Skip to main content

Understander: A Science Annotator

Published onMay 15, 2020
Understander: A Science Annotator

You're viewing an older Release (#2) of this Pub.

  • This Release (#2) was created on Jul 14, 2020 ()
  • The latest Release (#3) was created on Feb 12, 2021 ().

An interface for an ecosystem of shared research papers, commentary, and concepts.
In acrobatics, an understander is the giant on whose shoulders others stand

Look up a paper or concept: see its latest incarnation (a la Semantic Scholar), perhaps its progression, and what people are saying about it. Add comments, for private use or to share.

Underlay view

  • A set of [hashes] Dn of related documents (versions), tagged w/ version type

  • A set of notes Nn, tagged w/ note-type (comment, link, review)

    • These may link to (part of) a source, or may connect two documents.

  • A set of annotators An, tagged w/ role (reviewer, journalist)

  • A set of annotator-groups Gn (badges, associations, guilds)

  • A time-range T0-n of creation dates

A search for a document should turn up a cluster of documents.
Associated with them is a set of revisions and annotations, some of which are identified with their authors, some of which have known roles or are part of known editing/reviewing groups.
This interaction should be filterable by a config of D, N, A, G or T, and perhaps a choice of sort order for many of those dimensions


Intermediaries can add context in a few ways.

Document equivalence class: Associate hashes of similar documents with one another (a la fatcat). if you find what looks like a revision or another instance of a work, link the hashes for a more useful equivalence class

Document revision tree: explicitly indicate the (stated, inferred) branch history of a doc

Timestamping: estimating creation times where ambiguous helps understand revisions over time.

Note classification: tag notes by type: comment, analysis, link, citation, original-source. Tag also by semantic meaning (e.g. for citations themselves): positive, neative, informative, rebutting, superceding.

Person classification: identyfing capabilities, qualities, and network-context of people (and other sources of notes) helps viewers prioritize, filter, and display their comments.

Annotator grouping: whether a review team, guild, parodist, authenticator, or [other] badge provider: naming groups and their members, and the type of badging they do, helps filter and display their notes

Overlay view

A reader or annotator, presenting the above, might have a configuration including

  • The default DNAGT filters + priority order

  • A primary document version to view (e.g. latest/most complete)

  • How to render a composite doc (e.g., primary doc + diffs from other v.s)

  • How to display notes (e.g., heatmap highlighting, margin bullets)

  • How to handle creation of new notes by the reader (private or public)

You should be able to browse annotations of a document even if you can’t see the whole doc yourself. Both via available comments, and including Google’s and Semantic Scholar’s approach to sharing contextual snippets around whatever you search for.

Bonus: extension to a discovery + publishing hub

Extending this to an annotated hub of science could readily include:
+ searching directly by term or DOI,
+ viewing the results in the reader,
+ publishing new works: getting an identifier and feedback on initial annotations that could be made to make it more accessible and useful.

No comments here
Why not start the discussion?