exact match searchers. Their aim was to examine how searchers who were experienced on one type of 

searching system performed when they searched on a different search system.   

Ten searchers were each given five different TREC topics and asked to provide a filtering query for 

each topic. They were analysed both on performance   how well they did in providing an effective 

query   but also on how they utilised new features such as automatic RF and special operators (such as 

synonym operators) provided by Inquery, [CCH92], the underlying retrieval system.  

The results showed that users can make use of the new features but that the take up rate was variable.  

Some searchers used the new features from the beginning, and used them effectively. Other searchers 

learned to use the features, what Cool et al classified as   combining old search strategies with new 

ones . Some searchers only attempted to use their existing search strategies in the new environment. The 

indications from this experiment is that the best match systems, which offer relatively weak indications 

of how they should be used

34

, may need to offer users more explanation on how they operate. This 

latter conclusion was also demonstrated in [WRJ02], who indicated that users may be unaware of basic 

best match principles, such as ranking of documents, when interacting with IR systems. Additional 

general conclusions from [CPB+96] were that interactive searching seemed to provide worse results 

and that users may need a mental model of how the system works to use it effectively. However, as 

reported by Belkin, [Bel97], even if people understand RF conceptually they can have difficulty in 

controlling it in operational systems. 

Experience on individual IR systems is important, general experience with any kind of IR system is also 

important. Hsieh Yee, [HY93], investigated the effect of subject searching experience and topic 

familiarity in interactive searching with particular reference to how searchers selected search terms.  

Her results indicate that experienced users   those with more than one year s searching experience, or 

who have attended a course in on line searching   differed in two ways from novice users.  

Firstly, experienced users were more flexible in their search strategies than novice users. Measuring the 

strategies used by the searches, using Bates's [Bat90] categorisation of search tactics, section 7.1, 

Hsieh Yee noted that novice users were more consistent in their search strategies whether the search 

topic was familiar or unfamiliar. However, experienced searchers were more likely to use different 

strategies according to how familiar they were with the search topic. Secondly, experienced and novice 

searchers selected terms differently. Experiences searchers used more synonyms and concentration on 

combining search terms than novice users. When searching on an unfamiliar topic novice users 

depended more on their own search terms, whereas experienced searchers used tended to use more 

thesarual terms, prepared term selection more heavily and spent more time preparing a search.  

The major conclusion for IR from this study is that the user's experience level has a strong affect on 

how a user searches. A particular conclusion for RF is that novice and experienced users may require 

different methods of selecting expansion terms or may require query expansion to be described in a 

different manner. We shall return to factors that affect a user's selection of terms in section 7.3. 

7.2 User characteristics 

We should also consider the characteristics of searchers. Borgman, [Borg89], reviewed a range of user 

characteristics that may play a role in determining the success or failure of online searching in IR 

systems. Borgman s analysis concentrated on Boolean searching but a number of the aspects she 

examined such as technical aptitude, educational background, personality type and the retrieval task 

will be pertinent to all interactive searching.  

Other individual aspects that affect searching behaviour include the task the user is trying to achieve 

[VH00], and the searchers professional discipline [Fid91]. Heuer, [Heu99], also suggests that people in 

different domains use information differently. In addition, Heuer suggests that people often want better 

quality information rather than simply more information so adequate techniques to cut down the amount 

of potentially relevant information may be important.  

Peters, [Pet89], examined the transaction logs of an OPAC library system to classify searches in which 

no results were obtained. A high proportion of the searches (39%) were due to documents not being in 

34

 Although the Boolean model may be more difficult to use, the fact that it forces the user to structure their 

queries may actually make it easier to understand how to interact with the system.  

 41