1,721,039 research outputs found
No More Pencils No More Books: Capabilities of Generative AI on Irish and UK Computer Science School Leaving Examinations
No full textSupplemental data for:
Joyce Mahon, Brian Mac Namee, and Brett A. Becker. 2023. No More Pencils No More Books: Capabilities of Generative AI on Irish and UK Computer Science School Leaving Examinations. In The United Kingdom and Ireland Computing Education Research (UKICER) conference (UKICER 2023), September 7–8, 2023, Swansea, Wales UK. ACM, New York, NY, USA, 7 pages.
https://doi.org/10.1145/3610969.3610982
ChatGPT Prompts and Output for:
1. Leaving Certificate Computer Science Higher Level Examination 2022
p.2
2. Leaving Certificate Computer Science Higher Level Examination 2021
p.69
3. A-Level Computer Science Examination 2021 Paper 11
p.120
4. A-Level Computer Science Examination 2021 Paper 21
p.146
5. A-Level Computer Science Examination 2021 Paper 31
p.170
6. A-Level Computer Science Examination 2021 Paper 41
p.19
Resolving Perception Based Problems in Human-Computer Dialogue
Get PDFWe investigate the effect of sensor errors on situated human computer dialogues. If a human user instructs a robot to perform a task in a spatial environment, errors in the robot\u27s sensor based perception of the environment may result in divergences between the user\u27s and the robot\u27s understanding of the environment. If the user and the robot communicate through a language based interface, these problems may result in complex misunderstand ings. In this work we investigate such situations. We set up a simulation based scenario in which a human user instructs a robot to perform a series of manipulation tasks, such as lifting, moving and re-arranging simple objects. We induce errors into the robot\u27s perception, such as misclassification of shapes and colours, and record and analyse the user\u27s attempts to resolve the problems. We evaluate a set of methods to alleviate the problems by allowing the operator to access the robot\u27s understanding of the scene. We investigate a uni-directional language based option, which is based on automatically generated scene descriptions, a visually based option, in which the system highlights objects and provides known properties, and a dialogue based assistance option. In this option the participant can a.sk simple questions about the robot\u27s perception of the scene. As a baseline condition we perform the experiment without introducing any errors. We evaluate and compare the success and problems in all four conditions. We identify and compare strategies the participants used in each condition. We find that the participants appreciate and use the information request options successfully. We find that that all options provide an improvement over the condition without information.
We conclude that allowing the participants to access information about the robot\u27s perception state is an effective way to resolve problems in the dialogue
Robot perception errors and human resolution strategies in situated human-robot dialogue
Get PDFWe performed an experiment in which human participants interacted through a natural language dialogue interface with a simulated robot to fulfil a series of object manipulation tasks. We introduced errors into the robot’s perception, and observed the resulting problems in the dialogues and their resolutions. We then introduced different methods for the user to request information about the robot’s understanding of the environment. We quantify the impact of perception errors on the dialogues, and investigate resolution attempts by users at a structural level and at the level of referring expressions
Development of a Workflow for the Comparison of Classification Techniques
Get PDFAs the interest in machine learning and data mining springs up, the problem of how to assess learning algorithms and compare classifiers become more pressing. This has been associated with the lack of comprehensive and complete workflow depending on the project scale to provide guidance to its users. This means the success or failure of the project can be highly dependent on the person or team carrying it.
The standard practice adopted by many researchers and experimenters has been to follow steps or phases from existing workflows such as CRISP-DM, KDD and SASSEMMA. However, as machine learning and data mining fields involve complex comparative experiments, there is a need of having complete workflow which when applied provides efficient and effective results. Though existing workflows offers many benefits, a successful comparative experiment requires more than outlined steps of workflows. Conclusions based on results drawn from a more complete workflow will yield more reliable results and experimenter can stand with confidence while comparing classifiers.
