Birmingham City University Open Access Repository
Not a member yet
8823 research outputs found
Sort by
Advanced Development of a Customer Assistance Chatbot using Multimodal Large Language Models to Enhance User Experience
No full textThe quality of online customer assistance plays an essential role in shaping the overall user experience across various industries. As customer bases grow, the volume of queries and usage issues also increases, necessitating ef-ficient and cost-effective solutions to manage these demands. The rise of Large Language Models (LLMs) has made automating chatbot solutions es-sential for delivering effective services, such as technical assistance, trouble-shooting, and personalised recommendations. With the abilities of Retrieval-Augmented Generation (RAG) architectures, applying LLMs to domain-specific tasks requiring specialised knowledge has become increasingly im-portant. In this paper, we experiment with different approaches to develop a multimodal chatbot capable of responding to user queries in various formats, including text, images, tables, and audio, across multiple languages. To ad-dress the limitations of LLMs in providing accurate answers related to private knowledge sectors, as well as the constraints of traditional RAG models in retrieving information from general knowledge on the web that is not includ-ed in the private knowledge base, to tackle this this study presents a multi-agent approach. The chatbot leverages private domain knowledge bases while maintaining the ability to access web-based resources for out-of-domain queries using a multi-agent paradigm. Our study evaluates the impact of different vector stores, embedding models, and chunking parameters on model performance. To refine our results, the evaluation was conducted in two stages, narrowing the focus to the optimal model parameters that achieved higher accuracy. Performance was assessed using RAGAS and hit image rate evaluation methods, ensuring robust measurement of retrieval ac-curacy. This study demonstrates the applicability of the proposed approach to diverse datasets across various domains, effectively addressing business-specific queries. To produce a smarter, and more capable customised AI chatbot assistant that enhances the overall user experience
Theatre Censorship in Restoration London: the case of Charles Killigrew, Master of the Revels
No full text
International Internships: Preparing Students for Rights and Justice in Action
No full textThis chapter explores intercultural competency as a critical skill for law students, using Birmingham City University’s American Legal Practice module as a case study. Part I outlines the American Legal Practice model for context. Part II links culture and intercultural competence with legal education. Part III introduces strategies to foster students' intercultural competency, categorised as student-led, module design, and institutional approaches. Infused with examples and reflections from American Legal Practice, the chapter offers educators practical insights for applying this field of research
A systematic review of circular economy literature in healthcare: Transitioning from a ‘post-waste’ approach to sustainability
No full textThe healthcare sector generates significant waste and environmental challenges, making the adoption of circular economy principles increasingly urgent. While research on circular economy adoption in healthcare has grown, a critical gap exists in understanding how the field evolves and the primary themes driving this transformation. This study employs bibliometric and content analysis to systematically review the intellectual structure and current state of circular economy research in healthcare.
Analysing academic papers published between 2014 and 2024, this review identifies four key themes: healthcare waste management, sustainable product design, economic and policy frameworks, and education and stakeholder engagement. The findings highlight an imbalance in the research landscape, emphasising operational challenges strongly, while systemic enablers remain largely underexplored. Although Europe leads in adopting CE practices, significant research gaps persist in Asia and North America. Key barriers, including regulatory constraints, resistance to change, and concerns around patient safety, continue to impede the effective implementation of CE, particularly in the reuse and recycling of medical devices.
This study proposes research on three key areas: evaluating the impact of the existing economic and policy frameworks; sustainability education, aimed at embedding circular economy principles into healthcare training programs; and operationalising circular supply chains, focusing on reverse logistics for medical device recovery and recycling.
This review contributes to Step 1 of circularity by addressing waste minimisation at the source. It also identifies gaps in research and geographic disparities to advance Step 2 of circularity, which is focused on resource recovery and reuse. Finally, it provides actionable recommendations for Step 3, which aims to build systemic resilience and reduce carbon footprints through circular supply chains and sustainable procurement
Impact of different annealing methods on MEX-printed polyetherketoneketone parts
No full textMaterial extrusion is an established category of additive manufacturing, primarily using thermoplastic materials for product development. The continuous quest for improved material properties and superior surface finish has driven significant interest in the development and application of advanced post-processing techniques. In this work, Antero 800NA has been used, which is a Polyetherketoneketone thermoplastic of the polyaryletherketone family, with excellent mechanical and thermal properties as well as high chemical resistance. To enhance the performance of Antero 800NA parts, three annealing methods were used in this work. They include oven annealing, fluidised bed annealing, and metal plate annealing. Temperatures of 160 °C, 180 °C, and 200 °C were used for three different time intervals of 1, 2, and 3 h. Metal plate annealing showed overall better dimensional integrity and surface finish compared to oven and sand annealing. However, sand annealing showed reduced dimensional variations at the 2-h annealing time-interval for all three temperatures. Tensile testing showed varying results with sand annealing, oven annealing, and metal plate annealing showing higher tensile strength at 160 °C, 180 °C, and 200 °C, respectively. Microstructural analysis showed a compact and cohesive structure with finer microvoids for high tensile strength specimens and larger voids as well as delamination for low tensile strength Antero 800NA specimens. Metal plate annealing also demonstrated higher hardness values, whereas oven and sand annealing showed consistent but lower values compared to metal plate annealing. These results provide a comprehensive comparison of three distinct annealing methods for Antero 800NA parts that can offer valuable insights into optimising post-processing techniques for engineering applications requiring enhanced performance of additive manufactured Antero 800NA parts
Biometrics to Necrometrics: What the dead can tell us about war. A human security approach on collecting and analysing conflict data from the dead
No full textThis paper discusses the use of biometrics on the dead of conflict from a human security perspective. Human security interprets conflict and the effects of conflict from a human point of view. Data extracted from the bodies of the dead aid to provide valuable information to assess the human security situation and are therefore vitally important. This also has a legal aspect to it. Which obligations does international law set on the use of metrics from the dead in conflict? The project Iraq Body Count demonstrates in clear terms the importance of taking the human measure into account
Artificial Intelligence in Pharmacovigilance: Leadership for Ethical AI Integration and Human-AI Collaboration in the Pharmaceutical Industry
No full textPurpose
Pharmacovigilance plays a vital role in ensuring medication and vaccine safety, yet it faces persistent challenges, including underreporting, resource-intensive processes, and regulatory complexities. Artificial intelligence has the potential to enhance efficiency, but its adoption requires strategic leadership to navigate automation feasibility, ethical dilemmas, and socio-economic implications.
