13463 research outputs found
Sort by
Language teacher AI literacy: Insights from collaborations with ChatGPT
Get PDFThe transformative potential of generative artificial intelligence (GenAI) in
language education highlights the importance of AI literacy among teachers to ensure its ethical and effective integration into teaching practices. Although studies have examined the application of AI in language education, there is a lack of comprehensive reviews focusing on the interaction between language teachers and ChatGPT, a GenAI
tool, particularly in fostering human–AI collaboration within educational contexts. This review addresses this gap by synthesising findings from empirical studies. The Scopus database was used as the primary source for this review. A total of 19 journal articles, published between 2023 and 2024, were identified. The review first analyses the research participants and research methods of the selected studies. Key themes are organised into five dimensions: AI foundations and applications, AI ethics, a humancentred mindset, AI pedagogy and AI for professional development, which are derived from the framework proposed by Miao and Cukurova (2024. AI competency Framework for teachers. Paris, France: UNESCO). This review adopts their framework as an
analytical lens for evaluating both the opportunities and challenges associated with integrating ChatGPT into language education. The findings highlight the importance of a balanced approach to AI integration to safeguard educational integrity. By offering actionable insights for teachers, curriculum designers and policymakers, this review
presents a roadmap for the responsible adoption of AI in language education, ensuring that teachers remain central to the learning process
Can Work Be Meaningful Under Algorithmic Management? A MacIntyrean Perspective
Get PDFAlgorithmic management is deeply changing the way work is performed and the interaction between managers and workers in organizations. It also heavily affects the conditions for meaningful work highlighted by existing literature. Therefore, organizations need an appropriate framework to enable meaningful work when adopting algorithmic management systems. This article presents a normative study of the conditions for work to be meaningful in this new scenario. To fulfil this purpose, it adopts a MacIntyrean approach, according to which work is meaningful when it embodies practice-like characteristics. The article identifies the main threats of algorithmic management and characterizes the normative conditions organizations should meet to enable meaningful work. In addition, the article explores the strategies of resistance that workers use to live up to the standards of meaningful work when organizations are not capable or willing to provide those conditions
Sustainable human resource management and job satisfaction—Unlocking the power of organizational identification: A cross-cultural perspective from 54 countries
No full textSustainable human resource management is gaining importance in organizations due to its role in developing a sustainable work environment and well-being. This paper discusses the relationship between employee perceptions of sustainable human resource management and job satisfaction in 54 countries. We propose that sustainable HRM is positively associated with job satisfaction but that this relationship is moderated by employees’ identification with
the organization and country-level individualism-collectivism. Thus, we suggest national culture functions as a second-level moderator of the relationship of sustainable HRM with organizational identification on job satisfaction. Findings from the multi-level analyses using data from 14,502 employees nested within 54 countries provided support for our hypotheses, namely that employee perceptions of sustainable HRM were positively associated with job satisfaction and that this relationship was more pronounced for employees with lower levels compared to higher levels of organizational identification in individualistic rather than
collectivistic countries. These findings bear important implications for both theory and practice
FilmAgent: Automating Virtual Film Production through a Multi-Agent Collaborative Framework
Get PDFVirtual film production requires intricate decision-making processes, including scriptwriting, virtual cinematography, and precise actor positioning and actions. Remarkable progress in automated decision-making have utilized agent societies powered by large language models (LLMs). This paper introduces FilmAgent, a novel LLM-based multi-agent collaborative framework designed to automate and streamline the film production process. FilmAgent simulates key crew roles—directors, screenwriters, actors, and cinematographers—within a sandbox environment, integrating efficient human workflows. The process is divided into three stages: planning, scriptwriting, and cinematography. Each stage engages a team of film crews providing iterative feedback, thus verifying intermediate results and reducing errors. Our evaluation of generated videos reveals that collaborative FilmAgent significantly
outperforms individual efforts in line consistency, script coherence,
character actions, and camera settings. Further analysis highlights
the importance of feedback and verification in reducing hallucinations, enhancing script quality, and improving camera choices. We
hope that this project lays the groundwork and shows the potential
of integrating LLMs into creative multimedia task
Learning to evaluate video captioning systems through human assessment
Get PDFMultimodal content analysis has attracted the attention of numerous researchers in the computer vision community. One of the most representative tasks of this sub-field is video captioning. There are numerous difficulties involved in creating these descriptions, ranging from effectively conveying the essence of the scene to generating text that is grammatically accurate and flows smoothly. In this thesis, we examine an inherent issue with these techniques. The question being asked is “How can we effectively evaluate video captioning?”. Our research has identified several shortcomings in the metrics, such as skewness in the length of the sentences or the practice of only measuring textual similarity on a small set of human reference captions. The widely used TRECVid video-to-text task is used as the basis for our investigations. Shortcomings are identified by comparing the correlation of various metrics against the direct assessment. To improve the quality of the evaluation, we propose to fine-tune a large language model to maximise the correlation. Our results show that this metric uses other qualities to evaluate the system output and obtains good performance. This idea is then extended using contrastive learning to learn an embedding space where a more human-like similarity can be used in the evaluation. We show that vision and language models can be used to measure the similarity between visual and textual features. We also identified discrepancies in how human judgment scores were distributed.
