Week 9 Discussion

Chapter15.pdf

Home >Information Systems homework help >Week 9 Discussion

15.2 Five-Stage Search Framework

In designing the advanced search interface, a five-stage search framework may help to coordinate design practices and satisfy the needs of first-time, intermittent, and frequent users. The five stages of action, illustrated more fully in Box 15.1 , are:

1. Formulation: Expressing the search 2. Initiation of action: Launching the search 3. Review of results: Reading messages and outcomes 4. Refinement: Formulating the next step 5. Use: Compiling or disseminating insight

Information seeking is an iterative process, so the five-stages can be repeated many times until users’ needs are met. Users may not see all the components of the five stages, but if they are unsatisfied with the results, they should be able to have additional options and change their queries easily.

PRINTED BY: Fran Blow <[email protected]>. Printing is for personal, private use only. No part of this book may be reproduced or transmitted without publisher's prior permission. Violators will be prosecuted.

2/19/2018http://e.pub/r6kac6j9uer7lvx1pcjc.vbk/OPS/xhtml/fileP7001012430000000000000000002...

Box 15.1 Five-stage framework to clarify search user interfaces

1. Formulation ◾ Use simple and advanced search. ◾ Limit the search using structured fields such as year, media, or location. ◾ Recognize phrases to allow entry of names, such as “George Washington.” ◾ Permit variants to allow relaxation of search constraints (e.g., phonetic variations). ◾ Control the size of the initial result set. ◾ Use scoping of source carefully. ◾ Provide suggestions, hints, and common sources.

2. Initiation of action ◾ Explicit actions are initiated by buttons with consistent labels (such as “Search”). ◾ Implicit actions are initiated by changes to a parameter and update results

immediately. ◾ Guide users to successful or past queries with auto-complete.

3. Review of results ◾ Keep search terms and constraints visible. ◾ Provide an overview of the results (e.g., total number). ◾ Categorize results using metadata (by attribute value, topics, etc.). ◾ Provide descriptive previews of each result item. ◾ Highlight search terms in results. ◾ Allow examination of selected items. ◾ Provide visualizations when appropriate (e.g., maps or timelines). ◾ Allow adjustment of the size of the result set and which fields are displayed. ◾ Allow change of sequencing (alphabetical, chronological, relevance ranked, etc.).

4. Refinement ◾ Guide users in progressive refinement with meaningful messages. ◾ Make changing of search parameters convenient. ◾ Provide related searches. ◾ Provide suggestions for error correction (without forcing correction).

5. Use ◾ Imbed actions in results when possible. ◾ Allow queries, settings, and results to be saved, annotated, and sent to other

applications. ◾ Explore collecting explicit feedback (ratings, reviews, like, etc.).

2/19/2018http://e.pub/r6kac6j9uer7lvx1pcjc.vbk/OPS/xhtml/fileP7001012430000000000000000002...

15.2.1 Formulation

The formulation stage includes identifying the source of the information (i.e., where to search). Users may want to limit their searches for flights to certain travel sites, or limit a text search to help documents and not the entire web (Fig. 15.3 ), or search only women’s clothing on a shopping website (e.g., women’s shirts). Also called scoping, this limitation of the source can lead to better results but also lead to failures when the constraint remains active and users forget about it. Clearly displaying the source of information is important.

Even if it is technically feasible, searching all libraries or the entire web may not always be the best approach. Users often prefer to limit the search to a specific library or collection in a library (e.g., within the manuscript collections of the Library of Congress or within Wikipedia). Keywords or phrases can be specified, and structured fields such as year of publication, volume number, or language can be used to further limit the search scope. A text box and a few menus may be enough in most cases, but form fill-in (Section 8.6 ) allows users to specify more detailed searches in databases (e.g., search for nonstop flights between three local airports and New Orleans across a range of possible dates).

