RQ4 · structural recurrence
Cross-saga comparison
Only Njáls saga has a hand-annotated evaluation; the other three run the same bootstrap entity table and rule pipeline without curator review, so treat their absolute numbers as provisional (see Methods). What is worth comparing is whether the same structural shape recurs.
Basic network statistics
| Saga | Nodes | Edges | Density | Avg. clustering | Communities | Modularity | Assortativity | Balance ratio |
|---|---|---|---|---|---|---|---|---|
| Njáls saga (Brennu-Njáls saga) | 160 | 382 | 0.030 | 0.235 | 10 | 0.553 | -0.152 | 71.4% |
| Laxdæla saga | 121 | 227 | 0.031 | 0.211 | 9 | 0.495 | -0.159 | 93.8% |
| Grettis saga Ásmundarsonar | 119 | 213 | 0.030 | 0.154 | 14 | 0.466 | -0.165 | 75.0% |
| Egils saga Skallagrímssonar | 95 | 165 | 0.037 | 0.148 | 11 | 0.524 | -0.127 | 90.0% |
Does the shape recur?
Degree assortativity (all four negative → hub-and-spoke, not clique-of-hubs)
Structural balance ratio
RQ5 · centrality vs. narrative prominence, by saga
| Saga | mentions ~ degree | mentions ~ betweenness | mentions ~ pagerank | dialogue ~ pagerank |
|---|---|---|---|---|
| Njáls saga (Brennu-Njáls saga) | 0.787 | 0.588 | 0.793 | 0.62 |
| Laxdæla saga | 0.786 | 0.678 | 0.675 | 0.721 |
| Grettis saga Ásmundarsonar | 0.647 | 0.573 | 0.56 | 0.504 |
| Egils saga Skallagrímssonar | 0.674 | 0.589 | 0.64 | — |
Spearman ρ between raw mention/dialogue counts and network centrality. Consistently strong positive correlation across all four sagas is itself informative: it suggests structural centrality and textual prominence are not independent measurements of “importance” here, largely because both are downstream of how often a character is written about at all.