Some experiments on reducing bias by collecting multiple scores per caption or filtering outliers were carried out. These finally motivated us to create a new dataset for video captioning evaluation. This new dataset divides the qualitative scores into five sub-aspects and the captions were also post-edited. The final dataset provides the basis for a more comprehensive and robust reference upon which to compare
metrics for video captioning evaluation
The invisible minority: a biographical narrative study of gay men’s stories of intimate partner violence
Get PDFIntimate partner violence (IPV) is a serious social problem. There is limited research on gay men's accounts of abuse. Due to IPV being socially constructed as predominantly occurring within heterosexual relationships, abuse in sexual minority relationships remains concealed
within society, rendering gay men, as victims, an invisible minority. This thesis utilised Biographical Narrative Interpretive Method (BNIM) to interview six gay men and examined how they account for abuse in their life stories. Three cases (Will, George and Sam) were analysed using all ten stages of the BNIM analytic process. The remaining three cases (Tom, James, and Cole) were incorporated using a streamlined narrative analysis methodology.
This study revealed that gay men accounted for abuse individually, experiencing IPV that closely resembled abuse found in heterosexual relationships. This included sexual, physical, financial abuse, controlling behaviours and technology-related abuse. Sexual minority abuse,
distinct from traditional abuse, was also identified. This included outing, encountering heterosexist and hostile attitudes from family members, and restrictions from the LGBTQ community.
This study identified that gay mens' perception of their masculinity shaped the framing of IPV victimisation within their life narratives. Drawing from Connell’s masculinity theory, it was found that discourses of masculinity, femininity and heteronormativity informed how gay men
articulated their public and private accounts of abuse. Participants drew upon desirable attributes and values associated with heteronormative masculinity, mobilising narrative techniques such as minimization, generalisation, and avoidance.
Three narrative strategies were identified which characterised how gay men accounted for their IPV experiences. This included the ‘Fixer Narrative’, ‘Invisibility Narrative’ and ‘Vulnerability Narrative’. These narrative strategies were underpinned by men’s desire to affirm masculinity whilst also concealing vulnerability.
