Federated Search implimentation case study

MetaSearch-Two Clicks to full text
Judi Briden
University of Rochester River Campus Libraries


Usability Studies indicated that search was Google for the students
On Academic library sites – users face serial failure
Can’t find articles
Make assumptions the librarly fails to match
Seldom ask for help
Librarians just know about books

Follow a UCD design process
Very collaborative
Multiple iterations before and after release
Ethnographic research

Did studies with students describing how they research
Did video studies of dorms
Did disposable camera study with users
Had students show where they go on campus for a day
Learned a lot about the undergraduates
Planning on continuing the Ethnographic study, as it was very valuable

Goal – reduce the serial failure
Skip complicated paths to databases
Don’t make me choose before I search
Accept what I type in
Don’t expect me to read before search
Use labels I understand

Original title – Articles and More
New title – Find Articles

Started simple, added complexity slowly, based on usability and ethnography.
Final result was similar to ours, full page, radio buttons have descriptions
General search, sciences, social science, arts and humanities

Search takes 10- 30 seconds to come back – big issue
Tabbed basic search and advanced search
Highlighting the terms was very important
Added part of abstract
List source
More>>> full abstract, subject terms, native database

Title link automagically brings back full text, PDF, etc. – if no full text available, moves to catalog with location- looking to do a map, but no time yet. If not in catalog, go to interlibrary loan.

Advanced search for people who know about particular databases. Still one search box. Advanced search shows number of hits per page.

Set up searches for each course, with specific databases to be searched for each course. This is done at the enterprise search level. I wonder if we could do something similar for service line, or communities, at the enterprise search level, rather than the community level? Maybe not. . .


Looking to add a Google scholar tab in find articles
Take advantage of customizations

jbriden@library.rochester.edu

User Perceptions of Federated Search

User Perception of MetaLib Combined Search
Rong Tang
Simmons College

Study
Investigate user perception of Metalib combined search in the environment of Washington Research Library Consortium

Methodology – user a questionnaire and screen shots. Respondents answered questions related to the search process and drew a sketch ti illustrate their understanding of system operation.

Objectives
Understand users perceptions of Metalib
Compare librarians and students perceptions of federated searching process
Identify areas of confusion and explore the implications for delivering information literacy program.

Uses of Metalib combined search
Students used it for full text
Librarians used it for additional search after a fruitless search
Some use it for quick search

Useful but hard to figure out: 69% of Students
Not useful & hard to figure out: Library ~40%
Only 27% of the students mentioned the presence of “Find It” button, even thought 85% of them indicated that MCS is used to locate full text.

Librarians used technical terms to explain the process
Students were more interested in obtaining full text through federated searching. Librarians paid more attention to retrieval performance, search strategies and relevance ranking.

If we are going to do federated searching, we should do a study of our own. This does not look to be generalizable.

Students
Learn background
System operation
Search strategies
Practice

Librarians
Teach
Background

Design Implications
Make explicit the retrieval and ranking procedures
Search results are produced in batches
Make more results option available on each display screen
MCS relevance ranking is not based on the entire combined results
Provide multiple search support mechanisms
Findit button design
Implies full text is available, needs to be only available when it will work
Cluster search results and use visualization tools to facilitate use

The role of search tools in information seeking by users

The role of search tools in information seeking by users
Karen Groves
Metalib product manager

Information vendor representative

Overview
ExLibris
What is metasearching?
Metasearching and the user experience
Discovery to delivery

ExLibris - global information provider, more than just the US.
Products - Aleph, SFX, Metalib, Digitool, Verde, Primo

Customers are mainly universities

How do I find the information I need?
By hitting all of the search systemsseparatelyy. . ..
Metasearch gives a common interface to a broad range of resources

Resources that are searched with many metasearch applications
Library Catalog
Collections Databases EJournals
Archives
subjectt Gateway
Newspapers
Search Engines
Digital repositories
Etc

Use Facets in metasearch, combine metadata and available sources.

Metasearch
What do you need? Building search interface
Configuring connectors to search and retrieve

Use accessprotocolss to do federated search
Non-http
Z39.50
http
xml

Send unstructured protocols via screen scraping or HTML parsing
Basically, advantage is that the interface removes the complexity for theend userr and allows one entry point and one search syntax

Product comes with a set of preconfigured connectors into various sources
Active in standards community

Why does this matter?
People want to find information when they need it, where they are.

Users gravitate to easy to use vehicles
What is FRBR?

