WikiCite 2016/Report/Group 1/Notes

From Meta, a Wikimedia project coordination wiki
Jump to navigation Jump to search

Links and notes[edit]

Goal[edit]

  • Find a consensus over how to represent cited sources (in Wikipedia and WikiSource etc.) on Wikidata
    • as an individual item on Wikidata
    • as a detailed and qualified reference (i.e., cited book in Obama's item)
  • We're discussing items and properties, how they work and which problems may arise:
  1. Wikidata lacks precise indications for a workflow related to book descriptions.
  2. It’s a problem to use the same reference(s) to source two or more statements on an item.
  • Articles are simpler
    • Example for an article: Q555438
    • Books more complicated (there are reprints, multiple editions, collections). For books, we have basically two levels: work and edition. Work is an abstract idea (eg. Hamlet is the work). Real books are single editions of that work (all translations, versions of Hamlet).
    • Example for a book (Hamlet): Q41567

Things to do[edit]

  • WikiProject Source Metadata - Source
    • Trying to match Zotero types to Wikidata items, for the "instance of" statement. (only left column is relevant)
    • This will help Katie (aude) and Marielle (mvolz) get Citoid working on Wikidata.
    • Most of these are now done. Remaining items have some ambiguity or require a new item to added.
  • WikiProject Books
    • review all properties for work and edition
    • define a basic workflow for users (when to create the double item?)

Second half: Analysis of WikiProject Books[edit]

  1. Can we get rid of {{P|1779}}? "Possible creator" can be "author" + qualifier "uncertain" (or something like that)
    Most similar qualifer is "uncertainty corresponds to" which wants a numeric value :).But agree, uncertainty can correspond to author, artist, etc. so having a qualifier seems a better approach than having a particular property for it.
  2. Edition level: the Lua model can recall some of the properties from the "Work item" to be shown on the "edition article" (BUT we need those data to be in the work item + to have a 2.-connection between work and edition!)
  3. Also discussed how to represent multiple contributors (difficult! but there is precedence for this via people like CASRAI so we could look at that)

Second day[edit]

  • Aim: Audit existing models
  • Review of articles: start pages and end pages, however not all articles have page numbers, or numeric page numbers
  • Andrea: move all the relevant data for citations to Wikidata and develop tools that fetch that data for Wikipedia citations (like already used in infoboxes)
  • Need to make this as user-friendly as possible so as not to intimidate new users who aren't familiar with Q codes etc. Maybe give people a preview of how it will look?
  • Example syntax ideas:
    • Q codes for properties: <ref>{{ cite journal|author1=Q1234|published-in=Q1235... }}</ref>
    • No local bibliographic data, all loaded from WikiData: <ref>{{ cite journal|wikidata-id=Q1234|page=1234|accessed=May 26, 2016 }}</ref>
  • Suggestion: If you retrieve an item it should automatically populate the date retrieved.
  • Question:
    • When to auto-create an item in Wikidata for a reference? After a certain amount of uses?
  • Looked at proposal: Autofill & Ref Name Unique Identifiers - Existing Wikipedia Cite Pathway into Wikidata
  • Looked at commonalities between how different types of item are cited.
  • We looked at the document from Erika (BrillLyle) Dropbox which tries to map the identifiers across the different items that are cited.
  • Suggestion: Lambert suggested using HathiTrust (Q3128305) as another place to get full-text.
  • Inventarie
    • Lydia: inventaire.io which pulls data from Wikidata
    • Special:AboutTopic - search for topic (useful for smaller Wikipedias, e.g., Esperanto)
  • Property 'cites'
    • Property 'cites' added (P2860) followed by a request for deletion from Jura - Tobias (Tobias1984) noted we have fast-tracked proposals in the past at conferences.
    • Added cites property (one value) to Q21032780 "Homo naledi, a new species of the genus Homo from the Dinaledi Chamber, South Africa" (first usage?)
  • Edition?
    • Should it be added to the work or the edition of the book?
    • Connect the edition to the work, and the property 'cites' should link to the manifestation then links to the work.
  • Example: OCLC control number (P243) as property to work label for Pride and Prejudice should be at the edition level.
  • SPAQRL used, found 159 cases (so is a manageable task to clean up).
  • Talked about equivilant on Wikipedia infobox, has OCLC when it shouldn't.
  • Discussed challenges of educating editors.
  • Naming conventions
    • Challenge of finding an operative way to talk about a work. Markus' suggestion, intellectual unit.
    • Suggestion (Merrilee): If you're writing about something you're writing at the work level, if you're citing something it's at the edition level. Erika (BrillLyle) said edition can be confusing for non-library communities.
    • Suggestion: How to guide for people to work on? Aim to create a definition tree for users.
  • If it's a new work use 'based on'. Issue of work versus edition is tough - may need to form a separate group to work on this (and how this works between Wikipedia and Wikidata)
  • More on edition
    • Find an edition and decide what kind of properties we want for an edition: looked at Alice in Wonderland (Q20872991)
    • Language of work on name property confusing as it's trying to serve multiple functions (P407). Idea to narrow the focus of this property.
    • In doing things at the edition level we can link to Wikisource but may lose link to Wikipedia, unless an article is written specially for an edition
    • Each edition can only have 1 Wikisource entry and/or 1 Wikipedia entry
  • Non-Scholarly Sources
    • Looking at often-cited non-scholarly sources like newspaper articles etc. Ta-Nehisi Coates Wikipedia (en) as an example.
    • Found there are 6 blog posts in Wikidata.
    • News articles: Extra information added if content has been archived - archived URL, new URL retrieved on.
    • Web articles: URL, retrieved from, date linked.
    • News and web are pretty well-established and simple. Books are a mess!

See also[edit]