Forward Pairwise Report
SMDG as query side and Txt-18 as corpus side. Includes corpus heatmap, pair heatmaps, stream profiles, top-k match modes, and click-to-read sentence context.
Corpus-scale Tibetan embedding analysis
Tibetan Intertext Lab
Static research reports and reproducible tooling for exploring semantic affinity, directional coverage, and candidate intertextual parallels across Tibetan corpora.
This landing page is a report hub. The current corpus run is one checkpoint, not the final shape of the project.
Current Reports
Reverse Pairwise Report
Txt-18 as query side and SMDG as corpus side. Useful for checking direction sensitivity in decoder-style asymmetric embedding protocols.
Bidirectional Survival Report
Joins forward and reverse runs, then prioritizes pairs whose broad, local, or directional signal survives both directions.
What This Site Shows
Layer 1 analysis: full sentence interaction matrices, document-pair aggregation, heatmaps, directional coverage profiles, and top sentence-pair evidence with local context.
Next Research Layer
Layer 2 should add null calibration, continuity-aware passage detection, and specificity correction so descriptive similarity becomes stronger inferential evidence.
Code Pathways
Use scripts/run_bidirectional_corpus_pairwise.py for CLI runs, or run_bidirectional_corpus_pairwise(...) / TibetanResearchSDK.bidirectional_corpus_pairwise(...) from Python.