Possible to offer everything today
Call the process Find it + Get it
Supporting a Knowledge creation cycle
Discovery -> analysis -> publication -> reuse

Study on Tagging

Patterns and inconsistencies in collaborative tagging systems, and examination of tagging practices

Examination of social bookmarking systems compared with classic taxonomy.

Definition of terms:
Tagging – act of generating a dynamic taxonomy or folksonomy.
Tagging – alternative format Tag Cloud
Post is the measure in study – consists of actual tagging of a document.

Citeulike and Del.io.us
Citeulike - Academic use – more often used by

Controversy

Tagging is Good: dynamiuc distributed classification, related tag networks, tag clouds show extent of collection, user terminology, diversity, consensus by active users

Tagging is Bad: mob indexing, no controlled vocabulary, poor browsing experience, no thesaurus, consensus by a mob is no consensus

Research Questions
What patterns of tagging emerged? Frequency and co-word analysis.
What extent do these enhance the traditional methods?
What extent do they defy the traditional methods?

Study protocol
Used 64 URL’s on Del.icio.us

Analysis
58,728 posts
165,831 tags
Number of unique tags 18,904
Per URL max 1252, min 23
Average posts per url 917
Max 5172, min 53
Average tags per URL 295
Max 13809 min 49

Interesting to see that even with 49 tags, consensus started to form.
6% do not tag
Users use 1-3 tags 65%

High Frequency Tags
Productivity
Gtd
Lifehacks
Tools
Web2.0
Web
Organization
Css
Blog
Wiki

Issues with tags

Spelling variations occurred

A number of synonyms or related terms

Acronyms

Cotag Graphs
Looking at where tags are clustered together to determine when things belong together.
They did not see things clustering together well.

Non Subject Tags
Affective Tags
Cool – 906 occurrences
Time and Task Tags
Toread: 939 occurrences
3049 unique tags identified as time and task 16%

Intrinsically time sensitive
Express response for user
Users were not consistent in applying tags, or in the spelling of tags, etc.
Users demand further finer grained indexing that taxonomists usually apply

eMail Management

eMail Management

People see email as being a temporary unimportant content form.
People typically are unaware of what records retention methods they should do.

Three issues in archive
1- reliable storage
2- comprehensive metadata schema
3 - missed the third

People with no folders were the most happy. None of them reported problems finding older email.
All of the data on this is available at the following website:

http://ils.unc.edu/digitaldesktop
megan@ischool.utexas.edu

Taxonomy Standards and Location

Taxonomy building, dealing with Location
Ron Daniel
rdaniel@taxonomystrategies.com
Taxonomy Strategies

Helped define the Dublin Core

8 common taxonomy facets
Organization
Content Type
Industry
Locations
Function
Topics
Audience
Products and Services

Don’t make more lists, relate the data sets and map into the same taxonomy. Use fewer lists.
People use location lists all the time, and they often mean different things. This makes it hard to interrelate things, since they have different reasons for calling the different things by the same name.
Great slide on what makes a good taxonomy – look for it when the deck is published

Using ISO 3166 for location
Breaks into Countries Sub-regions Changes
Source is the UN statistics Division
They have a standard for unassigned regions

Automatic Classification: Some facets are easier to auto classify than others. Entity classification goes fairly well. Subject, less so.

Website Taxonomy

Website Taxonomy
Kathryn La Barre

Interested in faceted classification. But did not see it as useful in a library situation.

Webopedia web site taxonomy definition

Organization and access for web content for browsing, navigation, search

Faceted Classification discussion list?

People use facets on 24/200 sites surveyed
Used in basic search, browse, navigation only, advanced search, integrated search and nav.

Common facets were
Who,what, where, place

US Department of labor given as an example.

Facets carry throughout the site.

By topic, by organization, by location – principles of orders within facets, each facet has it’s own logic – not always alpha, sometimes by importance.

Took each example, showed graphically, then broke down into a wireframe. Nicely showed her point.

Cranfield studies showed that facets can be rapidly overwhelming.

Principles for practice: Theory and Practice

Are there a list of fundamental facets - Platonic ideal facets? Developing a list of facets as a possible list of fundamental facets.

There are existing metadata standards that exist, how much do facets map to those standards?
Most sites have backend databases, and the facets are based on that metadata usually.
How do we reuse legacy metadata?
Looking to coordinate this with OWL and RDF / Semantic web.

http://leep.lis.uiuc.edu/publish/klabarre/overview.html

Posters

One interesting area I have not been covering so far are the posters. In the hall there are multiple poster presentations, which change every 4 hours. This makes it hard to take notes directly into the PC as I have been doing with the sessions. In between sessions I am going to try and put up some comments on the posters I have seen.

