2020-05-13 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

2020-05-13 Adding Citations

Speakers

• Semantic Scholar: Sebastian Kohlmeier

• Citation Hunt: Guilherme Gonçalves<ggonc...@google.com>

• Citation related work in Wikidata.org / WMDE by Lydia Pintscher

Self-Introduction: 

• Aaron Halfaker: WMF Principle Research Scientist, AI team for detecting vandalism, quality, topic modeling.  Studies quality control dynamics in Wikis like Wikipedia. 

• Elan: Google, WikiLoop Team, WikiLoop Game

• James Hare: working on Wikibase for COVID19 research. Documenting works on Wikidata. / Citation on Wikipedia

• Maria: Google Open Source Team / WikiLoop communication lead.  Interested in citations.  Makes open information reliable for users. 

• Thad: Big data, different sort of citation, cite / referenceable datapoint, most interested in machine-readable cite formats

• Vinay: Google Brain, help create a tool to assist WP editors create stub articles.

• Lydia: Product Management for WD at WMDE.

• Guilherme Goncalves: SRE at Google, Citation Hunt author. Think of Citation as a good way for micro-contribution

• Sebastian: Senior PM of Allen Institute for AI

• SJ Klein: federated data, Wikipedia + Knowledge Futures group 

Citation Graph by Allen Institute for AI, Sebastian Kohlmeier, Link to Slides

• Allen Institute for AI: contribute to humanity through AI...

• Semantic Scholar: science and technology academic article search engine.

• COVID 19 research repository "CORD-19". parsed and extracted large corpus of related academic articles

• CitationGraph

• 186M+ Papers, 1B+ Citations, expand coverage with Microsoft Academic Graph and publishers

• Citation Classification: background citation, result citation, method citation.

• Citation Alert

• CitationGraph + Wikipedia Collaboration

• Citation Template Integration (S2CID) / Semantic Scholar Citation ID

• Initial work with S2 Author IDs in WIkidata

• Integration with Citation Bot

• LoopRequest: Want to be able to recommend citations for WIkidata, using similar text.

Citation Hunt by Guilherme Goncalves

• https://citationhunt.toolforge.org

• Random browsing

• Browse by category

• Internationalization

• Leaderboard

• Usage

• #1Lib1Ref: one librarian, one reference. link

• Training materials, physical events, tracking.

• Now on its 5th year, multiple languages, twice a year 3-week long.

• Starting again next week!

• https://twitter.com/WikiLibrary/status/1227521465333010433

• The Wikipedia Library/1Lib1Ref/Lessons/2018 - Meta

• Wiki Education

• Various editathons, smaller campaigns, organic traffic

• Wikimedians in residence help with this!

• Links throughout Wikipedia itself (e.g. Wikipedia:Contributing_to_Wikipedia)

• https://citationhunt.toolforge.org/en/stats.html?days=30

• Documentation

• https://meta.wikimedia.org/wiki/Citation_Hunt

• https://github.com/eggpi/citationhunt

• Latest development:

• Custom Citation Hunt (DEMO)

• Citation Detective announcement email

• Citation Detective WikiWorkshop article

• Next steps:

• More languages and campaigns

• Integration with other tools, e.g. SuggestBot

• UX review, especially mobile

• More types of backlog, e.g. OABot

• Recommending citations :)

Wikidata and Citation (Lydia) Link to Slides

• Wikidata grows and needs automation!

• Automation related to citation

• Hoped for cleaner data, but not always so.

• We can do statistics on this when we have enough instances of a value (country demographics).  But things like author names may only appear once or twice: pretty poor.

• Prototype: generated a bunch of refs, got feedback from the community

• Next: Wikidata game instances for this; early result: +300 refs this way [how does this compare to 1lib1ref in terms of participation, time-per-cite?]

• Want more input on how to make this more effective!

• Ideas: 

• 1) schema.org: look at Type, to infer meaningful/high quality cites

• 2) models that could be trained: 
  A: how suspicious is a cite?  A1: How important is it to find a source for this, if it will be kept?  A2: where a source exists, how important is a doublecheck that the source makes the claim it’s associated with?
  B: how trustworthy is a claim [cited or not]?  Context includes the age of the account that posted it, the amount of context for the cite, the # of other claims for that entity, the amount of visibility/traffic the entity-page and the property get, the trustworthiness of any source, as per A above

• 3) connect w/ 1Lib1Ref and CitationHunt 

• 4) look across the web for pages w/ text that makes a similar claim (entity E w/ value V for property P: look for {E,V,P} in proximity anywhere in a) SemSchol, b) CommonCrawl, c) Google)

LoopRequest and LoopOffer

• AI2 offers citation recommendation candidates, Citation Hunt to suggest (long term) Guilherme + Sebastain

• Citation Hunt to support Wikidata, Guilherme + Elan + Lydia

• "Special” LoopOffer: Thad is there to help (and for free!) with respect to schema.org

• WikiData make use of 1Lib1Ref  Guilherme + Lydia

• Late request: intern to help w/ WP-on-IPFS , SJ + Santhosh

• Thad: can we use "Google Crowdsource" for abbreviations, other properties?
  → can highlight schema.org props that are often abbreviated (not canonical strings)
  → could help lydia  ]

Interested Guests to invite:  Thad, Daniel

Victor,

Google Search