20140217/Reversing the Panopticon

Intro

The idea

Sensebase

Problems

Sensebase design

Next steps

[edit]

TA3M event

http://www.heterotopiastudies.com/wp-content/uploads/2012/07/large.panopticon.jpg
(Not a real panopticon since the viewer can be seen)

The panopticon is an institutional design whereby authorities can watch any subject with a minimum of effort. A reverse panopticon makes it possible to better track the actions of those in positions of authority.

David Mason will discuss a system designed to collect semantic content and link it to documents as they're browsed. For example, it can determine connections between people mentioned on a current page - those involved in politics, big business or crime. Content can be collected and annotated through automated crawling or by a user browsing their topic of interest. An investigative journalist can easily edit and annotate semantic links of people and their association as they follow a line of inquiry. General questions can be asked from the repository of all the links they visited, such as 'politician' and 'Saguenay'. The tool's flexible design and open source nature could allow anyone to participate in creating annotators and add meaning, connections to content. The system can support individual or federated instances.

[edit]

google "reversing the panopticon"

About 32,800 results

Today's version of the reverse panopticon is a Web based system for organizing data behind content.

• hypothes.is

Supporting content

• Corporate registries - Sunlight foundation
• Wikipedia, Wikidata and other reusable content

What other implementations do people know?

[edit]

https://github.com/vid/SenseBase MIT license

• An annotating system featuring a teams of computer and humans
• Designed for science research (Proxiris) and health systems (PatientSense)
• Amenable to be distributed and support personal databases

[edit]

Problems it can solve

• Augments publicly available data as you browse across the web
  • Research tool
    • Health systems
    • Science
    • Journalism

• less sinister, more constructive
• Link useful information
• Ask questions
  • How many articles about Cuba mentioned politicians
  • When did an Australian prime minister mention aboriginals

Problems it can create

• Bad information
  • Especially in computer systems (80% accuracy at best)
• Signal to noise
• Sensationalism vs subtle details
• Favouring the favoured

[edit]

• Semantic Web / Linked Data principles so any content can be described
  • Page level: Quote, ValueQuote
  • Document level: Category, Value

• Workflow
  • Many automated annotations to few human-vetted

• ElasticSearch
• NodeJS
• Bayueux (Faye)

Elements

Dashboard

• Drag and drop content
• Triaging process for annotations
• Advanced searches
  • Facet search
• Select annotators, status display
• Manage team
• Chat

Annotators

• AFINN sentiment
• DBPedia Spotlight
• Classifier
• Genozymes pipeline (Proxiris)
• Structural

Open Annotation

• Exchange annotations between bases

Text oriented data store

• supports fuzzy search, ranges, "more like this," resilient and scalable
• Team of specialized distributed software agents that are loosely coupled
  • Uses pubsub
• Triage process - validated and unvalidated

Proxy approach, indexes content as you browse

• Provides an in-page annotation and insight tool
• Re-annotate content based on new information

Scraper

• Start from current link
• re-run as required
• Continue scrape based on relevance
• Browser based

[edit]

• Annotators it's easy and useful to add
• Connecting projects
• Developers
• Front end (Semantic UI), back end (NodeJS)
• Applications