SEARCH
A gentleman from Rutgers University has a big poster on search. He investigated several different methods of improving search, ranging from asking the user to enter more data via a series of boxes to automated methods. His research showed that, with his set of data, incorporating a weighting based on the top 30 documents for a search term - based on expert opinion of the top documents significantly improved the relevant search. I think we could incorporate this once we have our logs. We could look at the top queries and start modifying when those queries are run. He had another interesting idea, in which he used the index of the persons desktop, gathered via google desktop search, into the relevancy ranking. This was not as good as the expert evaluation, but did improve relevancy quite a bit.

WIKIPEDIA
Interesting review of the research around wikipedia and the available areas for further research at http://en.wikipedia.org/wiki/user:phoebe_ayers

CREDIBILITY
Interesting poster that showed that peoples use of sources is not based on the source itself, but how helpful they found the source.

US Government - management of people

James Dobbins
US Government – working with Army, Navy, etc.

Established a set of categories to determine CSF for projects – can I use this on the Usability projects? Get slide deck and see if it makes sense – or are our projects too short for an interview method to make sense?

Speaks to the need to ensure the people involved are suited to working in a virtual environment.

Air Products study of search and improvement

Mary Moulton

Air Products and Chemicals – Allentown PA

Taxonomy Specialist, KM Solutions

20,000 employees

Operations in 30 contries

Continuous improvement of innovation

Improve the ideas that enter the innovation pipeline

Develop an improved search tool for technical information

Looking at Knowledge Reuse

Developed a method for measuring the ease of knowledge retrieval.

Each searcher was tasked with finding 3 search goals by any means

Test preformed in searcher’s workspace

Test proctor recorded searcher’s strategies and success in finding goal

Results

65% failure rate in computer searching

Intranet search tools not precise enough

Knowledge stored inconsistently

Numerous unconnected systems

Searcher often does not know where to start

If they did know, 3 +- 2 minutes

Role and years at company not predictors of searching section

Interviewed Technology Managers

Result

Knowledge Retrieval Important

Knowledge Retrieval effectiveness low

Solution

Restrict cope of search to technology applications

Built taxonomy and metadata repository for browse and search

Custom thesauri

Reused the AP online Directory for interface

Develop behavioral aspects around information seeking

Got high level management support for content management

Use Microsoft sharepoint search

DT search as crawler

Used multiple different tools to create the solution

Spent time weighting content

Now moving into improving content

Navigation issues – memorized path to content

Needed to overcome the negative impact of previous problems with search

Idea Management at Cargill

Idea Management at Cargill
Anne Rogers

New service within Cargill that directly supports innovation.

Are ideas knowledge that can be managed?

By managing ideas, Cargill was able to systematically review the concepts, evaluate them, get back to the employee, implement the good concepts and save money. Would not share amount saved.

Developed a structure around the idea process and sold it to everyone as the right way to generate ideas.

Feedback
Capture -> Evaluate -> Act ->

Need to be systematic, intentional, focus on the type of ideas, develop effective practices.

Defines group as doing – Just in Case, Just in Time, Embedded in daily work products.

“Don’t want to solve the business model for the buggy whip”

From imagination to return of investment

From imagination to return of investment
Matt Chapman
Matt_chapman

Creating a tacit knowledge transfer process
KM at beginning and end of the innovation pipeline

Accessing the knowledge
Time based events – small group, asks big group for help, get solution
Ask the experts – small group advertise skills, big group share problems, and get the solution.

Converting Knowledge to value
If there is no innovation pipeline, then things sort of happen – rather than building on the past, we get corporate amnesia.

Could we / should we be the enabler of service line innovation?

Ask specific questions – leads to better concepts.

Proven 927% ROI
Idea Central
Open Innovation
Innovation Arms Race
Imagine accessing the knowledge of your customers, partners, suppliers, etc and get the value from all of them?

Information Searching Tactics of Web Searchers

Information Searching Tactics of Web Searchers
Mimi Zhang
mzhang@ist.psu.edu

Another interesting paper that could help us develop and study our web logs. If used with the IE browser tool below, then this could give us a good image of our users.

Bates (1979)
A move made to further a search

Query Reformulation of excite users
32% did not submit more than one questy
When they did nodify a query
34% modified queries included the same amount of terms as the original ones
19% users reformulated their queries by adding a term
16% of them modified by removing a term

Web query reformulation model
Build up on saracevic’s stratified model
Three categories content format and resource
8 sub categories

This study
How do web searchers reformulate their queries?
What are the results of this effort?

