• 검색 결과가 없습니다.

Chapter 2 Literature Review

2.2 Interactive Information Retrieval (IR)

2.2.2 Interactive IR Model

This section presents a discussion about several representative interactive IR models. The interactive nature of the users’ information searching behaviour has become a primary focus of the interactive IR models that have been developed since late 1980s. The IR models have an emphasis on the dynamic interaction between the searcher/user, the information need, and the information environment.

2.2.2.1 Bates (1989) Berry-picking Model

Bates (1989) described the online search as an evolving berry-picking process in which the changes in the search strategy were due to the experience of a variety of sources and were the result of new information encountered which provided new ideas and directions to the original search query. Her model illustrates that the

11

results of each search query would provoke cognitive thoughts in the user to make continual judgments regarding the relevance and interoperability as the information was sought and used.

Based on the basic implication that the information searcher is also the information user, the berry-picking model more closely discovers users’ actual information behaviour than previous traditional linear IR models, in that it considers the users’

dynamic and continuous cognitive responses during IR interaction and their effects on the follow-up search queries. Yet Bates’ ideas on the model have never been empirically validated. Thus, whether the users’ cognitive thought is the factor resulting in changes to the search query still remains a question. Additionally, the forms of the users’ cognitive response have not yet been investigated. A major aim of this dissertation is to investigate the interplay of multiple cognitive aspects on the users’ information search behaviour.

2.2.2.2 Ingwersen (1992, 1996) Cognitive IR interaction Model

As one of the earliest IR interaction models, Ingwersen’s (1992, 1996) model was the first to illustrate that a dynamic interaction process occurred at multiple levels within the “cognitive space” of the user and the “information space” of the IR system.

The multiple levels of interaction were said to occur not only between the user and the IR system, but also between the user and the information objects within the system. The model focused more on understanding the actual information system being used during the interactive cognitive processes. Ingwersen’s model provides a way to understand the process of how information is being retrieved and ultimately used.

Ingwersen (1996) claimed that a wide ranging influence of factors should be considered in IR research, such as social environment, IR system, information

12

objects, search intermediary and user. He incorporated all these variables into the notion of polyrepresentation. His models have presented a reasonable synthesis of studies regarding IR interactions with empirical evidence. As Robins (2000) pointed out, however, the problem with Ingwersen’s model lies in determining the way to get input from the users’ cognitive space into the request model builder, since the differences among the four components of user cognitive space put forth by Ingwersen are subtle. The study presented in this dissertation attempts to better understand the components of the users’ cognitive space in Ingwersen’s model.

2.2.2.3 Saracevic (1996) Stratified Interactive IR Model

Conceptually borrowed from human–computer interaction, Saracevic’s (1996) theoretical model described the IR interaction between the user, the IR system and the information objects through the system. His model was also based on the assumption that users interacted with IR systems in order to use information. He originally emphasized that understanding the reason why a user sought out information was an important part of discerning the influencing factors during that interaction.

The stratified model involved three strata of IR interaction:

• a surface level—the interactions between the user and the interface of the IR system.

• a cognitive level—user-made judgments regarding the results given by the system. Both the users’ thinking and system’s information objects were identified as cognitive entities.

• a situational level—a context-driven interaction, influenced by the need for original information and how the user or system might categorise, or even iteratively change the need.

13

In terms of a theoretical framework, it was a comprehensive model covering all of the three IR interaction levels. However, the details on what and how the changes occurred as a process of IR interaction were not fully established.

2.2.2.4 Belkin (1996) Episodic Model of IR Interaction

Belkin is another pioneer who advanced the interactive viewpoint in information retrieval. Belkin’s (1996) episodic model was based on his anomalous states of knowledge (ASK) hypothesis (Belkin, 1980), which modelled a user who turned to the information system with a high level of cognitive uncertainty. The users’ state of knowledge was anomalous. As such, he/she could not adequately present his/her information need to the information system. Belkin considered that the real problem in IR was how to represent the users’ anomalous state of knowledge.

Interaction with the information system led to the users’ altered state of knowledge, which enabled him/her to define, reformulate and re-focus the information need, and eventually to contribute to his/her underlying information problem solving.

Although the model provides a research framework for interactive IR, it lacks a treatment of the social/contextual facets of the user information problem and the corresponding effects on IR interactions.

2.2.2.5 Spink (1997) Interactive Feedback and Search Process Model

While focusing on understanding how the interactive process actually took place, Spink’s search process model (1997) was developed based on the empirical research. User judgments, search strategies and the interactive feedback loops within the search process were presented. The model reflected that a variety of feedback mechanisms were the major influencing factors in the interactive IR process.

14

Importantly, the model demonstrated that a users’ interaction with the system could consist of multiple feedback transactions, leading to additional inputs or queries which could in turn result in different feedback and new inputs. The strength of Spink’s model is that it observed IR from an interactive point of view and, in particular, that it provided a complete investigation of the feedback mechanisms.

Yet the feedback loops were identified within the discussions between the user and the search intermediary. That is to say, in her model, a feedback loop was incurred when one of the participants gave feedback to the other, followed by a judgment or an action taken. The recognition of feedback was based on analysis of the discourses between the two participants. Apparently, it was not applicable under a Web searching context in which the searcher is the actual user. Another weakness of the model is that it lacks appropriate explanations about the underlying cognitive changes during the occurrence of feedback loops. The present study seeks to address both of these issues.

The above interactive IR models have laid the foundation for the development of later Web search models, which are discussed in the following section.

15