Two sentence classifiers, compared across the whole corpus
Each Greek sentence is tagged by two independent classifiers
(both gpt-5.4-mini, temperature 0):
original-myth-history-other) —
the simple prompt that forces every sentence into exactly one of
mythic, historical, or other.greta-inspired-myth-history-other) —
two independent flags judged on each sentence's own content, calibrated to the
Greta/Rosie Book 3 base rates; a sentence may be both or neither.The two agree on the bucket for 82.0% of 11,302 sentences.
| Bucket | Original | Greta-inspired |
|---|---|---|
| mythic | 31.8% | 25.5% |
| historical | 30.4% | 27.4% |
| both | — | 1.6% |
| other | 37.8% | 45.5% |
Rows = Original, columns = Greta-inspired; the shaded diagonal is agreement.
| Original \ Greta-inspired | mythic | historical | both | other |
|---|---|---|---|---|
| mythic | 2,668 | 42 | 77 | 811 |
| historical | 82 | 2,763 | 95 | 494 |
| other | 137 | 288 | 12 | 3,833 |
| Book | Sentences | Agree | Original | Greta-inspired | |||||
|---|---|---|---|---|---|---|---|---|---|
| myth | hist | other | myth | hist | both | other | |||
| Book 1 | 1,354 | 82.0% | 34% | 36% | 31% | 27% | 32% | 3% | 38% |
| Book 2 | 1,258 | 82.5% | 46% | 14% | 40% | 36% | 13% | 1% | 50% |
| Book 3 | 959 | 80.8% | 32% | 33% | 35% | 24% | 28% | 3% | 44% |
| Book 4 | 1,070 | 75.3% | 24% | 50% | 26% | 21% | 37% | 2% | 40% |
| Book 5 | 944 | 81.9% | 30% | 23% | 47% | 24% | 23% | 1% | 52% |
| Book 6 | 879 | 79.5% | 16% | 40% | 45% | 11% | 46% | 1% | 42% |
| Book 7 | 930 | 84.5% | 24% | 42% | 34% | 18% | 37% | 2% | 43% |
| Book 8 | 1,473 | 85.6% | 32% | 23% | 45% | 27% | 21% | 1% | 51% |
| Book 9 | 1,139 | 82.8% | 42% | 20% | 37% | 34% | 19% | 1% | 46% |
| Book 10 | 1,296 | 82.8% | 31% | 30% | 38% | 24% | 27% | 2% | 47% |
Click a book to see every sentence where the two classifiers disagree.