Studied a 24 hour log from 9/9/2002

30% of the time people modified the nouns ie - subtracting nouns, noun after term noun before term.

Study really gets into the semantics of the search terms.

About 30% of time people subtract something – noun, phrase or conjunction

TC = topic change, so often times people changed the topic

This paper looks very interesting – I think we can get some interesting data from it that might be applicable to the search engine. We need to do something similar with our data, once we have it.

A Goal based classification of web infromation tasks

A goal based classification of web information tasks
Melanie Kellar – Dalhousie University

This was interesting because of the development of a customized browser for tracking user behvaior on the web. I think this tool could be used internally to study usage of the KWeb over time, specific search and navigation behavior. In conjunction with our web logs, search logs and other metrics we might be able to get a better understanding of how users interact with our tools and where the major pain points are.

Trying to define a framework for the high level tasks people perform on the web. Looking for implicit behavior

Choo
1 – informal search
2- formal search
3- undirected viewing
4 – conditioned viewing

Morrison
1-Find
2- Collect
3- explore
4- monitoring

Sellen
1-Finding
2 Information Gathering
3- browsing
4-
5- transaction

Used a custom web browser that annotated usage and allowed people to notate the purpose of the web session.

Used – fact finding, info gathering, just browsing, monitoring, other / transaction

Used a focus group to develop a new set of categories (card sort)

Looking for specific information, entertainment, transactions, information gathering, routine/hobby, monitoring

Developed a new classificiation system
Web information classification
Information seeking information exchange

Generalizability vs realism vs precision

http://www/cs/dal.ca/~melanie

Designing for uncertainty

Designing for uncertainty

1st speaker was working on a ethnographic study of two researchers and their experience with systems and uncertainty

Strategic uncertainty
Enables creativity and innovation
Working through uncertainty a mediating strategy
Knowledge generation
Information discovery
Focus formulation

University of technology Sydney

Uncertainty and understanding intermingle
Clarity about one aspect would emerge as chanllenges surface about others
There were specturms and degrees
Certainty . uncertainty
Releveance / non relevence

Power of allowing people to communicate uncertainty, share uncertainty, and enable them to work through this in some positive way.

Don’t reduce uncertainty, but allow people to work through and with it – tolerance

Uncertainty in work based settings
Jennifer Berryman

Study with public sector workers in AUS. – knowledge workers
Working with them about how they make decisions.
Very aware of their stakeholder groups, their industry and area as well as the politicians they talk to.

Work environment is complex and dynamic

Tasks are unclear
Goals ill-defined
Competing

Difficult to judge how much information they needed, what information they needed etc.

Iterative process as they sought feedback – never positive that they had enough information.
Created a lot of ambiguity in their work environment, but people were used to it and did not see this as an issue.
Uncertainty is not a barrier neither a negative experience nor a positive stimulus, part of the natural work environment.

Individual differences in the experience of uncertainty
Jannica Heinstrom

Motivation and uncertainty
Low motivation – meet difficulties with feelings of anxiety and confusion
High motivation – meet difficulties with feelings of confidence and optimism

Fast Surfing
Hasty and superficial information seeking
Common among persons with low conscientiousness and high sensitivity
Fast surfing design - easy to use and motivation with support features

If we create a positive environment, even fast surfers will be able to use the environment

Broad Scanning
Explorative and spontaneous information seeking
Common among outgoing, competitive and open persons
Broad scanning design – tend to browse, and need associative with outreach services

Deep Diving
Planned and structured information seeking
Common among conscientious and organized people

Design for uncertainty
Encourage creative thinking and innovative solutions
Build in support and incentives
Open curious and motivated persons most receptive

Uncertainty and Information Literacy activities
Sanda Erdelez

Proposed framework for information uncertainty inside information behavior.

Two types of information behavior – seeking and “bumping into it unexpectedly”

There are research models that are being used, that do not match the natural behavior of information seekers.

Super encounters – people who bump into information rather than doing them in the step wise model that the research models create.

Tools need to be developed to better match this model – people bumping into interesting information, rather than a more step wise model

Supporting uncertainty in information seeking and retrieval
Marcia Bates

Overview of the topic presented and then discussion of the need to allow for and legitimize the process of uncertainty.