The study findings challenge the assumption that IPV is primarily a phenomenon perpetrated by heterosexual males against female victims. This original study makes visible the hidden issue of IPV within gay men’s relationships. It is hoped that the findings of this study prompt further
investigation into this understudied area
A search engine using graph-based structures for lifelog retrieval
Get PDFLifelog, a personal digital record of daily activities, is becoming popular and can operate as a form of a digital diary. Although having a wide range of applications, the basic underlying challenge of lifelogs is how to build a lifelog retrieval system that can retrieve a specific activity efficiently from large multimodal lifelog data. Recently, there have been many workshops organized to encourage researchers to solve that problem leading to the introduction of many lifelog retrieval systems with different approaches. In addition to the user interface, the backend search engine plays a critical role in the lifelog retrieval system. The fundamental approach for the lifelog search engine is the concept-based model which matches the objects in lifelog images with keywords in a semantic query. However, many state-of-the-art embedding multimodal retrieval models, such as CLIP, have been introduced recently where they encode both images and texts into the same vector space to perform the retrieval. This has opened the opportunity to apply this new generation search engine to the lifelog retrieval task. In this thesis, I propose applying embedding- based models to the lifelog retrieval system. Furthermore, I enhance these models by applying a graph-based structure to them. I also build an interactive lifelog retrieval system that follows the concept-based approach to serve as the baseline in my experiment. Through my interactive user studies, I observe that the graph- based enhanced retrieval model surpasses the conventional concept-based by a large margin in various evaluation metrics. In summary, this thesis primarily contributes to the field of creating an accurate lifelog search engine using graph neural network structures that can be employed to build an effective and efficient lifelog retrieval system to address lifelog retrieval tasks
Photonic and Signal Processing Technologies for High Capacity Short Reach Optical Interconnects
Get PDFThe development of integrated photonic technologies has enabled the miniaturisation and mass production of optical components for fibre-based optical transmission systems. This resulted in an exponential growth in bandwidth availability over the last few decades due to drastic cost reductions and increases in energy efficiency. However, as the world transitions to a carbon-neutral economy, the requirements imposed on the optical network get more and more challenging to meet. The use of intensity modulation with direct detection (IMDD) has historically been the preferred solution for short-reach applications, but coherent technologies have proven in recent years that they can outperform IMDD even after a few tens of km.
Some of the main technologies competing in the short-reach market, ordered in increasing complexity and performance, are directly modulated lasers (DML), externally modulated lasers, and coherent transmission. DMLs provide high power and energy efficiency, but the frequency chirp induced by the modulation makes it challenging to achieve high data rates and transmission distances. External modulation allows for improved performance at the expense of additional insertion losses, while coherent transmission provides great improvements in reach and spectral efficiency at the expense of increased complexity, component cost and energy consumption.
This thesis explores the use of photonic integrated circuits and low-complexity digital signal processing to enable cost-effective and energy-efficient increases in transmission capacity/reach in short-reach optical transmission systems. We study the use of the three technologies mentioned above for different use cases relevant for
current and future optical links ranging from intra-data centre point-to-point links to point-to-multipoint systems for optical access networks
Information Lifelogging: Leveraging Eye Movements and Reading Comprehension for Efficient Retrieval of Previously Encountered On-Screen Information
Get PDFThe progress of lifelog research has enabled individuals to comprehensively capture their daily experiences. As a result, previous studies have primarily focused on developing tools to organise and retrieve lifelog moments effectively. However, existing lifelog data often lacks the ability to capture the lifelogger’s focal points (their attention), despite providing information-rich first-person-view lifelog images of their surroundings and activities. Consequently, this limitation hinders the lifelog retrieval systems’ utility when lifeloggers seek to retrieve specific information they have previously encountered. To address this research gap, a subjective point of view, represented through eye movements, should be incorporated as a new modality into lifelog data, thereby enhancing the retrieval performance. In pursuit of this objective, this dissertation investigates the feasibility of retriving on-screen information by analysing lifelogger’s reading activities and comprehension level.
The primary contributions of this dissertation are as follows. Firstly, the development of LifeSeeker, an advanced interactive lifelog retrieval system, is developed and benchmarked in numerous lifelog retrieval challenges and competitions. By efficiently integrating various modality processing components (e.g., visual, text, location, biometrics) and user interaction components (e.g., search, filtering, browsing, relevance feedback) into a single interactive retrieval framework, LifeSeeker ranked among the top systems in these benchmarking activities, serving as the foundation for the rest of the thesis. Secondly, a novel reading comprehension dataset was created to explore the feasibility of recognising reading activities and estimating reading comprehension levels in daily life. Statistical tests and machine learning analyses on the dataset have revealed the strong connection between eye movement patterns, reading conditions, and reading comprehension. This led to a novel method for estimating reading comprehension with potential real-world applications. Furthermore, the longitudinal aspect of reading comprehension was investigated to examine the stability and generalisation of reading comprehension estimation models over time. Lastly, the reading comprehension estimation model was integrated into LifeSeeker as a new modality processor, resulting in a significant improvement in the system’s overall retrieval performance. In summary, this dissertation contributes to the understanding of reading activities and reading comprehension in real-world settings and showcases the potential of integrating reading comprehension estimation to enhance the retrieval of previously encountered information in lifelog data