Boolean Browsing in an Information System: An Experimental Test

Boolean browsing presents a two-dimensional word map in response to an initial user query. Users interact with this word map to obtain a list of suggested search terms, which then can be used in information retrieval. This experiment tested the impact of Boolean browsing on retrieval effectiveness with users of standard bibliographic data who had different levels of cognitive abilities, and who were working on different tasks. The results showed a positive impact for this approach to vocabulary selection in users with higher levels of cognitive abilities (particularly spatial scanning abilities), who were accomplishing a high-precision, low-recall task. The implications of this result/or the design of information systems are discussed. Users of information systems are frequently required to select (often without much guidance) appropriate words to describe their topic of interest. Sometimes the vocabulary that describes a topic contains specialized jargon that nonspecialists may not use. In other cases, the natural messiness of language (including synonyms and homonyms) can make vocabulary selection difficult. Information systems may use a variety of mechanisms to reduce the difficulties of vocabulary selection. Controlled vocabulary is one mechanism that can help. Subject heading lists and thesauri can be presented in such a way that users can see the relationships between words and select appropriate search terms. However, controlled vocabulary may also act as a barrier for the user, who has to learn to employ this restricted word set, and to navigate from unauthorized to authorized topic descriptors. A second solution is a vocabulary suggestion system that provides possible search terms to users. The simplest form of such a system, in use since the beginning of interactive information retrieval and still found in many information systems, is a "browse" command. Users enter a word or phrase, and are shown a vocabulary list (which may consist of either controlled vocabulary or natural language) that corresponds alphabetically to the word or phrase they enter. The shortcomings of such a system are obvious. The user may enter a word or phrase that does not match alphabetically with appropriate search vocabulary. In addition, vocabulary lists, typically generated from the inverted index of keywords or subject headings, can be very long. Traversing even a small part of that list by unidirectional browsing may be impractical. Recognizing these problems, Allen investigated an expandable vocabulary list for browsing that allowed users to expand and collapse a browse list by selectively showing subheadings (Allen 1993a). This solution provided greater economy of searching, but at the cost of a more complex vocabulary suggestion system. A step beyond a simple vocabulary list in the vocabulary suggestion system can be achieved by moving from one dimension to two. Vocabulary can be presented in a two-dimensional word map rather than in a one-dimensional list of terms. This approach was investigated (Allen 1998) and found to provide important advantages over traditional vocabulary browsing. In particular, the word map allowed users with lower levels of cognitive abilities to make more effective use of the information system. However, the advantages were achieved by adding complexity to the interface, and by requiring reasonably high levels of additional computing to create word maps in real time. In this investigation, a refinement of the word map (allowing Boolean browsing) was tested. This refinement increased the amount of interaction between the user and the vocabulary suggestion system. The objective of this investigation was to assess the usability of Boolean browsing by users who have different levels of cognitive abilities, and who are accomplishing different tasks. The research questions focused on how Boolean browsing enhanced search performance in a variety of search situations. …