User:Sj/AHS

Building out Annotated Hub of Science

• Focus on the global aggregate + mission

Potential tools: Codex!

Quick ideas

• Rapid grant idea: Develop design + workflow for offset-annotation tool for Wikipedia pages that interfaces w/ codex + hyp annotations
  • Gather existing ideas for relatedwidgets
  • Format: compare VicuñaUploader + WikiJournal.

Todo: Write up a use case; short version for a proposal. Engage interested annotator.

Codexian approach: make use of Wikidata. Select entities, link that. Clone a lot of entities and link them to WD entries.

Browser plugin that works via WD (or a broader wikibase?), not its own db. Actual implementation would be outside the scope

Build on Hyp to link text ranges to WD entities

• ^^^ when do you want to do this without direct links? In what context do you want this to be visible?
• creating a set of entities per page that aren't all yet merged into a broader namespace.
• building a broader Wikibase : a) support this in wikiland + b) merge it with the same in a local-graph. united by the mantra to annotate all the things!

• Translation idea: Describe how a C approach could help w/ translation: bundles of entities; entity extraction (+ mapping to wikidata)
  • Step towards multilingual wikidata; usage cases for lexemes + senses for a better global dictionary

Longer term

• WS project: Apply this to Wikisource
  • Use just the entity extractor; where to the entities go?
  • For annotations and links: where do the annotations go?
  • Links: Browser plugin w/ autocomplete for links in {current page in chapter in work, other context + regional context + local namespaces + WD}.

    Different from the [global KG] approach of canonical names and URIs : use every level of ostension

    • Ties to the Perseus project + Scaife?

• Hub of Science proposal: Apply this to scholarly articles
  • Build an interface for annotating + linking an article or doc to related concepts.
  • Make this useful + legible even where the source doc is not legible (wrong language, eroded by time, hidden by (c) )
• Todo: What is needed to make this work?
  • Moving beyond entity refs to entity modeling: See a model of the entities and their relations.
  • Models derived directly from text + annotations, not external knowledge. Axioms, claims, processes, data.
  • Combine this relational info (in the text) w/ external relations (from WD), make use of both local + regional [ontology/functions/connections]

• Consider a claim graph, with axioms + claims + interim steps w/in the paper.
  1. Identify atomic claims in the work
  2. Identify atomic sources -- and a mix of arguments a) in the paper, b) in cited sources.
  3. Try chaining claims/cites -- a) refine the source and target of a citation as precisely as possible. b) support clearly listing all cites [visual limitations]
  4. Project this forward: get notifications if anyone cites this claim // disproves this claim

A: Importing from PDF, or annotating PDF directly. What's the best representation of this?

Datasource could be a glossary attached to that doc.

B: Work through an example w/ plaintext of articles

Potential tools: Disambiguation

• "Who is 'We'?"