Intertemporal choice for browsing in informational retrieval

Azman, Azreen (2007) Intertemporal choice for browsing in informational retrieval. PhD thesis, University of Glasgow.

Full text available as:
[img]
Preview
PDF
Download (10MB) | Preview

Abstract

Browsing is an important strategy for information seeking in information retrieval (IR). Usually, browsing is guided by the information need of a user, where the documents are chosen by anticipating whether they could satisfy the user's information need. Therefore, the effectiveness of browsing depends on the ability of the user to make the right decision. However, user is unfamiliar with the document collection and the models underlying the IR system. Due to this limitation, the user is unlikely to make an optimal decision for his/her browsing strategy. Such a problem can be overcome by incorporating a recommendation model to suggest a good browsing strategy for the user. A good recommendation model should be based on modelling the decision behaviour of the user. However, modelling such behaviour is problematic. In this thesis, the intertemporal choice model is adopted to model the browsing behaviour of the user. It is based on the assumption that browsing is an intertemporal choice problem. The effectiveness of modelling the browsing behaviour of the users is evaluated in the context of browsing on mobile devices and post retrieval browsing. First, an implicit RF system is proposed for mobile devices to overcome the limitations of the devices, namely the small screen size and the limited interaction capability. A number of implicit RF models and display strategies are investigated to find the optimal setting for the system. The results suggest that the implicit RF system can be effective provided that an effective browsing recommendation model is incorporated. For this purpose, a recommendation system based on the intertemporal choice model is proposed. The effectiveness of the model is measured by the median average precision (MAP) and the expected search length (ESL) to measure the cost of browsing of the recommended browsing strategy. Second, the effectiveness of the model is evaluated for post retrieval browsing in the context of the subtopic relevance retrieval application. Post retrieval browsing refers to the sequential assessment of the top retrieved documents. In this context, a topic consists of a set of subtopics and a document can be relevant to one and up to all subtopics. The aim of the model is to produce a ranking such that it will take as little as possible to cover all relevant subtopics by browsing through the ranks of the documents. The results from both evaluations suggest that the intertemporal choice model could be effective provided that the parameters associated with the model are optimised and the value of the documents used in the model is as accurate as possible.

Item Type: Thesis (PhD)
Qualification Level: Doctoral
Additional Information: Adviser: Iadh Ounis
Keywords: Computer science, Information technology
Date of Award: 2007
Depositing User: Enlighten Team
Unique ID: glathesis:2007-73995
Copyright: Copyright of this thesis is held by the author.
Date Deposited: 23 Sep 2019 15:33
Last Modified: 23 Sep 2019 15:33
URI: http://theses.gla.ac.uk/id/eprint/73995

Actions (login required)

View Item View Item

Downloads

Downloads per month over past year