Computing and Information Systems - Theses
Permanent URI for this collection
Search Results
1 - 10 of 342
-
ItemAttentional Reality: Understanding and Managing Limited Attentional Resources in Augmented Reality( 2022)Limits of human attention restrict the amount of information we can perceive when using augmented reality (AR) applications. This leads to consequences when the unperceived information, either from the digital content or the real surrounding, is vital for the user's experience or safety. Despite such consequences, it is unclear how attentional resources are allocated in AR applications and what measures can be taken to improve attention management in AR. This thesis aims to better our understanding of attention allocation in AR applications, isolate variables related to AR that demand excessive attentional resources, and develop and evaluate adaptive techniques to improve efficiency of attention allocation in AR. Our findings show that users excessively focus on the digital content in AR at the cost of neglecting information from other sources. We demonstrate how the excessive allocation of attention towards the AR content is related to the task of scanning for and processing task-relevant digital content. Finally, we show how an intelligent adaptive agent, based on the theories of selective attention, can improve attention management in AR but faces challenges when users are less receptive to the agent's support. These findings and the resulting discussions presented in this thesis yield novel insights regarding user attention in AR and provide valuable lessons in designing AR applications.
-
ItemUbiquitous Material Sensing in Everyday Settings Using Miniaturized Near-Infrared Spectroscopy( 2022)Our computing systems are becoming increasingly smart with the growing number of sensors. However, we are yet to have a material sensing method that can be easily integrated into a computing system. For its applications in many fields including healthcare, agriculture, and food computing, there is a high demand to have a ubiquitous material sensing method that is mobile, low-cost, and versatile for various sensing tasks. Conventional material sensing techniques require either expensive equipment or complex procedures with rigorous training, and thus cannot be readily incorporated with a computing system. A promising method to enable ubiquitous material sensing is to utilize the emerging miniaturized Near Infrared Spectroscopy (NIRS) scanners. Nevertheless, existing knowledge and tools are mostly for conventional laboratorial settings, requiring expertise in NIRS and significant efforts to develop new material sensing systems. To alleviate this issue, this thesis aims to enable non-experts, including researchers and developers in various study fields, to utilize NIRS as a ubiquitous material sensing method in everyday settings. We present novel designs and prototypes using miniaturized NIRS that can be deployed in everyday settings for various material sensing tasks. In particular, we demonstrate prototypes for probing liquids such as drinks or alcohols, detecting gluten in bread, and reading through covered contents within paper sheets or 3D printed objects. We also conduct comprehensive experiments to evaluate the performance of our tools. In addition, we investigate design considerations that can impact end users' trust in using our material sensing tools in daily tasks. Our findings provide guidance for designing trustworthy material sensing applications, especially for users who are unfamiliar with the technology. Our work contributes towards establishing a knowledge base for ubiquitous material sensing using miniaturized NIRS. In particular, our results provide references on design, data collection, and evaluation for this emerging study field. Our methods do not require expertise in NIRS, allowing readily and rapidly developing new material sensing applications. Finally, we discuss the future directions toward ubiquitous material sensing in everyday settings. We envision that material sensing is becoming an important tool to significantly improve our understanding of our living context.
-
ItemTowards Multimodal, Interpretable and Robust Task-Oriented Dialogue Modeling( 2022)Over the last decade, task-oriented dialogue system has enabled canonical applications such as Amazon Alexa, Google Home, and Apple Siri to help humans accomplish specific tasks via natural language interfaces. Now while driving, we can simply ask Alexa to book a restaurant and check the weather forecast for us. Traditionally, these systems are built via the pipeline approach where different modules are separately trained and connected to generate responses. However, these systems suffer from the credit assignment issue, and requires labelled data for each component. Recently, end-to-end dialogue models have attracted increasing attention as they alleviate the drawbacks of the pipeline approach by directly mapping the user utterances to system responses. However, most existing end-to-end models leverage memory networks to generate responses, and they don't capture the structural information stored in the external knowledge base. In this thesis, we first propose a novel end-to-end dialogue model that incorporates graph structure from the knowledge base. Observing that existing end-to-end models are limited as it uses a single modality (e.g., text only) knowledge base and multimodal data (e.g., images) can provide more information about the entities of interest (e.g., restaurants), we next introduce a multimodal dialogue system that can leverage a multimodal knowledge base. Nowadays, most end-to-end dialogue models use pre-trained language models as the backbone and although they are found to be empirically effective, their black-box nature means these dialogue systems do not provide explicit reasoning. To this end, we propose a novel explainable neuro-symbolic architecture for improving the interpretability of response generation. Lastly, we found that pre-trained model based dialogue models exploit shortcuts in the training data, learning the dataset instead of the task. To tackle this challenge, we propose debiasing techniques based on contrastive learning and adversarial filtering to improve model robustness. Extensive experimental studies demonstrate the effectiveness and robustness of our dialogue systems.