In database searches (e.g., Fig. 15.2 ), users often seek items that contain meaningful short phrases (“Civil War,” “Environmental Protection Agency,” “carbon monoxide”), and multiple-entry fields can be provided to allow for multiple phrases. Searches on phrases have proven to be more accurate than searches on individual words. Phrases also facilitate searching for names (for example, a search on “George Washington” should not turn up “George Bush” or “Washington, DC”). If Boolean operations, proximity restrictions, or other combining strategies

Figure 15.3 The Yahoo! help search box has two buttons of different colors to search two different sources of information: purple for searching the help information and blue for searching the web. Pressing the purple button “scopes” the results to the help information only and shows results below a purple banner. Searching the web jumps to a different page (the normal search) that reuses the blue button color, helping users keep track of which source of information they are searching.

2/19/2018http://e.pub/r6kac6j9uer7lvx1pcjc.vbk/OPS/xhtml/fileP7001012430000000000000000002...

are specifiable, users should also be able to express them. Users or service providers should additionally have control over stop lists (which typically filter out from the search terms common words, single letters, and obscenities).

When users are unsure of the exact value of the field (the terms to be searched for or the spelling or capitalization of a name), the search constraints can be relaxed by allowing variants to be accepted. In a textual-document search, advanced search interfaces may give users control over variant capitalization (case sensitivity), stemmed versions (the keyword “teach” retrieves words with variant suffixes such as “teacher,” “teaching,” and “teaches”), partial matches (the keyword “biology” retrieves “sociobiology” and “astrobiology”), phonetic variants from soundex methods (the keyword “Johnson” retrieves “Jonson,” “Jansen,” and “Johnsson”), synonyms (the keyword “cancer” retrieves “malignant neoplasm”), abbreviations (the keyword “IBM” retrieves “International Business Machines,” and vice versa), and broader or narrower terms from a thesaurus (the keyphrase “New England” retrieves “Vermont,” “Maine,” “Rhode Island,” “New Hampshire,” “Massachusetts,” and “Connecticut”).

When searching in a simple list of items (e.g., searching a name in a list of contacts), the result list can be displayed as users type. The list shrinks rapidly and users can select the wanted item without finishing typing the name. When the collection is large, auto-completion can be applied to the search term instead, revealing most common search phrases that match the text already typed (Fig. 15.4 ). The

Figure 15.4 Auto-complete suggestions can speed data entry and guide users toward successful queries.

a. In a mobile phone address book, typing one character filters the list to all names that contain that character, and the list is updated continuously as users type.

b. Typing “helm” in Amazon’s search box shows suggestions for “helmet light” or “welding helmet” but also suggestions to narrow the scope of the search to relevant departments.

c. On the Adobe website, suggestions include products (e.g., typing the beginning of the word “video” suggests several video-editing tools).

auto-complete list is updated as users continue typing, which helps them recall terms of interest, limits misspelling, and speeds up the query initiation process.

2/19/2018http://e.pub/r6kac6j9uer7lvx1pcjc.vbk/OPS/xhtml/fileP7001012430000000000000000002...

Mobile applications may use context information such as location to narrow down the auto- completion suggestions. For example, searching on a map can narrow the history of previous searches to the ones relevant to the current location. This may allows users to find the exact location of a doctor’s office for a follow-up visit even if they don’t remember the name or exact address of the doctor—an impressive combined use of location, history and auto-completion.

For regular users who want additional control, advanced command languages can be offered to search databases (e.g., SQL; see Section 15.4 ). On the other hand, new users will benefit from the reading of typical phrases (which can be placed next to the search box), direct links to often searched items (e.g., sales or popular topics), and carefully designed tips. The seamless integration of search with navigation and browsing will allow users to switch to menus of choices when they are not able to come up with search phrases or to review sample materials to better understand what is available even before they start composing their query (see also Section 15.3 and Fig. 12.1 of NASA’s Earthdata search interface).

15.2.2 Initiation of action

The second stage is the initiation of action, which may be explicit or implicit. Many systems have a search button for explicit initiation. A magnifier glass has become the de facto standard icon for search when space is limited, but pressing the Enter key on a keyboard or pausing during spoken interaction may be the only thing needed to initiate the search.

