User Interface and Visualization Lecture for CS 410

User Interface and Visualization (Lecture for CS 410 Intro Text Info Systems) April 13, 2007 Cheng. Xiang Zhai Department of Computer Science University of Illinois, Urbana-Champaign

IR System Architecture docs INDEXING Query Rep Doc Rep SEARCHING Ranking Feedback query User results INTERFACE judgments QUERY MODIFICATION

This Lecture • From “Retrieval” to “Information Access” • Principles for designing a retrieval interface • Information seeking models • Elements of a retrieval interface

What is Information Access? A process for a user to find and understand information relevant to the user’s task from sources of information

Requirements of a “Full-fledged” Information Access System Interactive User modeling Search & Organize A process for a user to find and understand information relevant to the user’s task from Multimedia sources of information Integration & Domain modeling Context modeling Information need modeling& satisfying

A Typical Retrieval System ? Interactive User modeling Search & Organize A process for a user to find and understand Text information relevant to the user’s task from Multimedia sources of information Integration & Domain modeling Context modeling Information need modeling& satisfying

Next, we’ll see how a better user interface can add value to a typical retrieval system

So, What’s a Good Interface? • Offer informative feedback • Reduce working memory load • Provide alternative interfaces for novice and expert users • …. Human-Computer Interaction (HCI) is an active research area…

“Standard” Interaction Model (Hearst 97) Information Need static Query Send to System Reformulate Receive Results Evaluate Results No Done Yes Stop results

“Berry-Picking” as an Information Seeking Strategy (Bates 90) • Standard IR model – The information need remains the same throughout the search session. – Goal is to produce a perfect set of relevant docs. • Berry-picking model – The query is continually shifting. – Users may move through a variety of sources. – New information may yield new ideas and new directions. – The value of search is on the bits and pieces picked up along the way. (Slide from “Marti Hearst, UCB SIMS, Fall 98)

A sketch of a searcher… “moving through many actions towards a general goal of satisfactory completion of research related to an information need. ” (after Bates 90) Q 2 Q 1 Q 4 Q 3 Q 5 Q 0 (Slide from “Marti Hearst, UCB SIMS, Fall 98)

Elements of Retrieval Interface • Document selection • Query specification • Result examination • Interaction support (feedback)

Document Selection (Starting Points) • Lists • Overviews • Automated source selection

List of Collections: LEXIS-NEXIS

Overviews (Adding Structures) • An overview shows the topic domains to help users select or eliminate sources from consideration • Types of overviews – Category or Directory Overview – Automatically Derived Collection Overview (clustering, query-specific)

Category or Directory Overviews Me. SHBrowse Interface for viewing category labels hierarchically

Automatically Derived Collection Overviews Display of Scatter/Gather clustering retrieval results

Automatically Derived Collection Overviews (cont) A three-dimensional overview based on document clustering

Automatically Derived Collection Overviews (Kohonen Map) Web pages about entertainment

Automated Source Selection • Create a representation of the contents of information sources • Match it against the query specification • Suggest sources • Adaptive source suggestion…

Query Specification • Formulate a query – Select: collections, metadata descriptions, information sets • – Specify: words, phrases, descriptors… Five major styles – Command language – Form filling – Menu selection – Direct manipulation • – Natural language Query by examples

Form Filling & Menu Selection A view of query specification in the Web-based version of the Melvyl bibliographic catalog.

Direct Manipulation • • A graphical approach Properties: – Continuous representation of interest – Physical actions or button presses • – Rapid incremental reversible operations Examples: – Graphical depictions of Venn diagrams – Filter-flow model

Venn Diagram The VQuery Venn diagram visualization for Boolean query specification

Filter-Flow The filter-flow visualization for Boolean query specification

Examining results…

Document Surrogates (cont) An example of a ranked list of titles and other document surrogate information

Tile. Bars An example of the Tile. Bars retrieval results visualization

See. Soft An example of the See. Soft visualization for showing locations of characters within a text

Super. Book The Super. Book interface for showing retrieval results on a large manual in context

Categories for Results Set Context The Dyna. Cat interface for viewing category labels that correspond to query types

Organize Results by Hyperlinks The Cha-Cha interface for showing Web intranet search results in context displaying results on the query `medical centre'

Graphical Depiction of Links Example of a Web subset visualized by Mapuccino

Tables The Envision tabular display for graphically organizing retrieved documents

Using Relevance Judgments • • • A standard interface consists of a list of titles with checkboxes beside the titles that allow users to mark relevant documents Provide a choice among several checkboxes indicating relevant or not relevant The system either automatically reweights the query or generate a list of terms to augment the original query

Interaction Support …

Standard Relevance Feedback An example of an interface for relevance feedback

Studies of User Interaction • • Standard relevance feedback assumes the user is involved by specifying relevance documents Level of user control – Control: No relevance feedback – Opaque: Select relevant documents, get revised rankings – Transparent: Know how the system reformulate the queries and revised rankings – Penetrable: Stop in the midway through reranking. Users select terms and system presents revised rankings

Retaining Search History • User interface should show – The available choices at any given point – Movements made in the past – Short-term tactics & long-term tactics – Annotate choices and information along the way • – Bundle search sessions and save individual portions History mechanism – A list of commands executed earlier – Graphical history for tracking commands and results

History Visualization The VISAGE interaction history visualization

Integrating Operations • In general, user interfaces for information access do not do a good job of supporting strategies – Users prefer an integration of scanning and query specification – Unrestricted interaction may lead to erroneous behavior, and interactions between two different modes require more guidance

Integrating Operations (cont) A view of query history revision in the Web-based version of the Melvyl bibliographic catalog

Integrating Operations (cont) The Cat-a-Cone interface for integrating category and text scanning and search

Integrating Operations (cont) An interface for a starting point for searching over category labels

Trends and Research Issues • Research is expected to increase rapidly, primarily because of the rise of the Web – New information access paradigms – Interest in organization and search over personal information collections – Recommender systems • New ideas about how to display large, abstract information spaces intuitively

What You Should Know • Search/Retrieval is only ONE WAY to achieve the high-level goal of information access • A real application needs a carefully designed user interface • What counts as a good interface often depends on both the type of users and the domain of information • Many interesting research topics in improving a retrieval interface, but hard to evaluate…