-
ItemAnaphora Resolution in Procedural Text - from Domain to Domain( 2022)Anaphora is an important and frequent concept in any form of discourse. It describes the use of expressions referring back to expressions used earlier in text, to avoid repetition. Anaphora resolution aims at resolving these reference relations in discourse and forms a core task in natural language understanding. It mainly contains two anaphoric types: coreference and bridging. While much effort has been targeted at anaphora resolution, most research has focused on these two anaphoric types separately. Specifically, anaphora research mostly focuses on coreference, modeling it from different perspectives across various resources. Bridging, on the other hand, has not been studied comprehensively. Different work analyzes bridging differently, leading to inconsistencies in bridging definitions. The lack of attention to bridging also brings challenges in capturing comprehensive anaphora phenomena in discourse -- only modeling coreference is not sufficient to capture complex anaphoric relations in text. It is becoming increasingly important to have both coreference and bridging annotated. Additionally, most existing anaphora research is based on declarative text. Procedural text, a common type of text, has received limited attention despite the richness and importance of anaphora phenomena in it, leaving much room for further exploration. In this thesis, we focus on anaphora resolution in procedural text, studying both coreference and bridging based on two common types of procedural text, chemical patents and recipes, and show that our proposed anaphora frameworks are well suited for procedural text. The four research questions we address in this thesis are: (1) How to model anaphora resolution in chemical patents? (2) How to combine different types of anaphora resolution? (3) How to incorporate external knowledge into anaphora resolution? (4) How to generalize our anaphora resolution model to domains apart from the biochemical domain? We address the first research question by proposing domain-specific anaphora annotation guidelines for chemical patents, targeting both coreference and bridging and incorporating general and domain-specific knowledge via in-depth investigations. We resolve ambiguities in bridging definitions by limiting the anaphoric relations to four specific subtypes related to the chemical domain while maintaining high coverage of anaphora phenomena. We achieve high IAA on the created ChEMU-Ref corpus, well above existing bridging corpora and demonstrating the reliability of the created dataset. To address the second research question, we propose an end-to-end joint training anaphora resolution model for coreference and bridging, adopting an end-to-end coreference resolution framework (Lee et al., 2017, 2018). Through empirical experiments on off-the-shelf anaphora corpora, we show the benefits of joint training for bridging. However, the impact on coreference is not clear. We argue that it could be due to ambiguity in the definition of bridging. To validate our hypothesis, we further experiment on two high-quality anaphora corpora with clear anaphora definitions, the ChEMU-Ref and RecipeRef (details in the last research question) datasets, and show the potential in improving both tasks through joint training, indicating the benefits of joint learning of coreference and bridging on high-quality anaphora corpora. Next, we address the third research question from the perspective of the utilization of pretrained language models based on the proposed end-to-end joint training framework, experimenting on the ChEMU-Ref corpus. We show that even with simple replacements, replacing generic language models (e.g. ELMo (Peters et al., 2018)) with domain pretrained language models (e.g. CHELMO (Zhai et al., 2019)), models achieve better performance, suggesting the potential of incorporating external knowledge for domain-specific anaphora resolution. Further explorations on recurrent neural network based and transformer based language models provide deeper insights, and suggest that different approaches might be needed to fully utilize different types of pretrained language models. For the last research question, we generalize the anaphora annotation framework developed for chemical patents to recipes with domain adjustments by detailed analysis of the similarities and differences between these two types of procedural text. Through in-depth comparison, we propose a more generic anaphora annotation framework for procedural text, designing in a hierarchy based on the state of entities. Based on the proposed annotation framework, we create the RecipeRef corpus, capturing rich anaphora phenomena in recipes, maintaining high IAA scores, and suggesting the feasibility of generalizing this framework to other procedural text. We observe further improvement from transfer learning, i.e. pretraining on the ChEMU-Ref dataset and fine-tuning on the RecipeRef dataset, indicating the transformation of general procedural knowledge in this domain. In summary, this thesis studies anaphora resolution in procedural text, particularly based on chemical patents and recipes, two common types of procedural text, and fills the gap in modeling and resolving anaphora resolution in this area. This establishes a firm base and contributes towards further research in anaphora resolution over procedural text.
