XHTML Documents with Inline, Policy-Aware Provenance

Harvey Jones
Decentralized Information Group
MIT Computer Science and Artificial Intelligence Laboratory
Advisor: Daniel J. Weitzner

Copyright Problems on the Web

• Copying has outpaced control
• Licensing data is often nonexistent
• When it does exist, it often gets lost in the creation of new documents.
• Serious copyright problems exist on Wikipedia, etc. because of this
• Centralized control is inadequate for decentralized systems
• Creative Commons - Start of a solution?

Use Case

• Professor Doe composes set of lecture slides on cameras.
• Creates original content
• Copies several Creative Commons images from Flickr
• Copies several slides from similar presentation by colleague
• Originally for academic use
• Two years later, Professor Doe wants to publish the slides in a book
• Professor asserted fair use originally, now he needs to track down rights for all presentation content.

Solution - Inline Provenance Metadata

• Embedding Provenance Metadata
  • Use RDFa as a standard way of inlining metadata
• Policy-Aware Clipboard
  • Preserve metadata when creating composite documents
• Reason over this data to create policy aids
  • For example, enumerate acceptable Creative Commons licenses

Representing Provenance in RDFa

• DocumentFragment describes a quotation
• The DOM node the metadata is attached to is the content of the quotation

Demo

• Combining CC-BY and CC-BY-NC
• Combining CC-BY-SA and CC-BY-NC

Nesting

• Nested tree represents a data flow

Copy

• Start at deepest node that contains entire selection
• Go up through parents, copy each annotation that applies
• Go down tree, slice away all unselected content
• Add new layer of metadata representing this copy

Copy Example 1

Copy Example 2

Creative Commons Reasoning

• Start with CC-BY (most permissive license)
• For each license:
• Add restrictions (Make transition)
• If no transition possible, no license is allowed

Contributions

• A document format with inline provenance metadata.
• A method of excerpting from these documents that preserves provenance auditability.
• A Creative Commons reasoning engine.
• A tool that uses these components to reason over documents and recommend licenses.

Acknowledgments and More Information

• Thanks to Daniel Weitzner and the TAMI team, especially Hal Abelson, Tim Berners-Lee, Sandro Hawke, Chris Hanson, and Gerry Sussman.
• Demo photos taken from Howard N2GOT, _gee_, AussieGall, Drab Makyo, and Dusso Janladde.
• More Information
  • Thesis
  • Links to bookmarklets and sample documents
  • Decentralized Information Group
  • This work is supported under the Transparent, Accountable Datamining Initiative funded by the US National Science Foundation Cybertrust (05-518) program.