Special Section

Interface Design Concepts in the Development of a Web-Based Information Retrieval System

by Rebecca Denning, Marie Shuttleworth and Phil Smith

With the proliferation of Web-based search engines, information seekers now have a multitude of resources for satisfying their needs. The speed with which these software packages can be developed has resulted in a large number of diverse software displays that often reflect little care for usability and efficiency. This article discusses six principles that we developed for building and evaluating Web-based information retrieval interfaces as part of a recent project for the Eisenhower National Clearinghouse for Mathematics and Science Education (ENC).

Eisenhower National Clearinghouse

The Eisenhower National Clearinghouse for Mathematics and Science Education is charged with providing online access to information about curriculum materials (books, kits, software, videos, etc.) for K-12 math and science educators, parents and students. This audience is very challenging because it presents a wide range of previous experience using computers to access information. In order to serve this audience, the ENC developed an indexed database using a controlled vocabulary that includes fields such as subject, grade level, type of media and cost, as well as an abstract. Users can also access tables of contents and ordering information for curriculum materials online. They can search the database using either a free-text search or the controlled vocabulary. This project, therefore, faced a variety of the issues typically confronted by developers of Web-based information retrieval systems.

Design Concepts and Principles

In order to meet the needs of the target user population, we developed a search interface integrating both browsing and keyword searches. The team recorded questions that arose, as well as the human factors principles used as the basis for making design decisions, during its design and implementation. Thus, this paper presents a set of principles for evaluation and discusses how ENC's information retrieval system, named the Enhanced Resource Finder, does or does not fulfill these design goals.

Principle 1. Help the user develop an understanding of the operation of the interface and the search process.

A well-designed interface provides assistance to the information seeker so that the person can map his or her knowledge of the topic into the system. This assistance can take a number of forms.

Principle 2. Provide information to help users judge the value of continuing a search path.

System designers should keep in mind that many people have slow network connections and that almost everyone is short on time. Thus, any clues provided to help information seekers determine the value of continuing a particular search path are useful. It might be the case that they would more readily find their search terms if browse lists were provided both in alphabetical order and as a hierarchy of subject-related search terms. These arrangements reflect different ways of thinking about the associated list.

The hierarchical list is arranged by content area with alphabetical ordering for each subcategory. If searchers were looking for a textbook for a physics class, they could find physics under either "p" in the alphabetical list or under "science" in the hierarchical list. If the search terms are only presented in alphabetical lists, as shown in Figure 3, finding the correct label for a concept can be an arduous task. Currently implementation of hierarchical lists is under development at ENC in order to provide an alternative to the alphabetical browse lists.

Another classic way to help users judge the value of continuing a query is for the system to provide the number of records associated with each vocabulary term. Knowing that there are only two records associated with Distance learning is a potent clue as to how much information the person might find on this topic in this database (Figure 3).

Results lists need to be structured in such a way that the contents of the returned record are clear (Figure 4). The information most important to the audience should be clearly displayed in the summaries. This guides the information seeker not only in selecting which records might be of interest, but also whether the search resulted in the types of information desired. The search query should be visible while scanning retrieved results.

Principle 3. Assist the user in refining the search query or search topic.

The designer can use a number of techniques. First, provide clear and simple instructions, a Help page for those interested in query construction and the ability to reset or erase a query. In addition, summarize the query on the results page so that the information seeker can be reminded exactly what was searched. Another function that is becoming popular in some engines is to generate a list of related search topics in response to a query.

Principle 4. Avoid complex navigation.

A trade-off exists between having users scroll down a page as they complete a longer query form or designing the screen so that the query fits on one page. Further complicating matters is the wide range of monitors in use today. At ENC we tried to develop a flexible workspace that minimized scrolling or paging around. As can be seen in Figures 2 and 3, frames divide the screen into work areas. As the query is constructed, terms selected appear in the middle box. The information in the bottom frame changes based on users' selections.

Principle 5. Make the system actions explicit to the information seeker.

Instructions placed near related controls show functional grouping. Analogies using familiar objects guide many interfaces. Many people intuitively know what to do when presented with a form to complete. Boxes suggest areas to type into and buttons suggest places to click. Place functions near the object acted upon. For instance, Search and Reset buttons should be placed near the query in order to suggest applicability.

Summarizing terms included in a query is a common standard for results lists, but few systems indicate whether or how Boolean operators were employed. In order for a user to understand the system's actions the query construction needs to be clarified. Sometimes this information is available through the Help pages, but frequently it is not made explicit, leaving the user to guess how the results were obtained. Not only should the information seeker see the terms used in the query, but also the operators used to form the query.

Principle 6. Provide verbal labels suggestive of meaning.

We attempted to select verbal labels that were highly suggestive of the intended meanings and functions of their associated objects. Each window has a heading which signifies the type of information associated with it. Each button is labeled with the action that will occur if it is pressed. Finally, the label for each box describes the type of information that should be entered into it. Labels such as "Grades" and "Cost" clearly suggest the contents of a category. Alternatively, labels such as "Other" do not. Even "Help" can be misleading when the context is confused. For instance, in Figure 4, will "Help" take you to the description of the Home page or to more information on query construction? An information seeker might become confused in this particular case because whether the label refers to global and local navigation is ambiguous.

Summary

The six design concepts presented above are not intended to be an exhaustive list of principles to follow when developing or evaluating a Web-based information retrieval system. Rather, they highlight some high level concepts that should be considered in order to promote the usability and effectiveness of such systems. As demonstrated by the included figures, some of these design concepts were well followed in the design of ENC's Resource Finder, while others require enhancement. Design in the Web-page environment still involves much exploration and refinement, but we hope that our experience will assist others in this task.

Note: The authors would like to acknowledge the other design team members at ENC: Karen Abhari, Mike Connor, Tracy Crow, Doug Line, Marla Mayerson, Stephen Sarapata and Min-Te Sun. The Enhanced Resource Finder redesign occurred 1996-1997 and at the time of writing was located at http://watt.enc.org/main2.html

Rebecca Dunning is senior associate research scientist at Chemical Abstracts Service, P.O. Box 3012, Columbus, OH 43210; 614/447-3600; e-mail: rzd07@cas.org

Marie Shuttleworth is librarian at Eisenhower National Clearinghouse, Ohio State University, 1929 Kenny Rd., Columbus, OH 43210; 614/292-7784; e-mail: mshuttle@enc.org

Phil Smith is professor, Cognitive Systems Engineering Laboratory, Institute for Ergonomics, Ohio State University, 210 Baker Systems, 1971 Neil Avenue, Columbus, OH 43210; 614/292-4120; e-mail: phil+@osu.edu