An appealing alternative is implicit initiation, in which each change to any component of the formulation stage immediately produces a new set of search results. Dynamic queries in which users adjust query widgets to produce continuous updates (Shneiderman, 1994) led the way in demonstrating the benefits of implicit initiation, and they have been widely adopted in faceted browsing (Section 15.3 ).

2/19/2018http://e.pub/r6kac6j9uer7lvx1pcjc.vbk/OPS/xhtml/fileP7001012430000000000000000002...

15.2.3 Review of results

The third stage is the review of results, in which users review results in textual lists (Fig. 15.5 ) or on geographical maps (Fig. 15.6 ), timelines (see HIPMUNK in Fig. 1.7 ), or other specialized visual overviews of results. If no items are found, that failure should be indicated clearly. When messages are worded carefully (see error messages in Section 12.8 ) and useful suggestions are provided, users are less likely to abandon their search (e.g., leave a shopping website to never come back).

When results are presented in a list, it is common practice to return only about 20 results, but larger initial sets are preferable for those with high bandwidth and large displays. Previews consisting of carefully selected text samples (or snippets; see Fig. 15.5 ), human-generated abstracts, photos, or automatically generated summaries help users select a subset of the results for use and can

2/19/2018http://e.pub/r6kac6j9uer7lvx1pcjc.vbk/OPS/xhtml/fileP7001012430000000000000000002...

Figure 15.5 A Google Search result list. A summary is provided at the top (the total number of results). Each result includes preview information (or a snippet). Search terms are highlighted, including “Human- Computer Interaction Lab,” which is the expanded variant of the search term “HCIL.” The name of the top-level organization was added (here “National Center for Biotechnology Information”) to help users judge the trustiness of the information.

2/19/2018http://e.pub/r6kac6j9uer7lvx1pcjc.vbk/OPS/xhtml/fileP7001012430000000000000000002...

Figure 15.6 Searching for Annapolis on the real estate website Zillow returns a list of houses and dots displayed on a map. The two windows are coordinated; when the cursor hovers over a house in the result list, the location of the house is indicated on the map. A click on the house would bring all the details displayed in an overlapping window.

help them define more productive queries as they learn about the contents of the items (Greene et al., 2000). Translations may also be proposed. Allowing users to control how results are sequenced (e.g., alphabetical, chronological, relevance ranked, or by popularity) also contributes to more effective outcomes. If users have control over the result set size and which fields are displayed, they can better accommodate their information-seeking needs.

Highlighting the search terms in the snippet or other preview helps users gauge the value of results. Previous visits are noted. For websites, the URL is partially visible, and the name of the organization is provided—when available—helping users gauge the trustiness of the information. In database search, the preview information might indicate which collection the item belongs to or include a photo and important attributes (Fig. 15.6 ).

Additional preview and overview surrogates for items and collections can be created to facilitate browsing of results. Graphical overviews can indicate scope, size, or structure and help gauge the relevance of collections (e.g., with maps, timelines, or diagrams). Previews consisting of samples from collections entice users and help them define productive queries (Greene et al., 2000).

When the number of results is very large and metadata is available, a useful strategy is to provide an overview of the number of items in available categories (see also faceted search in Section 15.3 ). For example, when searching a library catalog, the number of books, journal articles, or news articles can be indicated (Fig. 15.7 ) and allow users to filter the results. When no metadata is available, one strategy is to automatically cluster the results based on content analysis (see, for

2/19/2018http://e.pub/r6kac6j9uer7lvx1pcjc.vbk/OPS/xhtml/fileP7001012430000000000000000002...

Figure 15.7 A search for “human computer interaction” powered by Summon for a university library catalog

returns a very large number of results. On the left, users can see the number of results for categories organized by content type, subject terms, or publication date. The box provides an overview of the results, reveals how the search was done (e.g., here the default search does not returns dissertations), and facilitates further refinement of the search. The menu at the upper right allows users to sort results by relevance or by date. Help is available with a “Chat now” button, which allows users to chat with a librarian (http://www.lib.ncsu.edu).

example, Yippy at http://www.yippy.com). This allows users to navigate a tree of hierarchically organized topics, but the quality and appropriate labeling of the clusters are often problematic, so this technique is losing proponents.

