Build a source map that separates origin from repetition
A source map should show where information originated and how each later object relates to it. Label relationships such as repost, translation, amendment, quotation, or independent observation. Without those labels, a pile of links can make one issuer message look like broad corroboration.
W3C provenance offers background for representing source relationships. The SEC exhibit discussion illustrates why documents can point to earlier source objects. The source-edge list below is a practical manual convention, not a complete provenance ontology.
Define nodes by source identity
Give each source a stable local ID with title, publisher, date, version, and direct reference. A node represents an identified object, not a whole website. The same publisher may produce an original announcement, a correction, and a presentation that serve different evidentiary roles.
Add edges only when the relationship is supported. A news story that explicitly quotes a release can be linked as a quotation or summary. A similar story with no established origin should remain an unverified relationship. Do not infer independent reporting merely because the wording differs.
Hypothetical source-edge list
Assume fictional Lakeside Freight issues a project notice and several related documents appear. All details are invented.
| From | Relationship | To | Evidence role |
|---|---|---|---|
| Issuer notice S1 | Translated as | Language file S2 | Same issuer message |
| Issuer notice S1 | Summarized by | News article S3 | Secondary lead |
| Issuer notice S1 | Corrected by | Correction S4 | Version relationship |
| Authority decision S5 | Addresses same approval | Project event E1 | Separate action record |
| Research note N1 | Uses passage from | S1 and S5 | Derived interpretation |
The map may show five documents but only some provide distinct observations. The authority record can independently establish its own decision. It does not necessarily verify every benefit the issuer expects. Independence must be evaluated for the specific claim, not awarded to an entire document without qualification.
Use the map to answer a claim-specific question
Choose one claim and trace backward to the documents directly supporting it. Mark what each source actually contributes. If all paths end at the same release, describe the evidence as one issuer origin with multiple distribution routes. More links do not change the underlying observation count.
For a claim using an agreement, follow references to the actual agreement and relevant amendment. Keep the index as a navigation node if useful, but do not treat its short description as equivalent to the contract text. The map should preserve the difference between route and evidence.
Verify edges, not just nodes
Open the source and confirm the relationship used in each important edge. Does the news article cite the release? Does the correction identify the original? Does the research note actually use the referenced paragraph? A map with real documents but guessed relationships can still mislead.
When a relationship remains uncertain, label it provisional and state what would establish it. The map is ready for review when another reader can follow the main claim paths and identify both verified and provisional connections. It need not include every peripheral mention on the web.
Limits and common mistakes
A source map does not decide truth automatically, prove causation, or make all official sources independent. Different organizations can repeat the same information. Conversely, one organization can provide separate observations with distinct roles. Evaluate the claim and activity involved.
Common mistakes include using domain names as whole source identities, counting translations as corroboration, and leaving edge meanings undefined. Another is drawing every connection as “supports” when some merely repeat, qualify, or contradict. Specific relationship labels make the map useful for reasoning rather than decoration.
A final record check
Test the map by temporarily hiding secondary sources. Can the core claim still be traced to its originating evidence? Then restore the secondary nodes and describe what each adds: access, explanation, challenge, or a distinct observation. This exercise makes repetition visible without dismissing useful secondary material merely because it is not the original source.
FAQ
Do I need a graph application?
No. A table of source IDs and labeled relationships is sufficient for many small projects. Use a diagram only when it makes the paths easier to inspect.
Does an independent publisher mean independent evidence?
No. It may be repeating the original source. Inspect the reporting basis and what distinct observation, if any, supports the claim.