-
ItemFrom Discourse and Keyphrases, to Language Modeling in Automatic Summarization( 2022)This thesis aims to enhance single-document automatic summarization by exploring four different spectrums: language model, discourse, keyphrases, and evaluation systems. First, progress on language models and automatic summarization has predominantly been in English, and it leaves open the question of whether they function effectively in other languages. To address this, we perform a case study on the Indonesian language by releasing two pre-trained language models (i.e. IndoBERT, IndoBERTweet) and two large-scale summarization corpora (i.e. Liputan6 and LipKey). While our findings suggest that the current progress indeed effectively works in Indonesian, we found particular challenges in evaluating Indonesian text summarization because of morphological variation, synonyms, and abbreviations in system-generated summaries. Second, modern summarization systems are built on pre-trained language models which serve as the foundation models. However, it is still unclear whether these language models truly learn the summarization task, or they simply memorize the pattern of the input document and human-written summaries. In this thesis, we argue that these language models are still imperfect, and investigate the benefits of discourse information and keyphrases for summarization systems. This is because discourse provides information relating to text organization, while keyphrases capture succinct and salient words about the text. To test this hypothesis, we first perform discourse probing on pre-trained language models to understand the extent to which they capture discourse relations, and introduce a novel approach to discourse parsing - which aims to recover the discourse structure given a document. We then explicitly incorporate discourse and keyphrases into summarization systems and found the qualities of machine-generated summaries improve. Lastly, despite significant progress in the development of summarization models, both automatic and manual evaluations of text summarization are less studied. Reliable and scalable evaluation is critical to measure the research progress in summarization, and ROUGE as the de facto summarization evaluation is inadequate. ROUGE only evaluates summary quality by comparing word overlap between machine-generated and human-written summaries, while broader aspects such as faithfulness (the extent to which the generated summary contains genuine details found in the document) and linguistic quality of summaries (e.g. fluency of the language) are not covered. The last contribution of this thesis is a comprehensive automatic evaluation framework for text summarization compiling prominent aspects used in the manual evaluations of prior works. This proposal is introduced as the FFCI framework that consists of four aspects: faithfulness, focus, coverage, and inter-sentential coherence, and we propose methods to automatically assess summarization quality based on these four aspects.