Social Networks & Virtual Communities

Social Networks & Virtual Communities
Suliman Hawamdeh
University of Oklahoma
www.ickm2007.org
Vienna August 2007
2008 in Columbus along with ASIS&T, two weeks

Shared a you tube video of a professor talking about his feedback, the prof was not aware of being taped. Demonstrating some of the loss of privacy we have now.

87% youth on line
89% send email
84% search for fun
81% online gaming
75% IM
43% purchase
Girls use email and IM more often than boys
7th grade is the first big bump in online presence.

Blogs create a sense of community
Sharing of information in particular

Society is moving from real word to online world. Stronger relationship on line, breaking ties in real world.

Networks are gaining value, having a network and maintaining it has value.

Social Software – lets people rendezvous on line
Examples – RYZE, Tribe, LinkedIn, Friendster

Cross Cultural Issues on Blogging

Cross Cultural Issues on Blogging

Two studies
Blogging in personal environment
Blogging in working environment

Challenges to virtual teams - Culture
HCI has not spent much time on this
A major issue is a western bias in studies
Are there any cultural differences on trust by revealing different personal information?
Corporate blog contents and management styles

Study 1 – trust in blogger profiles
We are interested in whether bloggers from different cultures reveal similar personal information
National culture effect vs virtual community effects
In other words, would people be dominated by their normal culture or would the virtual culture overcome their home culture and create a new culture

Hypothisis – US bloggers reveal more personal information than China or Korea
Cross cultureal comparison of interface, fields, in user profile, then sampled user profile to see what was filled in – content.

See differences in basic format of the tools – the US tools were more individualistic when compared to Korean or China’s sites
The Asian sites were more group blogs, team blogs, and provided more context around them. Support more collaborative activities.

The cultural universal characteristics were:

Willingness to share
Links to other bloggers
Step by step instructions on how to create a blog
Privacy policy

User profile fields
19 in US
10 in Korea
11 in China

8 culturally universal fields
Cultural specific fields – real name and personality in Korean, law in Korean that you have to give this information
China & US - contact info, blogging site introduction
Us astrological sign, zodiac year, industry occupation, favorite movie, favorite music, favorite book, interest

US provides more fields

No cultural difference in fields of title, links, nickname

Overall, American & Korean more like then Chinese
Comparison
More likely to reveal gender and age in US
Other factors might influence online trust
Economic factors at play?


Second Study – Corporate Blogs

Why do people blog inside a company?
4 reasons for blog:
Establish expertise
Alternative media
Corporate communications
Build community

Culture effects
Power distance in CN > US
CN more managers blog
US employees with different titles blog equally
Individualist . collectivism:
CN contents focus on group . company
Us contents focus on individual or individual group.

If technology matters, then these will not exist.

Dataset – Google China and Google USA

Measure
Title, content type, update frequency

No cultural difference
Bit on blog contents
Product blogs > in US
Event = in both
CN > knowledge

Frequent updating seems very important for attracting audiences in corporation blog
Difference in product blogs might be because US has more products than China.

Personal blogs show large cultural difference, business not as much.

The Political Blogosphere

The Political Blogosphere
Scott Robertson – Drexel University

Political blogs fall into 3 types:
Quasi journalistic editorial gossip style
Official
Personal

Blogs are large – but not as a percentage of electorate
People in a study tended to look at the candidates web site, less blogs and wiki’s

His research is around political decision making and the use of logs.
Broke research out into voting, decision making, information gathering, culture and belief

Voting systems should not be designed like ballots.
Voting systems should be designed within the context of existing political systems.
Developing a Political Portal that would survive between elections – his studies show people have no interest in between elections, but this might generate this interest.

Political blogs become part of online deliberation that leads to decision making
They have a “bursty” traffic pattern
Tend to be slanted to particular types
Very partisian
Can be personal
Within ideology
Increases in-group engagement
Neutral on out-group engagement
Personal and confrontational

Would want to design portal to increase social engagement
Increase participation
Embed the voting act in a context of political discourse

Will Health Practitioners Blog?

Will Health Practitioners Blog?

Deborah Swain
dswain@nccu.edu

Pilot study to see if used
if they share explicit and tacit knowledge
if blogs fit in a public health informatics framework
if leads to medical error collection and process improvement?

