eScholar

eScholar stores, preserves and disseminates digital copies of the research and scholarly output of eScholar faculty, researchers and students. These can include the following items:

  • Monographs
  • Pre- and post-prints of academic journal articles
  • Theses and dissertations
  • Major projects and papers
  • Reports/working papers and conference proceedings

Materials in eScholar are openly available to the world and discoverable through search engines such as Google Scholar. This high visibility, discoverability, and exposure can lead to increased citation. Contact Library Publishing for more information.

 

Communities in eScholar

Select a community to browse its collections.

Now showing 1 - 5 of 9

Recent Submissions

Item
Contextual topics: advancing text segmentation through pre-trained models and contextual keywords
(2024-09-01) Maraj, Amit; Vargas Martin, Miguel; Makrehchi, Masoud
Text Segmentation (TS) is a Natural Language Processing based task that is aimed to divide paragraphs and bodies of text into topical, semantically aligned blocks of text. This can play an important role in creating structured, searchable text-based representations after digitizing paper-based documents. Traditionally, TS has been approached with sub-optimal feature engineering efforts and heuristic modelling. In this work, we explore novel supervised training procedures with a labeled text corpus along with a neural Deep Learning model for improved predictions. Results are evaluated with the Pĸ and WindowDiff metrics and show performance improvements beyond any previous unsupervised TS systems evaluated on similar datasets. The proposed system utilizes Bidirectional Encoder Representations from Transformers (BERT) as an encoding mechanism, which feeds to several downstream layers with a final classification output layer, and even shows promise for improved results with future iterations of BERT. It is also found that infusing sentence embeddings with unsupervised features, such as the ones gathered from Latent Dirichlet Allocation (LDA), provides comparable results to current state-of-the-art (SOTA) TS systems. In addition to this, unsupervised features derived from LDA give the proposed system the ability to generalize better than previous supervised systems in the space. Furthermore, it is shown that with the use of novel language models such as Generative Pre-trained Transformers (GPT) for text augmentation, training data can be multiplied, while continuing to see performance improvements. Although the proposed systems are supervised in nature, they have the capability of fine-tuning a threshold variable that allows the system to predict segments more frequently or sparingly, further bolstering the practical usability of it. Due to the increasing competition in the supervised TS space, creating competitive systems often see contributions from larger research companies with more available resources (e.g., Google, Meta, etc.). However, unsupervised TS has been relatively unexplored in comparison with supervised efforts, since it is much more challenging to build a generalizable TS system. To this end, strong word and sentence embeddings are used to create an unsupervised TS system called “Coherence”, that blends the best of pre-trained models and unsupervised features to create a system that is capable of generalizing across various datasets, while achieving competitive results in the space. Since Coherence is unsupervised, inference is quick and requires no upfront investment (i.e., this technique can be picked up and applied to a domain without the need for fine-tuning).
Item
Development of an AI-driven robotic manipulation framework
(2024-09-01) Liu, Xiaolong; Ren, Jing; Lang, Haoxiang
This paper introduces a manipulator control system that leverages state-of-the-art artificial intelligence (AI), advanced robotics, and the latest computer vision techniques to enable intuitive and universal robotic control across different robotic manipulators. The intuitive user command interpretation is deployed in the system using ChatGPT API. The vision capabilities of the system rely on YOLOv7 for object detection, Semi-Global Block Matching (SGBM) for object localization, and Principal Component Analysis (PCA) for object orientation. Moreover, MoveIt is utilized for motion planning, and the simulation environment is powered by Gazebo. These components are seamlessly coordinated by a custom state machine integrating into a ROS-based framework. This system allows users to control manipulators completing complex tasks by commands without expert knowledge of robotics. Additionally, a cross-platform kinematics solver architecture is introduced to assess inverse kinematics algorithms for manipulators. Compared to the traditional kinematics solver implemented as ROS plugins, this architecture integrates the sophisticated mathematical capabilities of Matlab with the feasibility of ROS.
Item
Mothers of children with autism: current challenges and intricate dynamics influencing the quality of life and mental health of mothers caring for children with autism