-
ItemA Person-Centred Information Needs Framework (PcINF) for Hospital Discharge: Exploring personal information needs from hospital to home( 2022)Despite advances in the design and development of healthcare information systems, there is still a lack of adequate process and system support for person-centred discharge. A large body of research highlights the need for tailored and valuable person-centred discharge information that patients can enact and draw on for recovery when they leave the hospital environment. Former studies highlight the need to analyse the information needed for person-centred discharge. More specifically, there is a need to investigate to what extent discharge information needs for a person can be fulfilled from admission through the entire patient journey. Even though several studies focus on the notion of person-centred discharge, there is a lack of theoretical insight on person-centred information needs and the unique attributes of this information. The latter refers to the format, presentation style and time intervals of releasing discharge information to support adaptation during a patient’s recovery process when home. Hence, this study focuses on the critical research question: How can a person-centred information needs framework (PcINF) combined with process and systems thinking improve discharge information? and sub-research questions: i) What are current problems in hospital discharge information? ii) What obstacles prevent a person from utilising hospital discharge information at home? and iii) How can process and systems thinking improve person-centred communication at discharge? In response to these questions and to better understand the problems and challenges associated with person-centred discharge information, this study first developed an integrated person-centred information framework (PcINFv0) by synthesising Roy's (Roy, 1970) Nursing Theory and Adaptive Structuration Theory (AST) (Giddens, 1984). Using a single in-depth case study, this framework was then evaluated through an exploratory multi-method research approach. The approach combined qualitative data collection and analysis methods (medical records and healthcare systems analyses, individual interviews, and a focus group) with business process analysis and process mining techniques. Findings indicate the PcINFv0 is valuable for classifying information relating to lifestyle activities and a person’s social capacity, which impacts adaptation at home. The framework identifies two key aspects: standard information about patient stressors and rules and resources that a person needs to understand during their recovery period at home. In addition, the findings call for attention to the format, presentation style and time-driven release of person-centred discharge information to prevent further readmissions to the hospital. The study is both timely and relevant, particularly in filling the gaps in understanding the problems, challenges and information needs of person-centred discharge information systems. The study highlights the need for a deeper understanding of patient needs in a new era of digitalisation. The findings highlight the need to apply process and systems thinking to the discharge planning and fulfilment process from the perspective of the patient/person’s experience. This calls for more emphasis on the discharge process and the quality and value of the information delivered to the person. This research also paves the way to develop a new generation of innovative and intelligent person-centred discharge systems that empower personal recovery at home.
-
ItemAppropriation of formal and informal learning technology in higher education: the case of Saudi students in home country and in Australia( 2022)Currently, students in higher education utilise a wide range of technologies, some of which have been formally provided and mandated by universities (e.g., learning management system (LMS)) whilst others are more informal and used voluntarily (e.g., social networking sites (SNSs)). Both types of technologies provide many advantages to the learning process, including increased communication with peers and academics and better engagement with learning content. Despite the extensive research on the adoption and use of various formal and informal learning technologies, little is known about how these different platforms are appropriated simultaneously within higher education systems. This qualitative interpretive research is composed of three studies that explore the appropriation of formal and informal learning technologies by Saudi students in higher education in their home country and in Australia. The findings of this research show how technology appropriation patterns of the same technology differ in different educational contexts. Specifically, the thesis explores the transition of students’ attitudes and behaviours towards the appropriation of formal and informal learning technologies based on their physical relocation from Saudi Arabia to Australia. The research provides a new context-oriented model as an extension of Adaptive Structuration Theory (AST) with concepts from networked learning to explain students' appropriation of technologies in their learning interactions. The model describes the appropriation process undertaken when students deal with formal and informal learning technologies while addressing the structures of the technology, the environment and the individual characteristics that influence this process. Moreover, the model shows the emergent social and technological structures in the appropriation process and the intended and unintended learning outcomes. Recommendations are provided based on the research findings for different stakeholders involved in students’ learning experiences, including academics, administrators and policymakers in higher education in both Saudi Arabia and Australia.
-
ItemThe Intersection of Planning and Learning through Cost-to-go Approximations, Imitation and Symbolic Regression( 2022)This thesis explores the intersection between planning and learning methods for autonomous sequential decision-making. Planning is a model-based approach to autonomous sequential decision-making where action policies are derived automatically through a model of an environment. Alternatively, learning methods learn action policies through interaction with an environment. The planning and learning approaches can be likened to current theories of human cognition which propose a fast and associative system works in conjunction with a slow and deliberative one. From this observation previous work has conjectured that in order to create intelligent systems that are more general and robust than existing ones, a combination of planning and learning methods may be required. Two common high-level approaches for combining planning and learning are to use learning to help guide the search effort of planners and to use planners to teach learning algorithms. This thesis examines these two high-level approaches through the topics of cost-to-go approximations, symbolic regression and imitation. We propose and study a number of new algorithms which provide new insights into methods that combine planning and learning, namely, we introduce methods for learning value and policy functions from lookeaheads; learning from single demonstrations produced by planners; and learning heuristics for planning algorithms.
