NTIS Documentation

This documentation describes NTIS as a current, pre-commercial engineering and research project. Its central boundary is probabilistic perception; deterministic analysis.

Current architecture

1. Source and provenance. Deterministic source identity, source time, metadata, contribution records, attribution, and exact spans preserve what was available.
2. Perception and candidate artifacts. Structural and learned methods can emit bounded, unreviewed candidates with producer provenance. Confidence remains metadata, not analytical weight.
3. Adjudication. Explicit review accepts, rejects, supersedes, or corrects candidate interpretations while retaining history.
4. Canonical observation and reasoning state. Curated proposition observations and exact-span claim, premise, evidence, and conclusion elements form a ReasoningState.
5. Semantic alignment. Correspondence across states is an explicit accepted input, not an automatic semantic decision.
6. Deterministic comparison. The RCE compares eligible accepted states and reports structural transitions without scoring reasoning quality or cognitive condition.

Automatic-extraction limitations

Automatic perception remains experimental and human-supervised. Important unresolved tasks include proposition-holder resolution, temporal and status interpretation, reasoning-role classification, quotation/report attribution, and conservative abstention. Automatic relation extraction, semantic alignment, canonical comparison context, historical baselines, drift methodology, and prediction are not current capabilities.

Historical pages

The former pages on stylometry, Monte Carlo simulation, linguistic scoring, rhetorical classification, and signal calibration are retained as historical records. They are not active NTIS architecture and are intentionally omitted from current navigation.