Evidence and Sources
NataPulse is evidence-oriented. Every product conclusion should be inspectable through the sources, events, reports, or memory items that support it.
Source families
Section titled “Source families”The product can use evidence from:
- primary and regulatory sources;
- permitted financial news feeds and official interfaces;
- verified or reviewed social sources;
- broader social discussion;
- market data;
- on-chain data;
- curated quantitative signals;
- prior NataPulse reports and memory;
- permitted live web research in General Deep Research mode.
Coverage depends on enabled providers.
Provenance
Section titled “Provenance”A source reference may include the provider, publication or account, original URL, source timestamp, retrieved timestamp, and a relation to the event. Product pages expose only safe, relevant provenance fields.
Repetition versus corroboration
Section titled “Repetition versus corroboration”Ten outlets repeating the same wire story are not necessarily ten independent confirmations. Corroboration is stronger when evidence comes from genuinely independent source paths, such as:
- a primary filing and a market reaction;
- an official company statement and independent reporting;
- a social claim and an on-chain transaction;
- a market anomaly and a later news catalyst.
NataPulse evaluates source diversity, temporal proximity, source relationships, and quality rather than counting every mention equally.
Primary, secondary, and social evidence
Section titled “Primary, secondary, and social evidence”Primary evidence usually describes the underlying action directly: a filing, transaction, official release, or market print. Secondary evidence interprets or reports it. Social evidence can be early and valuable, but also ambiguous or manipulative.
The system therefore does not apply one fixed trust level to every source in the same family. A reviewed corporate account, an anonymous post, and an official filing have different evidentiary roles.
Evidence in product pages
Section titled “Evidence in product pages”- Event Explorer provides numbered source cards and a cross-source timeline.
- Reports contain references and quality warnings.
- Deep Research groups source evidence and shows limitations.
- Analyst Chat places citations alongside the answer.
- Social Radar displays a reliability or rumor assessment.
- Quantitative evidence carries a fixed probabilistic disclaimer.
- Observation vs published event
- An observation is raw provider input; a published event has passed validation and the publication gates.
- Importance vs confidence
- Importance is how much an event may matter; confidence is how well the evidence supports it. They are scored separately.
- Repetition vs corroboration
- The same claim repeated is not stronger evidence. Independent corroboration from distinct source families is.
- Cluster vs emerging narrative
- A cluster groups related events; an emerging narrative is a detected theme forming across clusters over time.
- Cross-source confirmed
- Marked when independent source families report the same event within a short window.
- Rumor
- Flagged when a single low-authority source carries an unconfirmed claim.