IW Meeting 2012-02-16

From Inference Web

Jump to: navigation, search

Contents

Meeting Information

Agenda

  • reminder - new format - one technical focus and short around the room updates.
  • aim for meeting length to be an hour.
  • James will present some strategies for indicator explanation with drill down.
  • volunteer for focus topic for next week.  (deborah is likely offline - will be determined in a few days).
  • around the room  - brief status on current work. 
  • please post your updates in advance on the wiki page (reminder paulo can not get to the titanpad page)
  • deborah requests this starts with cynthia with an update on any issues related to access behind login prompt

Discussion

Around the Room

  • Jim Mc: Finally integrated some of PROV into MAGETAB2RDF in preparation for CSHALS.
  • Tim: 

FUSE

Use Case Working Group

  • Design of sample data for providing an answer to Challenge Question 3
    • Two funding indicators, backed by feature data provided by NYU
    • A CADRE reasoning trace, designed to link hypothesis for answer to these funding indicators
  • Using these materials, we have developed corresponding storyboarding for our explanation component
  • Two modes being considered, based on determined needs for explanation component:
    • Abstracted:
      • Expose assertions of triggered hypotheses, paired with assertions of supporting indicators.
      • Here, supporting indicators will be presented through visualizations of supporting feature data.
    • Debugging: 
      • Exposes same information available in abstracted view.
      • Additionally, exposes reasoning trace(s) for hypothesis generation, based on presence of calculated indicators
    • Open issue: determining how much detail of indicator calculation to expose in abstracted view.  This will likely vary between indicators.

Presentation Strategies for Indicators

  • Based on supporting feature data + calculation steps
  • Significant variability in expected structuring for each kind of feature data.  Examples include:
    • Networks (citation, co-authorship)
    • Numerical trends (increases in number of (publishing authors, funding agencies))
    • Text snippets (highlighting language sentiment)
  • As indicators are made accessible via explanation, appropriate visualization widgets should be called, based on kind of feature data used.
  • To provide this functionality, we will need presentation classifications attached to each set of feature data.
  • Will need to work with collaborators populating the SEKB to establish this

Supporting Technologies for Explanation Component

  • Many tools currently used by RPI for explanation/data visualization are web-based
    • Data visualization:  Google Visualization API, D3.JS
    • Provenance (PML) browsing: IWBrowser
  • RPI currently determining feasibility of reconfiguring these existing tools to work within the FUSENET intranet.

Demos using material on our secure server

  • Currently, Cynthia working on modifications to IWBrowser, to access PML stored in our WebDAV server
  • Will allow us to use IWBrowser in demos to FUSE collaborators

actions

Jim McC todo - send email with the listing of prov terms integrated and how those map to pml


todo James  - please make a list of questions to ask our team if we think - a) we can get - b) what we want to obtain - c) discussion w/ Tim on encoding indicator/feature data in datacube format - will allow us to leverage existing visualization technologies.


for example on slide 21 of the ppt presentation, we show an increase in the types of funding agencies - i expect we want to be able to provide a drill down to get the list or at least the beginning of the list if the list is giant.  what would it take to get that?

todo - tim and james come up with an email suggesting that bae use the data cube vocabulary - http://publishing-statistical-data.googlecode.com/svn/trunk/specs/src/main/html/cube.html  trying to make it as easy as we can for bae. this might help us get some quick, reusable interfaces ala jim mcc's popscigrid demo

todo - cynthia - schedule a call with paulo, deborah, cynthia on the issues jiao brought (cc inference web) also send an email first with the questions and cynthia's provisional answers - inference-web

todo: tim to propose changing time to 3pm on thursdays. ok with deborah, james, and cynthia

todo: tim to send deborah draft at 3pm, call at 5pm to discuss

next week's topic: data cube

Facts about IW Meeting 2012-02-16RDF feed
Date16 February 2012  +
Personal tools
Navigation