Study consisted of:
Student Nurses (23 users)
One month (10 frequent users)
Learning Environment
Blog on reserved space server with security

Topics picked – support please, Challenging patients, help finding a job, hob contacts, medications, NCLEX Exam, patient rights, public, test reviews, uncategorized

Qualitative Data Analysis

Over 30 postings / week

Research Classifications of postings
Concerns and curiosity about the tool and blogging as an activity
Medical topics of interest to the bloggers
General learning and sharing of professional news and advice

Wanted to verify repeated process of information flow
Wanted to see sense of community
Became a learning organization
Temporary COP

Validation
Need to promote error reporting
Apply lessons learned
Protect privacy
Ensure integrity
Encourage new media
Validated use of blogs in health informatics
Need to establish a level of trust for sharing
Kept actual names private

Did a Social Network analysis on the data

Had a couple of active hubs in the blogsphere
Facilitator recommended
New Project with Health Practitioners
Survey, build, focus group, open house, e-home with bolgs and wikis
David Kirby Dave@kirbyimc.com

SIG - HCI meeting on Sunday

Study of dogpile compared to single search engines

# of queries per user ~ 1
# of words per query ~2-3
People don’t go beyond the first page at most into second page. Not changed over the past 10 years
People spend 1- 2 minutes per query
“Are you looking for” feature was used more often then other similar features
Label which engines retrieved which sites – not sure if this is useful or not.

Implications
1 – underlying information content has an effect on how users interact with the web information system
2 – content of web searching as significant carry over regardless of changes in access to content.

We may need to change human behavior rather than technology. No good ideas on how to improve people searching

Jim Jansen – jjansen@ist.psu.edu
Amanda Spink – ah.spink@qut.edu.au

Use of Boolean – pretty low, not significant, people not using advanced search
Web log analysis allows for starting point, need to do studies to verify the data. Book coming out on this.

Google video tapes search to learn how people actually use the tool.
It is not just algorithms that need to improve; it is also the function of people use.

Examination of user behavior during web information tasks
Melanie Kellar and Carolyn Watters

1 – look at high level tasks and behaviors
2 – web based monitoring fits into our categorization of web information tasks
3- ho
Field study on the web
Used a automatic method of gathering data on a customized browser
Categorized around fact finding

Significant difference in how people do things.

Talk tomorrow afternoon on the material of classifying web information

Browsing – more useful to provide an awareness of the rate of change to a monitored age

Fact Finding – users should be able to easily identify the factual information they monitor and should only be notified when this information changes
How do we deliver this information?
Information gathering
Stored queries may be useful
Email may be the most appropriate form of notification

Transactions most difficult - airfare was often cited as participants most complex monitoring activity

Evaluate monitoring tools developed based on the findings
Validate the design
A high level classification of web information task
Characterization of user behavior during web design

Exploration of interface design for digital libraries

3 phase study – design features, effects of these changes on end user performance, propose basic and advanced features for the design of digital libraries

Pilot study
Assumptions – need to know the user – who, what, when
10 participants
Various colleges within the university

Think out loud study on interface
Wire frame here looks just like our environment, top and left nav, search, content and utilities.
Desire to increase the scent in the content area.
Leveraged ben schniderman’s work – university of Maryland
Looks like that news site, we have used / looked at before.

Found that these concepts did not work – browsing was problematic, image retrevial did not work well, rss feeds.
Tony.moore@ischool.drexel.edu

Participants wanted more interactivity, more graphics and a “cooler” interface.

An HCI Research Tool to investigate consumer interactions with Retail web sites
Sub set of a larger experiment – doctorial thesis
No existing system that would allow you to develop and create a retail site and do all of the functions online.

Point of the study was the development of a tool that allowed for testing without interaction of a facilitator. Worked well.
Results showed that browse and search is task related, not personality based. Need to review the tasks that people undertake with the environment, determine if they are browse or search related and then build accordingly. By capturing everything, you can reexamine the data for later questions.

Beyond HCI presentation notes continued:

Richard Florida – world has creativity peaks – eg taipeo, Beijing, Singapore, Tokyo, San Francisco- creative centers

Gary Olson & Judy Olson – distance matters, closer is better: unplanned encounters, familiarity

Article in NY times on Second Life

Wikimania – conference on Wiki use.

History flow of wikipedia – shows how often the article has been changed.

Make information universally usable, sociable
Ben Schneiderman – do a google on this guy, many people mentioned him.

Creativity support for individuals, groups and communities
Information literacy & efficacy

What is Self Efficacy?

Boundaries between online and offline

5 take aways
1 – software environments change rapidly
2 – new information and usage paradigms
3 – relationships and collaboration amoun broad range of users
4 – social interaction around the need to connect & contribute
5 – what works online and offline is similar, but how it works is different

Send email and ask for presentation

CC2007 June 13-15 in Washington on creativity.

