Proceedings of the Human Language Technology Conference of the North American Chapter of the ACL, pages 247–249,
New York, June 2006. c©2006 Association for Computational Linguistics
Supporting Multiple Information-Seeking Strategies  
in a Single System Framework 
 
Xiaojun Yuan 
School of Communication, Information and Library Studies 
Rutgers, The State University of New Jersey 
New Brunswick, NJ 08901, USA 
yuanxj@rci.rutgers.edu
 
Abstract 
This study investigates the support of 
multiple information seeking strategies 
(ISSs) within a single system, and the 
relation between varieties of ISSs and 
system design. It proposes to construct 
and evaluate an interactive information 
retrieval system which can adaptively 
support multiple ISSs, and allow change 
from one ISS to another within an ISS 
space. It is conducted in a series of steps: 
iterative designing -evaluating of several 
systems supporting different ISSs; 
specifying an interaction structure for 
multiple ISSs; and, implementing and 
evaluating a dynamically adaptive system 
supporting multiple ISSs. The study aims 
to make a contribution to interactive 
information retrieval drawing attention to 
user interface design, and to HCI, in 
integration of multiple support techniques 
within a single system framework. 
 
Keywords 
Information-seeking strategy, interaction 
structure, user interface design, evaluation, 
information retrieval 
 
1   Introduction 
 
Traditional information retrieval (IR) systems ask 
people to specify queries using terms to select 
documents from the selected databases. Current 
research indicates some problems of such systems. 
(1) Human information-seeking behavior is more 
complicated than just query formulation and term 
selection. For instance, people need to get an idea 
of which domain or genre of information they 
need, then familiarize themselves with the content 
and structure of various databases. They need to 
learn about a domain of interest and extend their 
knowledge of this domain so that they can 
formulate more effective queries. They need to 
change their searching and seeking strategies to 
improve their queries. (2) Human information-
seeking behaviors are not discrete processes. These 
behaviors interact with one another during 
information-seeking processes (Belkin, 1996). (3) 
Interaction, not representation or comparison, is 
the central process of IR (Belkin, 1993). (4) Users 
with different goals need to use different 
information-seeking strategies (ISSs) to conduct 
searches. However, traditional IR systems only 
support one ISS, that is, formulating queries using 
terms to select documents from some databases 
(Belkin, 1993, 1995; Xie, 2000). (5) It is known 
that different ISSs can be optimally supported by 
different combinations of IR techniques (Oddy, 
1977). The existing diversity of ISSs indicates that 
a system which provides good support for one ISS 
is unlikely to provide good support for the others. 
A system trying to support all ISSs with only one 
technique will be able to support them at only 
mediocre levels. 
Therefore, the goal of the research is to design an 
interactive IR system which incorporates different 
IR techniques to adaptively support different ISSs. 
Specifically, our solution to these problems 
focuses mainly on the following two goals. 
(1) Constructing a single IR system in which 
support techniques are explicitly represented 
so that it is possible to shift from one 
combination of support techniques to another 
in real time, and appropriate support 
techniques are suggested to the user by the 
system. 
(2) Evaluating the effectiveness and usability of 
the system within controlled experiments. 
 
247
2   Research Problems 
 
We aim to investigate the following research 
problems: 
(1) Implementing and evaluating several systems 
which are tailored to scanning or searching. 
(2) Developing a structure for guiding and 
controlling sequences of different support 
techniques. 
(3) Constructing and evaluating a single system 
which supports scanning and searching 
through integration of different support 
techniques within a single system framework. 
 
3   Methodology 
 
3.1   Research Problem 1 
 
SYSTEMS 
 
Using the Lemur toolkit (LEMUR), we 
implemented and evaluated several different 
prototype IR systems designed to support scanning 
(situation 1) or searching (situation 2). Table 1 
describes the tasks, as well as features and support 
techniques for each system. 
 
HYPOTHESES 
 
Hypothesis 1: The system summarizing each 
database is more effective in supporting scanning 
tasks than the baseline system which provides a 
ranked list of documents with descriptions about 
which databases these documents are in. 
(E1.1/B1.1, situation1-task1) 
 
Hypothesis 2: The system providing table of 
contents navigation is more effective in supporting 
scanning tasks than the baseline system which 
lists ranked paragraphs. (E1.2/B1.2, situation1- 
task2) 
 
Hypothesis 3: The system presenting clustered 
retrieval results is more effective in supporting 
searching tasks than the baseline system which 
presents a ranked list of retrieval results. 
(E2.1/B2.1, situation2-task1) 
 
Hypothesis 4: The system supporting fielded 
queries is more effective in supporting searching 
tasks than the baseline system which provides a 
generic query search. (E2.2/B2.2, situation2- 
task2) 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Table 1. Situations, Tasks and Systems 
 
 
EXPERIMENTAL DESIGN 
 
Participants conducted four searches on four 
different topics that are suitable for scanning or 
searching. This is a within-subjects design. Each 
subject searched half of the topics in one system, 
then half of the topics in the other system. Within 
the topic block, the topic order was randomly 
assigned. No two subjects used the same order of 
topics and the same order of systems. The 
experiments were replicated by exchanging the 
order of the systems. 
 
