Wikipedia talk:Authority control integration proposal: Difference between revisions
Content deleted Content added
Stuartyeates (talk | contribs) reply |
→VIAF as main source: interpretation |
||
Line 200:
:: --[[User:P64|P64]] ([[User talk:P64|talk]]) 22:58, 27 April 2014 (UTC)
:::I think the thing to remember is that [[VIAF]] is categorically not a [[WP:Reliable_source]]. [[User:Stuartyeates|Stuartyeates]] ([[User talk:Stuartyeates|talk]]) 23:24, 27 April 2014 (UTC)
::::Well, authority control in a sense always is a case of [[WP:NOR]], since you'll never find somebody you can quote that this article and that authority record pertain to the same entity. Probably Kolja is aiming at the distinction between ''real'' Authority Files and ''virtual'' ones like VIAF. Records in Real Authority Files consist of a bunch of data (names, dates, professions, ...) plus the implicit assertion that this data coherently describes an individual entity which is not described by any other record. Ideally there also exists some editorial infrastructure apt to process corrections. VIAF records (clusters) on the other hand solely consist of a bunch of references to Real Authority Records plus the assertion that all records in the same cluster pertain to the same entity. Clustering is performed algorithmically, manual combination or deduplication is possible however reduced to the absolute minimum. Expectations on coherence and uniqueness are loosened very much in VIAF for obvious reasons (Often the individual authority files have divergent data, if all of this would be taken at face value VIAF clusters would be coherent too, but wouldn't be good at clustering any more). Now suppose Real Authority Files qualify as ''reliable sources'' (taking them as databases created under professional standards and so on) and the question arises wether VIAF should be considered a reliable source too: Yes, because the clustering is usually successful and helps to "level" the many glitches in the Real Data, or Neutral because a compilation of sourced items does not substantially change their nature, or No, because the algorithmic processing defies any assessment of reliability for reasons of principle (''because'' the clustering is usually more successful than the actual data would allow). I think as an epistemic question this cannot easily be decided. -- [[User:Gymel|Gymel]] ([[User talk:Gymel|talk]]) 01:24, 28 April 2014 (UTC)
| |||