(2024-09-01) Kokkoros, Peter; Alvi, Shahid
This study investigates the social, personal and institutional challenges faced by mothers of children with Autism Spectrum Disorder (ASD) from their perspectives, aiming to understand these challenges, their strategies for negotiating these challenges and the impact on their mental health and quality of life. Employing a qualitative approach, the study utilizes Interpretative Phenomenological Analysis (IPA) and Young's (1999) framework on social inclusion and exclusion to contextualize societal attitudes towards autism over time. The analysis revealed four main themes: high caregiver burden, stress from the emotional toll of advocating for their child, difficulties in accessing resources, and a lack of coping strategies for dealing with the diagnosis. Despite increased awareness and advocacy for autism, mothers still experience significant stress and life disruptions due to societal acceptance issues and resource limitations. The findings highlight the need for a deeper understanding of caregivers' daily struggles to inform more effective support strategies and emphasize the necessity for tailored support systems to address these unique challenges.
Item
Perspectives on dementia education: a phenomenological investigation in institutional and community-based settings
(2024-10-01) Gabel, Glory; Sun, Winnie
Introduction: This study explores the factors influencing the knowledge uptake of dementia education among health care workers (HCWs) in institutional and community-based settings. It also examines the current barriers and facilitators within dementia education programs. Methods: An interpretive phenomenological approach using in-depth interviews with a diverse sample HCWs (n=10) from institutional and community-based settings. Braun and Clarke’s (2006) thematic analysis was used to identify the main themes and sub-themes related to HCWs’ experiences with dementia education and working with people with dementia (PWD). Results: Four major themes were identified from the interview data, including learning styles of HCWs; facilitators to dementia education; barriers to dementia education and future recommendations. Conclusion: Findings will guide improvements in the training of HCWs by creating engaging and specialized needs-based dementia education programs to enhance knowledge uptake and application of knowledge in healthcare settings, and ultimately address the diverse needs of PWD.
Item
Optimizing training designs and elevating reporting standards in simulation-based medical education
(2024-09-01) Elliott, Leanne K.; Wattie, Nick
Introduction: This thesis explores key elements of simulation-based education for healthcare professionals, focusing on protocol reporting, technical skill retention/transfer, and training microstructures. Study 1 developed and assessed new protocol reporting and evaluation tools (PRT and PET) to measure the quality of reporting (QOR) in simulation studies. Using a modified Delphi method, the PRT and PET were created and applied alongside the TIDieR Checklist and CONSORT Statement to evaluate 17 randomized controlled trials. Results revealed significant differences in QOR scores across the tools, highlighting strengths and areas for improvement in reporting practices. The PRT and PET have the potential to enhance QOR, enable accurate study replication, and assist in the identification of optimal training designs. Study 2 examined the long-term impact of a simulation-based mastery learning (SBML) curriculum versus a competency-based curriculum on pediatric emergency medicine (PEM) physicians’ video laryngoscopy (VL) skill retention and transfer six months post-training. A multidisciplinary panel set a minimum passing score (MPS) of 32/36 (89%) using the Mastery Angoff method. The mastery group outperformed the competency group in skill retention and transfer, suggesting that SBML better sustains VL skills amongst PEM physicians. Study 3 analyzed the microstructures of competency-based and mastery-based training interventions for VL. Behavior coding software determined that mastery learners took longer to reach their MPS, engaged in more partial versus whole practice trials, spent more time verbalizing preparatory steps and aftercare plans, and received more feedback from their instructor compared to learners in the competency group. These differences likely contributed to differences in VL skill learning, as observed in Study 2. Conclusions: This thesis emphasizes the importance of the microstructure in simulation-based training interventions. Detailed and accurate reporting and evaluation of training microstructures is necessary for the advancement of the field, and the newly developed PRT and PET provide a means to do so. The microstructure of the SBML intervention more effectively facilitated VL skill learning, compared to the competency-based intervention. These findings offer valuable insights for educators, researchers, and program facilitators in medical education.