Tools for studing communities – tendency to try and aggreage individuals into groups, need to use other methods to get at groups – Delphi method, etc.
Look into these tools as we go forward.
Challenge is to be seen as people who create and add value, not just studies.

Interaction: Beyond Retrieval

Interaction: Beyond Retrieval

Looking beyond current state of keyword searching and hyperlink browsing.

Open ended inquiry, complex decision making, how do we go from here

Conceptualizing interaction – Karl Fast

Jigsaw puzzles
You sort, you rearrange, you have different strategies, whole pictures, or small sections etc
Path to completion is not a straight line, instead is a more chaotic path.
If it were a straightline path, we would want to minimize the number of interactions. Any step off that path would be a mistake. This is obviously not the way the world works
What is interaction?
Does interaction have epistemic value?
Do we need more interaction, not less?

We make a distinction between Micro and Macro level interaction
Action and reaction are coupled together to create interaction.
Knowledge in the head and knowledge in the world.

Interaction can serve as a epistemic extension of representations.

Many different kinds of interactions

Task based micro level interactions
Animating Filtering Annotating fragmenting chunking probing Cloning
Interaction versus interaction technique
Descriptive framework, not a prescriptive framework.

Invent = Interactive Visual ENvironmenTs

Representations activities
Digital Objects <> interactions <> Actors
Ecologies

Fast & Sedig (2005) examining the role of information visualization in the reconceptualization of digital libraries. Journal of Digital Information
http://journals.tdl.org/jodi/article/view/jodi-169/69

karl.fast@pobox.edu

From Information Seeking to Sensemaking
Yan Qu – University of Maryland

Dervin’s Sensemaking Model


Situation
Help Gap

Needs (gap)
Use Bridging

How do people express their needs?
Structural information Need

Need
Produce a TV program Iron Chef
Sub needs
Learn about different cuisines
Where to find good chefs
Sub sub needs
Learn about French Cuisine
Understand position in Hierarchy in a kitchen

Information need is not a single need – there are rich structures within a single need

How can an information system help users to manage a set of structured needs?

Purpose of Knowledge gaining

Where do people obtain structure ideas when building knowledge representation in an information sekeing task? Existing knowledge structures from other people’s sensemaking process
How do people comprehend a large textual dataset?
Different stages and activities involved in information comprehension
Users behavior shaped the cost structure of interactions

New focus of information seeking study
To support he creation of representations for external explict knowledge
To build interactions

yanqu@umd.edu

Interaction
Interaction is a kind of action that occurs as tow ir more objhects have an effect upong one another

Only content has effect on the user
Is this what I want?
Did I ask the right question?
Now, I know a bit more . . .but

How do we interact with content?
We still need the system and interface
We need to understand the difference between interacting with system or interface
Interacting with content is much more difficult and much more important.

Interact with interface
Anticipation
Autonomy
Consistency
Exploratory

Interact with content
- Help me think
- Choose an effective representation structure
- help people think about the content they are seeing
Guide the user through dynamic content structure
Make best use of interactive tools

Examples
AuthorMap
Visual Content explorer

Example Author Search
Why do we limit author searching to author string matching
What other content will help the user?
The number of dovuments the author ahthored
Subjects of those documents
Related authors
Relationships of those authors
Subject areas of those authors
Showed a content map of authors to demonstrate the relationship between authors

Dynamic concept prepresentaiton
Studied from a poing to view of networks
Built upon recent discoveries in the science of netwoeks
Naturally occurred hubs

Design for interacting with content will be different than system interfaces
Research and prototypes are needed to explore in this direction


Towards a more intergrated model for IR

Berrypicking model (Bates 1990)

Ir is an evolving process

User interfaces for Ir should support multiple concepts
It becomes diffcult to:
maintain query context,
maintain task context
manage information objects systematically

More integrated IR environment to support advanced IR tasks.
Need more collaboration between IR and HCI communities

Interaction with content with natural language in an integrated environment

Why Natural Language
It is natural?

Handles complexity

People don’t use Boolean / formal query languages
Visualization techniques are not generic
So we need robots, something that understands us

How can we help the machines
Organize our data
Share our knowledge
- telling them is hard
- answering their questions is easier

XML to organize our knowledge –
Interactive dialog with machine
Machine can’t understand everything, so aim for what they can, then make them ask for help.
Collaboration
People – machines
People – people
Integrate with domain knowledge
Integration with visualization
Integration with full text techniques

Keynote from Sunday

Keynote – Network Analysis

www.orgnet.com

speaker talked to networks, social, internet, intranet as being scale free networks. They grow on a log log basis.
Erdos – Reyi model – Nodes and Links
Degree= how many links

Scale free network
First mover advantage – k(t)~ta
Fitness of the node modifys the first mover advantage, and somewhat removes it.
K(t)~tna
Where n = fitness

Fit –get – rich
Versus
Bose – Einstein condensation model

Robustness & Fragility
Network has a point at which it breaks down; as nodes are removed.
Scale Free -> no issues with removing items, if random items are removed. If a targeted attack made on the hub nodes, then model quickly disintegrates.

