Summary 02 — User Interfaces for Search

Source: 02_UserInt.pdf

1. Table of Contents (Topics Covered)

Gist (2–3 paragraphs)

This lecture shifts focus from the algorithm to the human and the interface — the window through which users see a search system. A good UI helps users express their information need, formulate queries, pick sources, understand results, and track progress. Search behavior depends on task type, domain expertise, and available time. A key distinction is information lookup (fact-finding, answered by a discrete piece of data) vs. exploratory search (learning and investigating, requiring many iterations and synthesis over time).

Modern models stress that searching is dynamic: users learn as they go and adjust their needs (the berry-picking model), often using quick imprecise queries refined by local navigation (orienteering). Studies reveal human limits: people are poor at judging relevance for unfamiliar topics, look mostly at top-ranked results, are biased toward the top one or two, and underestimate how much relevant material they’ve missed. Web queries are typically very short (1–3 words), and Boolean operators are rarely used.

Results are shown as document surrogates (title, URL, metadata, and a query-biased snippet) in a vertical list (SERP). Techniques like KWIC (keywords in context), highlighting, and blended/universal search (images, scores, etc.) improve relevance judgment. Results can be reformulated (spelling fixes, term expansion, relevance feedback) and organized (flat/hierarchical/faceted categories, or automatic clustering). The lecture closes with visualization experiments (TileBars, VIBE, Word Tree, thumbnails) and evaluation rooted in Human-Computer Interaction — discount usability methods, heuristic evaluation, longitudinal studies, and A/B (bucket) testing, emphasizing that subjective preference often matters more than raw metrics.

2. Key Terminologies — Meaning & Use Cases

Term Meaning Use Case
Information lookup Fact retrieval / question answering with a discrete answer “What’s the capital of X?”
Exploratory search Learning + investigating over many iterations Research, business intelligence
Sensemaking Iterative building of a conceptual representation from a collection Legal discovery, epidemiology
Berry-picking model Need evolves as the user sees results; query reformulated repeatedly Realistic web searching
Orienteering Quick imprecise query + local navigation to home in Typical short-query behavior
Navigation/browsing Recognizing info by following links vs. recalling keywords Directory/catalog browsing
Query specification How the need is entered (search box, links) Search entry form
Conjunctive query All query terms must appear (AND) Google ~1997 onward
Document surrogate Summary representing a document in results Title + URL + snippet
Snippet Extracted text summary, often query-biased SERP entries
SERP Search Engine Results Page The results list
KWIC Keywords In Context (query-biased summaries) Highlighting query terms
Blended / universal search Mixing media/answers into text results Images, scores, maps
Relevance feedback User marks relevant docs; system rebuilds query Query reformulation (limited UI success)
Faceted metadata Multiple category dimensions per item Filtering by facets
Clustering Automatic grouping by similarity Auto-organizing results
A/B (bucket) testing Show new design to a random subset, compare logs Live UI evaluation

3. Process Workflow Diagram

flowchart TD
    A[Information Need] --> B[Query Specification
search box / links]
    B --> C{Auto-complete /
suggestions}
    C --> D[Submit Query]
    D --> E[Results Display
SERP: surrogates + snippets]
    E --> F{Relevant?}
    F -- No --> G[Reformulate
spelling, term expansion, RF]
    G --> B
    F -- Partly --> H[Organize: categories / clustering]
    H --> E
    F -- Yes --> I[Navigate to document]
    I --> J[Learn / Sensemaking]
    J -->|need evolves: berry-picking| A

4. ELI5 — Complex Terms Explained Simply