In development
A reference system for the movement of public meaning
Discourse Meridian is being developed to track how public interpretations are observed across media and institutional documents, with statements linked to their source evidence. First domain: European defence autonomy. Explore the stored corpus below; analytical discourse-change findings are not yet available.
- Month
- 31 / 54
- Highest class reached
- L6 · Strategy
- Steps above L1
- 5
- Observations (synthetic)
- 0
Synthetic illustration only — not EDIS results. Grey points are background observations; dark points represent the illustrative frame. Orange squares mark its schematic ladder steps. Move across the plot or use the keyboard-accessible month control.
Every conclusion decomposes into six parts
The unit of analysis is the evidence span: a passage with exact character offsets, its frame, stance and conditionality, and the actor who holds the proposition. Report statements are built only from registered sets of spans.
- 01ClaimWhat changed, stated precisely.
- 02MeasurementThe defined metric and the method version that detected it.
- 03EvidenceThe dated documents and spans that support it.
- 04Counter-evidenceSignificant sources that cut against it.
- 05CoverageWhat the corpus included and left out, in five layers.
- 06UncertaintyWhat the system cannot assert.
No metric without a definition, a denominator and an error model
Fact
Countable and verifiable: coverage, provenance chains, threshold crossings, dated adoption sequences across segments.
Model-derived
Always with uncertainty: frame and stance distributions, frame divergence between segments, bounded shifts.
Research
Never in reports. A method stays here until it passes a published validation study.
- No causal attribution from observational data.
- No event, crisis or conflict prediction.
- No truth adjudication. Frames are measured, not fact-checked.
- No intent attribution. Actors are not classified as malicious, coordinated or inauthentic.
- No single-number ranking of actors' importance.
- No persuasion or message-testing features.
The project commits to the same versioned method for every client. Segments are defined in advance by institutional properties, rather than content. The current corpus guide discloses its coverage and limitations.
A real corpus. A method still in development.
The corpus guide shows stored document and passage counts, an indexed EDIS source record, downloadable tables and input fingerprints. It does not report frame/stance changes, adoption sequences or validated model performance.
A retrospective evaluation with a temporal holdout is planned. The evaluation design requires external protocol review and a signed preregistration freeze before evaluation-period computation. No deposited DOI or completed evaluation is claimed.