2020-05-27 Notes and Feedback Form for today's WL3C

[Zainan Zhou (a.k.a Victor)]

Hi WikiLoop Coalition, bcc every individual participants of today,

Thank you for attending today's WikiLoop Coalition Conference Call (WL3C), here are the unrefined notes today.

Useful Link

- Page on meta.wikimedia.org

- Join our Google Groups to subscribe to future calendar invites and emails. Don't tell your friends who you may find this group relevant to.

- Provide your feedback for this session

2020-05-27~ Vandalism Patrol & Useful KG standards  

Participants: 

• James Hare

• Thad Guidry

• Sebastain Kohlmeier

• Zainan Zhou

Summary

The following topics and LoopRequest and LoopOffer were discussed:

1. Vandalism patrol + Wikiloop Battlefield

2. Useful KG standards for wikilooping

Raw Notes

Vandalism patrol 

• Wikiloop Battlefield - staying neutral while providing an interface. 

Maintaining community interest w/ a simple + universally useful tool. 

Q: why not provide ‘good faith‘ label options?    A: simple, avoid value judgement

• Community members care a lot about careful assessment.  Collecting facts, review.
  Where we can contribute: speed up that evidence-gathering.
  Goals: lower the barrier to contribute
  Can get many more labels from regular readers/users than from wiki editors - working on capturing this.  (comparison: you get billions of spam emails, many fewer vandal edits)
  
  

• Loop requests from WLB: contributions + interface feedback (link to slide w/ invitations)

• Loop offer: University of Dallas - Prof Andrews + students may be interested in collaboration + helping w/ such work
  
  

• Other models?

• Predictive models -- Is anyone training a model to predict when a future edit (next edit on a page, next edit by an article) will trigger ORES?  Related: models of the potential challenges to a current edit, based on metadata other than (author age) and (edit diff):  recent changes to the article, recentchanges for the entire wiki, other topic modeling and 

• Do trends matter to spam? Could we feed G + W trends into tools to help informhelp focus attention where there will be the most trouble?  Cf. early-onset protection (perhaps very gradually titrated, less-restrictive protection options?)

• Similar projects that interface w/ or use ORES : ?
  Q: does ORES bad-actor data track ‘repeat actors’? Possibly not. 
  Related: huggle doesn’t track good edits.
  
  

• Loop request:  Tracking across projects
  Behavioral fingerprints, textual fingerprints?  Tracking individual problems and sock-puppets across projects.
  ~ Track across namespaces and wikis
  ~ Track across different tools (compare spam-remediation - talk to blacklist maintainers, WP plugin designers) wikis, blog

• Other counter-spam approaches?   [many problems considered solve in academia don’t scale well, so simpler min-approaches are implemented.tguidry linkedin]

• MS Sharepoint spam? 

• YT misinfo spam : problem description is considered much different.  Comments: be nice.  Videos: don’t cross some bright lines (clickbait, illegal)

Useful KG standards (and categories) 

• What are we using / what do we need?
  Beyond schema.org: useful protocols, ontologies, specifications

• TG: works often on linked data protocols, and related ontologies
  How can we improve linked data implementations within Wikidtaa?
  
  

• Examples:
  ~ data shapes + constraints (ShEx), Mutexes (from FB)
          Request: Export catalog of mutexes + descriptions -- BigMama + others
    canonical functions to enforce constraints (??+)
          Request: Is there a current project doing this? 
          Look into → OKFN catalog, Linked Open Vocabs (terms),  [SJ/TG]
          Wanted: incompatible type-matrix,  meta-catalogs  [support unmerge-requests]

                    Framework for communal agreement on incompat + other type-annotation
~ canonical ways to refer to a function (code + a VM to run it?) : 
        Future: WikiLambda, other code-wikis
        Request:  Existing collections of [functions] that would make good Z-objects

~ dataset [ontologies]: update-tempo, patch-mechanism, detailed source tree?
        Request:  Good current standards for (maintainable, shared datasets)
        What standards does G!Dataset Search care about; where to find feeds of changes + submissions, how to connect bots for reading + writing  
        What other data seas* have helpful approaches to tracking inflows?

         
~ universal PIDs: concordances of IDs and authority files. (VIAF, Lens ID, …)
        compare the fragility of url-shorteners and the IA work to preserve them. PIDs are basically url shorteners... 

• other KG standards - which [W3C, IEEE] ones to attend to?  (Wikidata is often an effective substitute for more brittle + specialized standards, how to engage those processes to combine forces?)

https://datacommons.org/ -- [Q for us: connected w/ the Open Data Commons?]
    To what extent does this offer good examples of [dataset ontologies + PID-sets]?

Some metadata fields for Datasets are within here: https://schema.org/Dataset 
    Idea: Typeset a process/bot to use/process data from specific sources.
    Idea: Define a ShEx? shape from a subset of Dataset fields that provides the needed affordances  [ draft some rules ]  → Start by drafting rules as comments?
Orig Proposal:  http://blog.schema.org/2012/07/describing-datasets-with-schemaorg.html

Related:  https://arxiv.org/abs/1903.03096

[aside: Leigh Dodds @ODI on data trusts + related norms]

Victor,

Google Search