Wikidata aims not to create a database of facts about the world, but to collect references to such knowledge. As such, Wikidata will be a secondary knowledge base. To every statement in Wikidata there will be the option to add references to primary or referenceable sources. This means that in Wikidata we will not discuss what the population of Germany actually is, nor will Wikidata provide this information, but it will provide the information what the population of Germany is according to a specific source, like the 2012 CIA world fact book.
For each statement in Wikidata we will thus also collect references for this statement. The question is, how should references and their connections to statements be be modeled and exported?
Some assumptions about references
The following assumptions are to be discussed:
- Eventually, a reference is just a pointer. It is not the primary goal of Wikidata to model detailed properties of references. In its simplest form, a reference is just an unconstrained string, pointing a human reader to some source.
- Some references may also be items. For instance there may be a Wikipedia article about a book, that is also used as reference
- References may be connected by the following relationships, that are defined based on their use as reference:
- There is a hierarchical relationship between a broader reference A and a narrower reference B if and only if every possible statement referenced by B can also be referenced by A. For instance a specific chapter, section, page or sentence in a book is a narrower reference compared to the whole book.
- There is a derivation relationship between a reference A and a derived reference B if most of the statements from B can also be referenced by A. For instance a translation or an edition of a text is derived from the original. Editions and translations may alter, drop and add some statements but it is assumed that and most statements can also be referenced by a different edition or by a translation.
- A reference may provide contradicting statements, for instance a book may contain several chapters where different authors give contradicting information.