Investigation of quantum-inspired modelling in interactive search based on information foraging theory

Abstract

This thesis investigates the use of the mathematical formalism of quantum mechanics formodelling users' information needs from the viewpoint of Information Foraging Theory(IFT). IFT has been successfully applied to model user behaviours and preferences in theinformation retrieval (IR) process, which motivates this work that hypothesises IFT canenhance user interaction and the effectiveness of typical IR and recommendation tasks, suchas multimodal query auto-completion and image recommendation. During interactive IRsessions, users' information needs often evolve, which requires more system assistance andsupport to capture these dynamics. The users' information needs in such an interactive sessioncompared to the typical unambiguous query terms tend to be multi-semantic and heuristic inthe way of natural languages. In an effort to solve this problem, an interactive multimedia andmultimodal IR system is developed based on a quantum-inspired mathematical frameworkutilising Hilbert spaces. Based on IFT, the key methodology involves characterising the users'multimodal information needs. The users' multimodal information needs are integrated intothe IR system through a projective transformation that follows mathematical constructs ofthe quantum probabilistic framework. The proposed quantum-inspired interactive frameworkis evaluated through an image retrieval task, which allows a multi-iteration of textual queriesfor finding specific images in a search session supported by auto query completion andvisual query cues. Our main findings are: in an interactive multimodal IR context, multisemanticqueries effectively help to confirm users' information needs in a session; fromthe spatial context of IR, our framework captures the dynamics of evolving informationneeds and reflects the historical interaction of the multimodal IR process. This dissertation,therefore, provides comprehensive insights and findings about the usage of the quantumprobabilistic framework in interactive IR. It exhibits the usefulness of the quantum-inspiredIR framework to fine-grained user aspects based on IFT, enhancing the representation andmodelling constructs to explicit the information retrieval behaviours and preferences.

Date of Award	Jun 2023
Original language	English
Awarding Institution	University of Bedfordshire
Supervisor	Haiming Liu (Supervisor) & Ingo Frommholz (Second supervisor)