PageGod Golden Narrative Scanner

Read full text Summarise with VIRAT
Authors: Joshua Kirchner
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Version 2 documents PageGod Golden Narrative Scanner, a deterministic narrative-anomaly detector implemented in local Python and scored against explicit answer keys. The submission-facing stable baseline remains v43.9-local-expanded-corpus in LOCAL_ONLY__ANOMALY_ENGINE_V38.py. On ALL_17, that stable baseline reports 204 matched findings, 26 unexpected findings, 114 missed findings, and F1 0.7445, with Blackwater F1 0.7273 and expanded-corpus micro/macro F1 0.7527/0.7516. The external audit still shows 6 of 16 accepted ConStory-derived expected findings were false ground truth. Version 2's main upgrade is the prose-tolerant experimental branch V44. On the independently authored blind manuscript INTENTIONS_WAKE_CH01-CH30_CLEAN_FINAL_v1.0, the historical blind winner was 8 TP / 7 FP / 22 FN (F1 0.3556), the entering active ceiling was 11 TP / 8 FP / 19 FN (F1 0.4490), and the confirmed V44 state now reaches 19 TP / 1 FP / 11 FN (F1 0.7600) with zero watch promotions in two consecutive runs with identical matched sets. The matched V44 promote packet also held the keyed floor packet at ALL_17 F1 0.7482, expanded-corpus micro/macro 0.8000/0.7975, Blackwater F1 0.7273, and filtered external 2 TP / 0 FP / 8 FN. The report also records closure of the Track 2 object-soul design gate, with post-scan reconciliation chosen as the first implementation path. This Version 2 record supersedes the earlier public Zenodo paper for the March 22, 2026 release.

Fields of Study & Subjects

Topic Modeling Computational and Text Analysis Methods Advanced Text Analysis Techniques

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
22 March 2026
Language
EN
Document type
Preprint