User talk:Henning (WMDE)/Wikibase/Concepts/Aliases

From mediawiki.org
Latest comment: 9 years ago by Denny

I generally agree, though I would suggest another alternative: Replacing aliases (and maybe even the label) with statements. This would reduce duplication, since users currently already represent many aliases as statements. It makes sense to have statements for aliases, since they are not just random â€șother labelsâ€č, but have a specific origin and meaning, which you cannot denote in the aliases list. I think aliases were only introduced for easy lookup. We need to get over this for units, anyway, so we could just as well get over it for everything. -- Adrian

I agree with the analysis - alias is a bad term. But it is, in the end, just a name for that concept in Wikidata, and has not much relation with the real-world concept of an alias. If you prefer more technical names think of it as in the following:

  • label + description: used to identify an item, i.e. the identifier
  • aliases: used to provide additional indexes for searching an item, i.e. inverted index entries

Additionally, every label is an also an inverted index entry.

(A single, unique label for every item is extremely helpful in all UIs, and together with the description it provides the identity criterion for all Wikidata items.)

It is true that many aliases are also values for a statement, but this is unlikely true for all, and even if it was, it would make things more complicated due to qualifiers, references, etc.

But in the end: labels, descriptions, aliases are really there to support the UI, and not part of the knowledge base proper. Aliases - or inverted index entries, if so preferred - are there only to facilitate search, and not to be part of the knowledge base.

Having said that, it would be a perfectly interesting solution to also add other textual or string-values to the search index, i.e. the values of P225 could also be added automatically as an inverted index entry (just as the label is). In particular, a property could be marked to say "use the values as index entries". Although this has possible side-issues (what if a value is correct for the statement, but would be embarrassing or problematic as a search term for the given item? should we do something about ranks? qualifiers?), it would still be a nice solution.

But label+description still have a special role, and cannot be completely replaced by statements, and I think that this is true for aliases as well. --Denny (talk) 21:20, 16 March 2015 (UTC)Reply