Design/methodology/approach
This study uses a systematic review with bibliometric and content analysis to address three core questions: the current state of artificial intelligence in pharmacovigilance, the feasibility of full automation, and the ethical dilemmas associated with its adoption. It explores six themes, including explainable AI, effectiveness, predictive applications, social media-based detection, challenges, and models used.
Findings
The findings reveal the growing use of AI, especially machine learning and natural language processing, to improve adverse drug reaction detection and streamline pharmacovigilance. Yet, full automation faces barriers like privacy concerns, regulatory gaps, and data biases. A strategic leadership approach, integrating AI-driven efficiency with human expertise, is essential to maintaining patient safety and public trust. Ethical concerns, including transparency, accountability, and fairness, must be addressed through responsible AI governance frameworks.
Research limitations/implications
The rapid evolution of AI technologies and regulatory frameworks means new insights are increasingly available. Future research should explore leadership strategies, regulatory adaptations, and governance models that ensure ethical and practical AI adoption in pharmacovigilance.
Practical implications
This study offers practical guidance for pharmaceutical companies, regulators, and third-party organisations to integrate artificial intelligence responsibly in pharmacovigilance. It highlights the role of leadership in delivering ethical AI adoption, shaping policy frameworks, and ensuring a balanced approach between technological innovation and human oversight in drug safety management.
Social implications
This study has significant social implications, particularly in enhancing patient safety, improving public trust in drug monitoring systems, and addressing health disparities. Identified challenges such as data privacy concerns, algorithmic biases, and regulatory gaps must be addressed to prevent AI-driven inequities in healthcare.
Originality/value
Unlike existing reviews that primarily focus on technological advancements or regulatory challenges, this research highlights the critical role of leadership in shaping ethical AI adoption and policy frameworks and balancing automation with human oversight. The findings will be valuable for policymakers, industry leaders, and regulators seeking to implement AI responsibly while maintaining trust and compliance in pharmaceutical safety management
Acoustic Information Retrieval for Interactive Sound Rendering in Virtual Environments
No full textThe adoption of Extended Reality across industry and research domains has incentivised the development of Head-Mounted Display (HMD) technology, driving the field towards better and more optimal techniques for efficient and realistic rendering, offering sensing and capturing capabilities. Virtual entities visualised through HMDs can interact with spatial features of the users’ surroundings, allowing for realistic context-aware interactions and improving task performance and the perceptual quality of the overall immersive experience. However, techniques for rendering realistic audio stimuli that respond to spatial features of the immersive environment are underrepresented, considering an extensive body of literature on Mixed Reality (MR) research domains. Perceptually valid sound rendering is key to realism in audio stimuli within immersive environments, as spatial features of the users’ surroundings can be considered, approximating fundamental characteristics of sound propagation in environments. This enables listeners to use natural hearing abilities that interpret sound propagation effects to sense space and entities in their proximities, affecting interactions in immersive experiences. This thesis reviews the current state of sound rendering techniques and their application and feasibility across several use cases, proposing, as a novel contribution, a pipeline that can generate context-aware realistic audio for MR applications. The development of this pipeline involves adopting computer vision techniques in the process of decomposing complex scenes to recognise acoustic characteristics of space, determining physical and structural features of the environment surrounding HMD users, and allowing audio stimuli to respond to spatial characteristics of the immersive environment. The experiments presented demonstrate applications of scene understanding techniques applied to virtual environments and reconstructions of real space to determine acoustic properties of surfaces and entities for automating the application of sound rendering. This is done by identifying the current state of automatic acoustic material recognition for virtual environments and proposing novel evaluation methods that test the efficacy of automatic systems for tagging acoustic materials in virtual environments. Proof-of-concept systems have been tested on state-of-the-art acoustic renderers to demonstrate their efficiency in real-world scenes. Participant testing using a prototype deployment of the proposed pipeline that measures the performance of psychoacoustics-related tasks suggests that audio stimuli generated using the proposed pipeline have a significant effect on task performance within Mixed Reality applications. Current directions are aimed at designing end-to-end pipelines for interactive, real-time applica-tions, with the ambition of adopting computer vision to understand the acoustic space, even in contexts of dynamic geometry typical of HMD technology, where the acoustic space is constantly updating based on the users’ surroundings