remnant carries the same semantic DNA as the stories of God preserving a faithful line through David.
Proverbs 31:10-31 ("The Excellent Wife") clusters with praise poetry
Cluster: Praise / Like / Loving. The famous acrostic poem shares deep structural and thematic similarity with the Psalms of praise.
Exodus 12:1-30 ("The Passover Instituted") clusters with Levitical offerings
Cluster: Offering / Aaron / Altar. The Passover narrative reads more like sacrificial law than historical narrative to the model.
Ezekiel 45-46 ("The Holy District; Feasts and Offerings") clusters with Moses/Joshua
Cluster: Moses / Joshua / Pharaoh. Ezekiel's future temple vision carries the same semantic fingerprint as the original Sinai legislation.
Section B
Concept Networks
We built a co-occurrence graph of every meaningful concept in Scripture, then used community detection to find natural semantic fields. These are the Bible's own topic clusters.
369
Concepts
6,762
Edges
25
Semantic Fields
100
OT-NT Bridges
What is a semantic field? When two concepts consistently appear in the same passages, they form a link. The algorithm groups tightly-linked concepts into "fields" — natural topic areas the text itself defines. Hub concepts (highlighted in gold) connect the most fields, acting as the backbone of Scripture's conceptual network.
Backbone Concepts — The Hubs of Scripture
The 25 Semantic Fields
Concept Trajectories
How concepts evolve: The same word changes its neighborhood as you move through Scripture. "Lamb" starts near offerings and Moses in the Pentateuch, then reappears near suffering and sacrifice in the Prophets, before finally landing next to Jesus, Spirit, and glory in the New Testament. Watch how the Bible's core concepts weave an unbroken thread.
Strongest Cross-Testament Bridges
Concept pairs that appear together in both the Old and New Testaments. High strength means the pair is frequent and tightly linked across the divide.
Section C
What the Transformer Sees
We ran every passage through a BERT transformer and analyzed its attention patterns. Where does the model look? What words matter most? What hidden connections does it find between the testaments?
88
Novel OT-NT Links
843
Passages Ranked
150
Content Words Analyzed
Most Important Passages
Importance = attention density + centrality + connection density. These are the passages the transformer considers most semantically central to the entire Bible. Notice how many come from unexpected places — not the "famous" verses, but passages that are deeply connected to everything else.
Most Attended Content Words
These are the words the transformer pays most attention to relative to their frequency. Common words like "the" get high raw attention but low frequency-normalized scores. The words below punch above their weight — the model treats them as particularly meaningful.
88 Hidden Connections
The transformer found 100 high-similarity OT-NT passage pairs. Only 12 are well-known prophecy-fulfillment links. 88 are novel — the model sees deep semantic echoes between passages that no standard cross-reference list connects. These are new discoveries.