Source representation
Each verse is represented using its Arabic text together with available Tafsīr al-Jalalayn content. Two verses without commentary use the verse-only fallback described in the artifact.
Qur’an Explorer · Research prototype
Surah Al-Baqarah semantic map
Explore how an experimental classifier places all 286 verses across six broad thematic classes. Switch between t-SNE and PCA, isolate a class, reveal the anchor verses, or inspect every result in the table.
Interactive visualization
Computational classification
Not Tafsīr
Research status. This prototype embeds verse text with Tafsīr al-Jalalayn using BGE-M3, then applies similarity-weighted k-nearest-neighbor classification against 19 hand-labeled anchors. The thematic labels are experimental computational outputs and require methodological and scholarly review.
The visualization includes marker shapes in addition to color and a filterable table containing the same 286 verses. Use the controls inside the frame to change projection, display anchors, or isolate a class.
How to read it
A map of model similarity, not a map of meaning itself.
The projection is useful for forming questions and auditing a pipeline. It does not establish an authoritative thematic interpretation of any verse.
Experimental labels
The six classes—Theology, Eschatology, Death, Prophethood, Ethics, and Jurisprudence—come from a small anchor set. They are research categories, not a complete taxonomy of Al-Baqarah.
Confidence boundary
The classifier leaves 112 verses unclassified rather than forcing a weak label. Sparse anchors and verses containing several themes limit coverage.
Two projections
t-SNE highlights local cluster structure; PCA provides a linear comparison. Relative proximity is exploratory and should not be read as theological distance.
Evidence view
Every plotted point can be traced back to a verse key and Arabic text. The table provides a linear, searchable version of the visualization.
Review required
Production use requires verified source versions and licenses, a documented normalization pipeline, broader reviewed anchors, held-out evaluation, and qualified scholarly review.
Study safeguard. Clusters and labels are computational observations. Return to the Qur’anic text, established Tafsīr, and knowledgeable teachers for interpretation.