Convert a news article (Layer 3 blended reporting) into a Layer 2 output:
Observed sequence
Attribution
Claims vs evidence
Unknowns
No motive inference
No evaluative framing
This protocol removes narrative layering without adding external fact-checking or interpretation.
Before Step 1:
Exclude headlines, subheadings, photo captions, sidebar links, embedded promotions, and unrelated image descriptions unless they directly describe the same event chain as the article body.
Do not treat media decoration as part of the event sequence.
Explicitly ignore ALL-CAPS promotional lines, unrelated image captions, and embedded video teasers.
A claim must be tagged as one of the following structural types. Tag based on structural container, not semantic content.
Exact words in quotation marks attributed to a named speaker.
The existence of the quote is the fact.
The content may be opinion.
Paraphrases are not F1.
A reference to:
bill text
court filing
official vote tally
transcript
report
dataset
institutional rule
primary document
Legislative provisions belong here.
If the article does not explicitly name the record, tag F?.
A time-bound action that occurred:
vote
speech
filing
announcement
meeting
release
resignation
arrest
Must be placeable in sequence.
Speech acts must be separated from their content (see Step 1 clarification below).
A number, percentage, measurement, or quantified result explicitly stated in the article.
If a claim appears factual but no source is named in the text, tag F? and mark:
“source not provided in text.”
Everything else must be tagged as:
Language that adds evaluative meaning beyond what is directly observable in the event, quote, or named record.
Claims about why someone acted, unless directly quoted or documented in a record.
Claims about what will happen or impact statements without evidence.
Cause-effect linkage without a named mechanism or evidence trail.
Factual-sounding claim with no named speaker or source inside the article.
Narrative emphasis, certainty inflation, moral labeling, partisan summary language, or loaded verbs.
Language that adds meaning without introducing a new record:
evaluative adjectives/adverbs
motive attribution
mind reading
certainty inflation
moral labeling
causal language without mechanism
summarizing language replacing numeric precision
abstract synthesis nouns (e.g., “reasoning,” “strategy,” “blow,” “impact,” “agenda”)
Break the article into short statements containing one claim each.
Do not merge multiple events into one claim.
If a sentence contains a speech act plus content (e.g., “X said that Y happened”), split into:
The speech act (F3 Event)
The content claim (F1 if verbatim, otherwise N5/F?)
Do not include excluded container material.
Assign exactly one structural tag (F1–F4 or N1–N6).
Tag by structural type, not by perceived accuracy or truth value.
Do not:
Double-tag
Collapse categories
Upgrade paraphrases into quotes
Upgrade article statements into named records
If a record is not explicitly named in the article, do not assume one.
Output in this order:
What the article says happened (sequence only — completed events)
What was said (verbatim quotes only)
What documents/data are explicitly referenced
What is asserted without support
Where narration was added and removed
Unknowns and missing links (article-scope only)
Minimal neutral summary (2–5 lines)
Separate sequence from procedural status if needed.
When rewriting into Layer 2:
Remove adjectives/adverbs unless part of a quote.
Remove motive unless directly quoted or documented.
Replace “caused / led to” with strict sequence separation unless mechanism is named.
Do not replace causal language with softer implied linkage (e.g., “following,” “after”) unless the article explicitly establishes sequence only.
Replace loaded verbs with neutral action verbs.
Preserve politically charged nouns unless violating other purge rules.
Do not replace precise numbers with summarizing language.
Do not introduce chronological connectors (e.g., “following,” “after,” “ensuing”) unless explicitly stated with a date or sequence marker in the article.
Example:
“All but one” must remain numeric.
Do not rewrite as “near-unanimous.”
Before finalizing, confirm:
Did I include container material improperly?
Did I bundle speech acts and claims?
Did I treat paraphrase as quote?
Did I upgrade an article statement into a named record?
Did I sneak in motive?
Did I use abstract synthesis nouns?
Did I compress numeric precision?
Did I imply causation via sequence language?
Did I convert someone’s claim into narrator fact?
Did I expand beyond the article’s scope when listing unknowns?
If yes, correct it.
[timestamp if given] [event] (source in text: article statement / named record)
…
“[quote]” (speaker, context)
…
[record explicitly named in article] (who produced it, if stated)
…
[assertion] (why unsupported: no named source / no record / implied motive)
…
Original phrasing: ___
Layer 2 rewrite: ___
Category: characterization / motive / rhetorical framing / certainty inflation / summarization / abstract synthesis
What is not provided: ___
What would be needed (within article scope): ___
(Do not speculate beyond the article.)
2–5 lines.
Sequence + attribution only.
No framing.
No motive.
No abstract synthesis nouns.
No compression.
No causal inference.