Selected Pair Heatmap
This is a downsampled view of the full sentence-by-sentence matrix. Full `.npy` paths remain in the pair artifacts.
Selected Pair Metrics
Use these as triage signals, not as final philological claims. Strong candidates still need close reading.
Whole-Text Stream Profile
Method: for every sentence in one text, the report finds its strongest sentence-level embedding matches anywhere in the other text, then plots the selected aggregation in the original sentence order. The upper line is SMDG to Txt-18 coverage; the lower line is Txt-18 to SMDG coverage.
What to read from it: Best 1 highlights the strongest local contact. Top 3 average and Top 5 average are more conservative because they require several good counterparts, not just one spike. Sustained elevated regions suggest broader passage-level coverage. This is a directional nearest-neighbor coverage profile derived from the full similarity matrix, related to accepted similarity-matrix, alignment, and text-reuse workflows. It is exploratory evidence for close reading, not proof of borrowing or historical direction by itself.