To help users identify items of interest, access to the full document is usually necessary, with highlighting of the terms used in the search. For large documents, automatic scrolling to the first occurrence of the keyword is helpful, as are markers placed along the scrollbar to indicate the

™

2/19/2018http://e.pub/r6kac6j9uer7lvx1pcjc.vbk/OPS/xhtml/fileP7001012430000000000000000002...

locations of other occurrences (see Fig. 9.8 ). A common issue when reviewing results within a document is to have the search box hide the location of the terms found in a document.

15.2.4 Refinement

The fourth stage is refinement. Search interfaces can provide meaningful messages to explain search outcomes and to support progressive refinement. For example, corrections can be proposed, such as asking, “Did you mean fibromyalgia?” when a term is misspelled. If multiple phrases were used, items containing all phrases should be shown first and identified, followed by items containing subsets. Progressive refinement, in which the results of a search are refined by changing the search parameters (e.g., search phrases but also time range, location, etc.), should be made convenient by leaving the search terms active—along with an easy way to clear them—instead of asking users to start from scratch every time.

2/19/2018http://e.pub/r6kac6j9uer7lvx1pcjc.vbk/OPS/xhtml/fileP7001012430000000000000000002...

15.2.5 Use

The final stage, use of the results, is where the payoff comes. Results may be merged and saved, disseminated by e-mail, or shared in social media. Users may want to feed the results to a bibliographic tool or be notified when new results become available. Sometimes direct answers or actions can be embedded directly in the result lists (Fig. 15.8 ), but most often search is only one of many components of a more complex analysis tool. For example, powerful environments are available for lawyers to review previous lawsuits and assemble supporting materials for their cases. Intelligence analysts might use tools such as nSPace from Uncharted Software (Fig. 15.9 ) to prepare evidence-based reports. Multiple searches can be specified at once, and names, dates, places, and organizations are

Figure 15.8 When possible (and important), provide information or simple actions without requiring users to leave the search results page. On the left, Google Search users get the answer to their safety- critical question at the top of the result list. On the right, Peapod shoppers looking for groceries can specify quantity and buy directly from the list of results after a search on “grapes.”

2/19/2018http://e.pub/r6kac6j9uer7lvx1pcjc.vbk/OPS/xhtml/fileP7001012430000000000000000002...

Figure 15.9 nSpace TRIST from Uncharted Software is used by analysts to produce evidence-based reports. For this (fictitious) criminal investigation, a user is reviewing a collection of documents (shown as icons at the top). The search history on the left shows three searches. Names, places, and organizations have been automatically extracted. Here the search term “Laboratory” and the person “John Panni” are selected. Snippets are shown on the bottom left. Analysts use TRIST to define dimensions of interest and to quickly identify documents of interest for use in a Sandbox (Fig. 15.10 ) (http://uncharted.software/nspace; Chien et al., 2008).

extracted automatically. Analysts review documents and export information into a Sandbox (Fig. 15.10 ) to organize the evidence they found, mark it as supporting or refuting hypotheses, and then generate reports.

Many searches are related to searches done in the past. Users often need to find the same information again or to continue a search started the previous day. A search history, bookmarking, tagging, and indication of past visits in the results (e.g., “You have visited this page 2 times, Last visit 1/6/2016”) all contribute to helping users re-find information. Keeping the visited links visually distinct reminds users of where they have been.

Designers can apply the five-stage framework to make the search process more visible, comprehensible, and controllable by users. The five stages are often repeated many times until users needs are met.

2/19/2018http://e.pub/r6kac6j9uer7lvx1pcjc.vbk/OPS/xhtml/fileP7001012430000000000000000002...

Figure 15.10 nSpace Sandbox from Uncharted Software™ allows multiple analysts to organize and present

the evidence gathered from research. A variety of tools such as node and link diagramming, automatic source attribution, recursive evidence marshaling, and timeline construction provide support for analysis and reporting.

2/19/2018http://e.pub/r6kac6j9uer7lvx1pcjc.vbk/OPS/xhtml/fileP7001012430000000000000000002...