Thesis
Cortical activity of relevance
- Creator
- Rights statement
- Awarding institution
- University of Strathclyde
- Date of award
- 2023
- Thesis identifier
- T17420
- Person Identifier (Local)
- 201888923
- Qualification Level
- Qualification Name
- Department, School or Faculty
- Abstract
- Despite decades of research, relevance remains a central focus of Information Retrieval (IR) research. Many theoretical approaches in IR assume that relevance is based on the mutual interaction of the system and user [1, 2]. Past studies have mainly focused on the system side, while user-centred studies are more recent and arguably more challenging to conduct due to no universally accepted research methodology nor established relevance definition [3, 4]. Despite many competing theories, researchers in general agree that relevance is an internal and subjective process. Therefore, experimental approaches investigating relevance should consider the underlying physiological, psychological and behavioural mechanisms involved [5]. With the development of brain imaging, a new multidisciplinary research direction (termed NeuraSearch [6]) has begun to investigate user relevance by analysing brain activity. The combination of information science, cognitive science, psychology and neuroscience has provided a unique insight into relevance phenomena and established the foundation for brain imaging research methodology within the IR field. Therefore, this thesis builds upon the successful NeuraSearch framework to gain a better understanding of relevance phenomena from a neuro-cognitive point of view, to test existing relevance theories, and to gain in-depth insight into mental processes that underpin relevance evaluation. To do so, we conducted a user study, during which participants provided relevance assessments in the context of the Question-Answering (Q/A) Task, during assessment with an electroencephalogram (EEG). Collected neurophysiological data were analysed using a data-driven approach, which offers a comprehensive overview of all the neurocognitive elements that play an essential role during relevance assessment. In this thesis, we investigated relevance as a binary (i.e. relevant vs. non relevant) and graded (e.g. highly relevant, low relevant, non relevant) variable. Additionally, we have explored the role of users’ cognitive context (namely the self-perceived knowledge (SPK)) on relevance assessment formation. Using a data-driven approach within the NeuraSearch experimental framework, we present the following research contributions: • By re-visiting binary relevance using a data-driven approach, we have not only confirmed the findings of previous studies but also shed light on previously not reported Event-Related Potential (ERP) component - P100. The data-driven approach has been proven effective in discovering novel ERP phenomenon, which have been shown to modulate early attention allocation [7] (see Chapter 4). • Relevance is a complex and context-dependent. Thus, this research investigated the impact of users’ SPK on binary relevance assessment. The results indicate that the SPK within the relevance context is associated with significant differences in cognitive processing related to attention, semantic integration and categorisation, memory and decision making (see Chapter 5). • So far, brain imaging studies have mainly considered relevance as a binary variable. The research presented in this thesis is the first to investigate relevance granularity. We observed significant differences in ERPs in response to words processed in the context of high-relevance, low-relevance and no-relevance. It is possible that differences in attentional engagement, semantic mismatch (between the question and answer) and memory processing may underpin the electrophysiological responses to the relevance assessment. The results support the concept of graded relevance and knowledge of the electrophysiological modulation to each type of stimulus may help to improve the search system design (see Chapter 6). Overall, presented findings may help to better understand the cognitive levels of individuals and recommend content based on their cognitive abilities, which would lead to an increase in search success. A better understanding of relevance is an important step toward improving personalisation in the IR process.
- Advisor / supervisor
- Moshfeghi, Yashar
- Resource Type
- DOI
Relations
Items
Thumbnail | Title | Date Uploaded | Visibility | Actions |
---|---|---|---|---|
|
File | 2025-06-27 | Private |