Dogear: Social Bookmarking in the Enterprise

David Millen presented dogear, a social bookmarking project at IBM Research.

Social bookmarking was pioneered by Joshua Schachter at del.icio.us
central storage for bookmarks
organize collection of bookmarks at one place

social tags characterize the content
"pivot browse" around a collection
allow subscriptions to tags (topics) or people

traditional social bookmarking sites allow people to bookmark things behind a firewall but doing so poses a security risk for companies; others outside of organization can't see what has been bookmarked

dogear supports Internet and Intranet bookmarks
authenticated identity (appropriate for large enterprise)
support for private and shared bookmarks
REST architectural style, dogear api

live demo of dogear - plugin to Firefox
dogear automatically grabs title, selected text for description, url, user adds tags (autocomplete feature), shows user the number of other people who have bookmarked it, checkbox for private bookmarks (default is shared to encourage sharing)

dogear also offers an "all people" view to see what is popular
can bookmark Lotus Notes databases

IBM Intranet search - offers dogear and Intranet search results

REST architectural style - controlled vocabulary of the URL
Bookmarks can be syndicated into other formats (e.g. blog)
dogear is designed for mashups

Dogear field trial results -
about 56% are shared bookmarks of content that is Internet
about 38% are shared bookmarks of IBM Intranet
very small percentage of private bookmarking

positive feedback from early adopters in early survey results
"corporate folksonomy"
can see common tags between members of an organization

Questions:
did you make any effort to reduce number of tags by identifying similar tags?
tag recommendation was an emerging feature, is diversity of tags in an organization a good thing? is still an open question

did users continue to use their own private bookmarks (e.g. browser-based)?
they don't have actual numbers on that, some people are interested in taking dogear bookmarks outside of the enterprise
some people have completely stopped using the browser-based bookmarks

how is dogear used within project teams?
using tagging in the context of activity spaces

Paper - Increasing User Decision Accuracy Using Suggestions

Pearl Pu presented a new recommendation technology called suggestions

recommender systems help solve information overload problem
social recommenders - predict interest from social context
decision recommenders - help users choose the most preferred item

high or medium stake products
social vs. decision
social requires little domain knowledge but requires effort to establish a profile

decision requires little initial effort and allows autonomous decision behavior (e.g. drive an expensive car but buy discounted wine)

the suggestion technique encourages users to offer more of their preferences
researchers assume that users will state preferences when they see attractive alternatives
look-ahead principle
Pareto optimality - simply put -- an item is optimal if it is not dominated by others in decision space

validated method with 40 real users in two groups
measured decision accuracy vs. decision effort

Paper - Co-Authoring with Structured Annotations

Qixing Zheng gave the presentation

co-authoring workflows include lots of annotations, insertions, deletions, and comments

hard to tell which groups of annotations are important or related to one another

users create "meta-comments" that are disassociated from the document (email), users create other artifacts (emails or documents) for these comments

common limitations of research and commercial systems - annotations are grouped only by system defined filters (status or author name); relationships between individual annotations are not captured

field investigation - found that most frequent email content about documents were summaries of edits (92%), to-do items (89%), discussions, and comments-on-comments (35%) (comments about earlier comments)

created a structured annotations model:
mandatory attributes and optional attributes
most important optional attribute is substructure
comments and edits are single annotations
meta-comments are bundled annotations
The Bundle Editor is a high fidelity prototype, she is showing a video of it right now
two pane window
document is at top, annotations are at the bottom
bundle editor has two tabs - all annotations and all bundles

evaluated bundle editor with another system
found that Bundle system is more accurate and reduced navigation time
participants also found it easier to find annotations and were more confident in their answers

questions:

is there any consideration of putting this into the open document format?
a possibility that is currently being pursued

have you done research of tracking bundles over time?
future direction is seeing how the bundles change over time (history)

Elizabeth asked how are they using bundles to support synchronous collaborative writing?
the bundles would be more spontaneous

could it work in other documents (e.g. architectural diagrams)?
yes, there are a lot of extensions - such as programming