TEXT COLLECTIONS 
 
There are two text collections: one is TREC 
HARD 2004 collection (HARD) which is suitable 
for situation1-task1 and situation2-task1, the other 
is a book database which is good for situation1-
task2 and situation2-task2. This database is 
composed of books downloaded from Project 
Gutenberg (Gutenberg). 
 
TASKS 
 
In this study, we used the simulated work task 
situation model (Borlund, 1997) to make subjects’ 
behavior as true-to-life as possible, hoping this will 
Situations Tasks
Experime ntal 
Systems
Baseline 
Systems
Support 
Techniques 
11 E1. B1. 
Scanning Identify 
best 
databases
Alphabetically 
ordered 
databases 
(showing  
summary for 
each)
Ranked 
documents 
(showing name 
of the database 
with the 
document)
2E1.2 B1.2 
Find 
comments 
or 
quotations 
from an 
electronic 
book
Table of 
Contents 
navigation 
within 
documents
Ranked 
paragraphs
21 E2.1 B2.1 
Searching Find 
relevant 
documents
Ranked 
clusters
Ranked 
documents
2E2. B2. 
Find the 
name of an 
electronic 
book
Field search Generic query 
search
Clustered 
retrieval 
results 
Fielded 
query 
Summary of 
each 
database 
Table of 
contents 
navigation 
248
make our results robust. Here is an example for 
situation1-task1. 
 
 
 
 
3.2 Research Problem 2  (Future Work) 
 
In order to guide the presentation of specific 
support techniques during the information seeking 
process, we need to specify an interaction 
structure. This interaction structure is equivalent 
to a dialogue manager, and can be used to control 
the interactions between the system and the user. 
We will employ the idea of interaction structure 
developed in the MERIT system (Belkin, 1995). 
This structure models human-computer interaction 
as dialogues and particular dialogue structures are 
associated with different ISSs. This structure will 
be incorporated into the system at the user 
interface level and act as the dialogue manager. 
 
3.3 Research Problem 3 (Future Work) 
 
SYSTEM 
 
The integrated system will allow the user to use a 
variety of ISSs and to seamlessly switch from one 
ISS to another in the information-seeking process. 
The user will be able to choose which ISS to use 
at any time. ISSs will be classified according to 
the goal of the interaction, the topic or task, and 
the information-seeking stage, etc. The system 
should be able to suggest to the user appropriate 
ISSs at the appropriate times, given the current 
state of the information- seeking process. 
 
HYPOTHESIS 
 
Hypothesis 5: The integrated system purposely 
designed for supporting both scanning and 
searching is more effective in supporting tasks 
requiring scanning and searching than the generic 
baseline system. 
 
EXPERIMENTAL DESIGN 
 
This will be a within-subject experimental design. 
The subjects will search the integrated system and 
then the baseline system. The experiment will be 
replicated by changing the order of the systems. 
 
4 Conclusion 
 
Our aim is to contribute to the field of interactive 
information retrieval drawing attention to the user 
interface design and HCI. The systems in research 
problem 1 have been implemented and the user 
studies were conducted. Future work will focus on 
the interaction structure and construction and 
testing of the integrated system. Through this we 
hope to improve information retrieval and human-
computer interaction. 
 
References 
 
Belkin, N. J., Marchetti, P. G., & Cool, C. 
BRAQUE: Design of an interface to support 
user interaction in information retrieval. 1993. 
Information Processing & management, 29(3): 
325-344. 
Belkin, N.J., Cool, C., Stein, A., Theil, U., Cases, 
scripts and information seeking strategies: on the 
design of interactive information retrieval 
systems. 1995. Expert Systems with Applications, 
9(3): 379-395. 
Belkin, N. J. Intelligent Information Retrieval:  
Whose Intelligence? 1996. In Proceedings of the 
Fifth International Symposium for Information 
Science (ISI-96), 25-31. 
Borlund,P. & Ingwerson, P. The development 
of a method for the evaluation of interactive 
information retrieval systems. 1997. Journal of 
Documentation, 53(3): 225-250. 
GUTENBERG.  http://www.gutenberg.org/ 
HARD.  http://projects.ldc.upenn.edu/HARD/ 
LEMUR.  http://www.lemurproject.org/ 
Oddy, R.N. Information retrieval through 
man-machine dialogue. 1977. Journal of 
Documentation,33(1): 1-14. 
Xie, H. Shifts of interactive intentions and 
information-seeking strategies in interactive 
information retrieval. 2000. Journal of the 
American Society for Information Science, 51(9): 
841-857. 
Topic: As a graduate student, you are 
asked to write an essay about air pollution 
for one of your courses. You are supposed 
to get  information you need from a system 
that is composed of several databases. 
Each database has lots of documents on a 
variety of topics. You believe it would be 
interesting to discover factors that cause air
pollution, but you have no idea which 
databases are good on this topic. 
Task: Please find out which databases are 
good for this particular topic, and rank the 
databases in order of likelihood of being 
good. Put your answer in the given space. 
249