-
ItemA Process Model to Improve Information Security Governance in Organisations( 2022)Information security is an increasingly important topic among senior organisational stakeholders (i.e. the board and executive management) as organisations acknowledge the potential for operational disruption, reputational loss, impact to share value and financial penalties. As information resources are a strategic asset to organisations, there is an expectation that these stakeholders will demonstrate their fiduciary duty of care by implementing information security governance (ISG). Compared to corporate governance, ISG is a relatively new and under-researched area. A review of the literature shows the lack of an ISG framework or model that: (1) incorporates the broad areas of ISG; (2) explains how to implement ISG; (3) is empirically grounded; and (4) identifies the processes required to be undertaken by various stakeholder groups involved in ISG. The practical requirement for an ISG framework or model to help organisations improve their implementation of ISG and the research gaps have led to the following research question: “How can ISG be implemented in organisations?” To address the research question, this research has adopted an exploratory research approach. First, a conceptual ISG process model was proposed based on synthesis of extant literature and detailed review of relevant frameworks and models. The conceptual ISG process model was subsequently refined based on empirical data gathered from 3 case study organisations comprising one financial institution in Singapore and two financial institutions in Malaysia. The refined ISG process model was finally validated in 6 expert interviews. This research addresses the aforementioned practice requirements and research gaps by introducing an empirically grounded ISG process model as a practical reference to facilitate the implementation of ISG in organisations. Specifically, the research contributes by: (1) developing ISG process theory, as ISG is a series of events occurring within an organisational context; and (2) developing an information-processing perspective on ISG, as the process model identifies the information and communication flows, and the relationships among stakeholder groups. In addition, the research has: (3) empirically examined and validated the ISG process model based on how ISG is practised in real-world organisations; (4) examined corporate governance theories to provide additional perspectives to ensure that the ISG process model is aligned with corporate governance objectives; (5) identified additional factors that influence the implementation of ISG requiring further research; and finally (6) expanded existing seminal research by introducing an empirically grounded ISG process model that has been developed based on synthesis of cumulative knowledge from previous research and validated with empirical data. This research is the most comprehensive study to date that has developed an empirically grounded ISG process model identifying stakeholder groups and explaining how core ISG processes and sub-processes interact. An ISG process model is easier to visualise for practitioners and easier to implement as it allows practitioners to structure their thinking according to the stages of the process model and change activities in their organisations.
-
ItemCredibility Assessment of Online Consumer Reviews( 2022)In the digital transformation era, online reviews have become an important source of information for decisions about purchasing products and services. Research shows that online reviews influence users’ purchasing behaviours and product sales. However, the overwhelming number of online reviews with unknown reviewers has made it difficult for users to find credible information. Thus, this thesis focuses on the credibility evaluation of online consumer reviews and develops a comprehensive credibility model to help differentiate reviews based on credibility. We, first, outline a baseline model for credibility assessment of online consumer reviews by building upon related literature and using the Heuristic Systematic Model (HSM) of information processing as the theoretical lens. We extend the baseline model by conducting several in-depth semi-structured interviews as a way of understanding how online shoppers assess the credibility of online reviews. Next, we identify important attributes that impact the credibility of online reviews and explore the moderating role of the reader’s perspective in this process through performing a user study on the Amazon Mechanical Turk platform. Finally, we develop a machine learning model to predict the credibility of online reviews and conduct a series of ANOVA analyses (analysis of variance) to differentiate the characteristics of fake and credible reviews. This thesis advances the state-of-the-art studies regarding the credibility of online consumer reviews by making the following key contributions: (1) synthesizing the related literature and providing a taxonomy of the key attributes that impact the credibility of online consumer reviews, (2) extending the HSM model in the area of the credibility of online reviews and explaining how users assess online reviews, (3) confirming important elements in the HSM including concurrency and complex effects of attributes in information processing, and exploring the moderating role of individuals in the credibility evaluation of online consumer reviews, (4) developing a predictive model for the credibility of online consumer reviews, and; (5) identifying different characteristics of fake and credible reviews.