Christopher Olston, Ed Chi
We propose a new modality for locating information in hypertext environments called ScentTrails. This modality combines the strengths of searching and browsing in a unified interface. In ScentTrails, a user types keywords representing the portion of her information goal that is initially known and compatible with keyword search. Then, a special proxy located between the content provider and the user customizes subsequently accessed hyperlink documents by adding search cues, which are indications that a hyperlink leads to content that matches the search query.
In an initial experimental study, test subjects completed a set of information location tasks relevant to the Xerox.com Web site significantly more quickly using ScentTrails than using existing browsing and searching technology. On average, ScentTrails improved user performance by about 50 percent.
Our work is motivated by the observation that two predominant yet imperfect interface modes currently exist for locating information on the Web: searching and browsing. Keyword searching allows users to identify pages containing specific information quickly; each search is tailored to a user's particular information goal, when formulated as a list of keywords. In contrast, browsing is advantageous when appropriate search keywords are impossible to determine. For example,
- The user's information goal may not be fully formed at the outset and the user may not be certain of what she is looking for until the available options are presented during browsing.
- Even if the full information goal is known at the outset, the exact terminology used on the Web pages may not be known, and therefore searching often will not yield the correct result.
- Certain information goals, such as ones that involve semantic predicates like "price < $200," cannot be expressed using existing general-purpose keyword search technology.
- Finally, browsing is appropriate when a great deal of information and context are obtained along the browsing path, not just at the final page.
Searching and browsing offer complementary advantages. However, neither modality is well equipped to handle rich or complex information goals consisting of multiple criteria; some criteria lend themselves well to searching and others are better suited to browsing. Regrettably, it is difficult to reap the benefits of both searching and browsing simultaneously by simply switching between the two modalities.
ScentTrails achieves smooth integration between searching and browsing by embedding customized search cues in existing Web pages. An example is shown in Figure 1. A portion of the Xerox.com Web site was modified by our ScentTrails proxy, which increased the font size of certain links to display search cues for the partial information goal "remote diagnostic technology." Links in a large font lead to product descriptions containing that phrase. Search cues may appear in different gradations (for instance, high, medium, low) depending on degree of relevance and distance in number of hops; search cue gradations are computed using an Information Scent algorithm operating over page content and link topology. By considering the search cues in conjunction with existing browsing cues, users are able to make informed navigational decisions and efficiently locate content matching complex information goals that lend themselves partly to searching and partly to browsing.
Carnegie Mellon University
Ed H. Chi
Palo Alto Research Center
This abstract is from a recent issue or forthcoming issue of ACM's Transactions of Computer Human Interaction (ToCHI). It is included here to alert interactions readers to what research is being done in the field of Computer Human Interaction. The complete papers, when published, can be found in ACM's Digital Library at www.acm.org/pubs/contents/journals/tochi.
©2003 ACM 1072-5220/03/0900 $5.00
Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee.
The Digital Library is published by the Association for Computing Machinery. Copyright © 2003 ACM, Inc.