Pre-process table columns to find the views and configurations that might be intersting e.g. showing the best correlation, showing items with some separations, etc.
Categorical data can be visualized on a treemap by assigning each category to a level in the tree. For example, a health survey can categorize people by sex and native language. This can be represented in a treemap by first splitting the data by sex (masculine, feminine) and then, for each sex, by language.
When several categorical data are available, users may want to choose the order for splitting (sex first, then language or the contrary).
This project follows-up on the project started last year called Dynamic Visualization of Categorical Data.
Proposed by Doug Oard and Ben Shneiderman
Assessment of information in context:
Archival appraisal to support retention decisions (if storage is limited)
Pre-release review and redaction (general, or specific to an intended use)
Access to information in context:
Creation of Encoded Archival Description (EAD)
Coupling content-based and context-based exploration
Display of information in context
Exploration of implicit information: Social network analysisENABLING TECHNOLOGY
Multiple standard file formats (e.g., RFC 822, RMAIL, Outlook, Lotus Notes, PROFS)
Display capabilities for a broad range of attachments
By user (optionally, by email address)
By group of correspondents (as a polyhierarchy)
By thread (recognized by subject, included text, reply-to/forward)
With Web content (through embedded URL's)
Time-oriented (e.g., Kim and Shin, 2000) vs. Relationship-oriented
Perspective-based (owner, correspondent A, correspondent B, ...) vs. global
Named entity recognition
Spelling correction (i.e., matching insensitive to common errors)
Management of access rights
Computer-mediated communications (USENET, email)
Computer forensic analysis (email)
Electronic records archives (presidential email)
http://www.nara.gov/nhprc/electron.html - grant proposals due October 1 "The NHPRC will enable the nation's archivists, records managers, and documentary editors to overcome the obstacles and take advantage of the opportunities posed by electronic technologies by continuing to provide leadership in funding research-and-development on appraising, preserving, disseminating, and providing access to important documentary sources in electronic form."
E876.W48 1995 - an example email collection:
3.5 inch disk, Regan/Bush white house email released under FOIA. Held by the Maryland Libraries (UMCP:HBKRES)
http://www.dgp.toronto.edu/~sasha/research.html - a researcher:
"Memory issues in time-based email organization". Presentation at the History Keeping in Computer Applications Workshop, University of Maryland at College Park (1999).
Visualizing networks is very difficult and no technique invented so far really scale. This project consists in transforming a network into a tree and adding special visualization techniques and interaction to visualize the graph structure superimposed on the tree structure.
To transform a graph into a tree, two techniques can be used. There may already be a tree structure naturally extractible from the graph. This is the case for web sites which are organized as a hieararchy in a file system. The graph can be displayed as a hierarchy using treemaps or node-link diagrams and non-hiearchical links can be overlaid either when the mouse moves over the nodes or using dynamic queries techniques.
For graph that don't have a canonical tree under them, a root has to be chosen and a tree can be consructed by pulling the nodes from the articial root, constructing a tree from the topology of the graph, avoiding loops. This tree can then also be visualized using treemaps or node-link diagrams.
Interactive techniques should be experimented to best show and explore the graph structure.
Some trees evolve: a file system, a web site, a software development tree, a organisation chart etc. This project consists in Visualizing this evolution and comparing trees over time using treemaps for node-link diagrams.
Three techniques can be used:
multiplexing in time,
multiplexing in space or
Multiplexing in time consists in switching between different moments of the evolution and showing them all at the same place. Transitions can be animated to best track where nodes go, if node appear or disappear of if some visual attribute change.
Multiplexing in space consists in showing several moments at the same time by splitting the view.
Superposition can use different colors to show old and new nodes over time.
The project can study each of these techniques of choose a subset of a specific task and compare them (tracking web logs for instance).