Modularity – pallar, ALB, Vicsek
Small communities have a long life, as long as members are stable over time
Large communities have a long life, as long as members change rapidly

Information diffusion in a network – best to hit the most active users, not just random users. Find the hub nodes and start there – quicker diffusion of information.

Looked at document life in a web site, typical number of hits Day 1 28%, Day 2 7%

In general, people do things in bursts. This is a result of their prioritizing things.

www.nd.edu/~networks

Information Retrevial Session on Monday

Visualization of search results

3d Information Visualization: an introduction and practical applications
Brad Eden – eden@library.ucsb.edu
Doctorate in music, edits a bunch of stuff

What is information visualization
The use of computer supported interactive visual representations of abstract data to amplify cognition
Readings in information visualization: using vision to Think

Tell me and I’ll forget
Show me and I may remember
Involve me and I will remember

David Rumsey Map Collection at http://www.davidrumsey.com/gis/3d.htm
www.virtualworld3d.com/p_travel.htm

Olive
www.otal.umd.edu/olive

Look to the atlas of cyberspaces for 3d sites

Lexington public library – aqua browser, topic map.

Using the structured metadata to present the connections between content.
Standard topic map

www.grokker.net

live plasma – displays music in a universe model

cubic eye – empty box you can play with
5 screens you can manipulate

Slide deck is available from his web site.

Information Visualization and Large Scale Repositories
Linn Marks Collins

NSF cyberinfrastructure vision ofr 21st century discovery

Data set – 60 million documents
Problem solvers as users
Higher order thinkers
Understand abstract representations of information
Tend to think in multiple ways
Would rather do science than learn a new interface

Two search result visualization products
Active graph
Interactive scatter plot of search results
By knowing her users she was able to develop a interface that is intuitive to a scientist, but not intuitive to anyone else
The interface shows the number of citations of articles, category, the journal, and the metadata. Incorporates filters. Very convoluted interface, with no clear functions, but – it works for the audience.
Increases information density
Quantitative and qualitiative data can be displayed
Feed back has been positive
Paper published last year on this tool

Aggregation of bibliographical data
LANL
More information in repositories = more time and effort to learn what is in the repository
Know what people are looking for, know the sources they go to, and what they find interesting – so why should they have to find it? Can’t we supply their needs directly?

This is interesting – could we do something similar? We know area they work in, we know their engagements, can we supply an information feed to supply most of their needs?

ScienceSifter
Uses information feed (RSS)
Created from local data
From external sources
Combine and filiter multiple feeds
Seeral viewing options
List
List with descriptions
Visualization with hyperbolic tree

Scientists work together with the channel editors to identify their reading needs
Creates the feed, then lets them read it and save for later
This is TPC’s!?!

Hyperbolic tree
Center group
Next is the key words
Each line from the keywords are the journals
Each line out of there is an article
Interactive, you can select the keyword, then the journals, then the articles

Features asked for is the ability highlight specific items, and save items for later.

Paper presented and available on this
Oncosifter another paper on hyperbolic tree displays

As more data is available, more work is needed to retrieve previous research, where analysis after the fact is more important than the classic experimentation.
IR and data visualization become more important.

Check for interesting papers by these people.
Mark Marinez
Tamara McMahon
Ketan Mane
Rick Luce
Miriam Blake
Jeremy Hussell

Collective Intelligence & Holistic Sense Making

Chaomei Chen, Drexel University

Sum of the parts > whole

What are the hot topics in subject?
How are the hot topics related?
How do they evolve over time?
What are and how do we assess the emerging insights?

Demonstrated a cluster representation of search results demonstrating the connections between papers, clusters etc.

Demonstrated social network analysis

Showed a interesting mash up between google maps and citation index on terriorism. Very powerful impact to see where and when things are displayed.

Retrieval vs visual Analytics

Retrieval
Recall
Discret search
Part formal

Visual
Recognition
Continuous foraging
Whole
Intuitive

Chaomei.chen@cis.drexel.edu
www.pages.drexel.edu/~cc345