This dissertation focuses on a range of issues from machine learning to statistics for the development of the classifier workflow. It represents in detail background materials which are the key to understanding how different experiments have to be carried out. It explains how different classification techniques work and their applications in different areas. It also explains how classification evaluations can be used in different domains. It also determines when an experimenter should use performance measures and how these measures correspond to performance estimators. Moreover, it explains how different settings can be obtained before committing to the experimentation step. Finally, a complete eight-phase classifier workflow which is platform independent will be provided. The workflow was then evaluated by expert users using close ended questionnaire
A Hybrid Agent-Based and Equation Based Epidemiological Model for the Spread of Infectious Diseases
Get PDFInfectious disease models are essential in understanding how an outbreak might occur and how best to mitigate an outbreak. One of the most important factors in modelling a disease is choosing an appropriate model and determining the assump tions needed to create the model. The main research questions this thesis addresses are how do we create a model for the spread of infectious diseases that captures heterogeneous agents without using an inordinate amount of computing power and how can we use that model to plan for future infectious disease outbreaks. We start our work by analysing and comparing equation based and agent based models and determine that an agent-based model’s stochasticity and ability to capture emerging results (complex and hard to explain results from interactions of agents) means that the agent-based model has an advantage in modelling the in dividual actions and complexities that make one infectious disease outbreak differ from another. Focusing on agent-based models, we take the model in two direc tions adding complexity and scaling up the model. Although adding complexity allows us to produce robust results, it increases run time so modelling anything beyond a small population is not feasible. Thus we focus on scaling up the model (from a town to a county) and determining what trade-offs need to be made to keep the model computationally tractable. With our scaled up model we look at characteristics of a town that come from its place in a network of towns, looking at how the centrality of a town affects how an outbreak spreads from a town and enters a town. We determine when a town has a high in degree centrality the i centrality of the other towns are not as important with respect to whether the outbreak will spread to the other towns. The additional agents in the scaled up model lead to an extended run time. In order to reduce run time we make an assumption about the importance of heterogeneous mixing when there is a large number of agents infected and create a hybrid agent-based and equation based model that switches between an agent based disease component and an equation based disease component based on a threshold of the number of agents infected. The hybrid model is able to save time compared to a fully agent-based model without losing a significant level of fidelity. This allows for the model to be scaled up to larger geographies and populations. Scaling the model to larger populations is essential in studying and testing the efficacy of interventions that would not be applicable at a smaller scale. To show this we use the hybrid model to analyse the effects of school closure policies across a network of towns, showing that closing both the town where an outbreak starts in and the town in the region with the highest in degree centrality can help mitigate an outbreak
Visualising The Complex Features Of Source Code
Get PDFSoftware development is a complex undertaking composed of several activities that include reading, writing, and modifying source code. Indeed, previous studies have shown that the majority of the effort invested in software development is dedicated to understanding code. This includes understanding the static structure, dynamic behaviour, and evolution of the code. Given these particular characteristics, as well as the high complexity of source code, it is reasonable to consider how visualisation can facilitate source code understanding. This work proposes to extend existing software development tools with visualisations that can be used to encode the various complex features within a source code document. Further, this work establishes a design space which includes a series of visualisations that are meant to complement existing textual views of source code, as found in source code editors. Several prototype visualisations are presented in this work which utilise various visualisation techniques, such as tree visualisations techniques and visualisation techniques based on the code-map metaphor. This work also presents several experiments, where the results of these experiments indicate that combining existing software development tools with visualisations can have a positive effect on source code understanding. Additionally, the result of the experiments show that the less a visualisation is abstracted from the original textural representation of a source code document, in terms of line, token, ordering, and character information, the more likely it is that viewers can link the visualisation to the code and back
Dynamic Estimation of Rater Reliability using Multi-Armed Bandits
Get PDFOne of the critical success factors for supervised machine learning is the quality of target values, or predictions, associated with training instances. Predictions can be discrete labels (such as a binary variable specifying whether a blog post is positive or negative) or continuous ratings (for instance, how boring a video is on a 10-point scale). In some areas, predictions are readily available, while in others, the eort of human workers has to be involved. For instance, in the task of emotion recognition from speech, a large corpus of speech recordings is usually available, and humans denote which emotions are present in which recordings
Drive-Based Utility-Maximizing Computer Game Non-Player Characters
Get PDFThis research examines the emergence of the five-string fiddle in contemporary North American fiddle culture within the past ten years. By interacting with leading artistlevel practitioners, the research documents the evolution and impact of the instrument to date in exploring the possibilities the five-string fiddle presents for musical performance and innovation. North American vernacular music and, in particular, the contemporary fiddle playing landscape, exemplifies virtousic and innovative idiomatic technique and improvisation as central to an overarching musical explosion, evidenced in the music of many high level, multi-stylistic contemporary practitioners. Within contemporary American fiddle performance, it is compelling to observe how many of the most innovative and highly regarded players now perform on five-string fiddles. The research uses a qualitative research methodology, drawing on interviews conducted with seven leading American fiddle players, each of whom has adopted the five-string fiddle in their own musical practice. The participants represent a rich cross section of American fiddle culture. They emerged naturally during the course of the literature review, and in-depth listening research, as particularly relevant sample cases. All participants were identified as leading exponents of the diversities encompassed in American fiddle music, between them sharing extensive professional recording, performance and academic experience, and all playing on five-string instruments. The research is further illuminated through practice, reflecting on my own musical work in illustrating how I have personally adopted the five-string fiddle, drawing influence from the research in demonstrating some wider possibilities of the instrument. This enquiry is important as it addresses the lack of specific research to date regarding the five-string fiddle, despite the significanance it holds for some of American fiddle music\u27s leading exponents, and consequently, for fiddle music itself. Equally significant, is the role of the instrument in facilitating the performance of innovative extended instrumental techniques, in particular, the five-string fiddles association with the rhythmic/percussive \u27chop\u27 bow techniques, now, so conspicuous within contemporary groove-based American string music. ix The findings of this research established the definitive emergence of the five-string fiddle, and subscribe that the five-string has now become a widely accepted part of the mainstream instrumentation in American music. This understanding emerges clearly through the words and practice of the participants. From this perspective, the research identifies the musical reasons that inspire the instruments popularity and elaborates through practice, the musical possibilities it presents to others.
behaviour selection systems that have been used successfully in industry. The evaluations show that UDGOAP can outperform these systems in both environments. Another novel contribution of this thesis is smart ambiance. Smart ambiance is an area of space in a virtual world that holds information about the context of that space and uses this information to have non-player characters inside the space select more contextually appropriate actions. Information about the context comes from events that took place inside the smart ambiance, objects inside the smart ambiance, and the location of the smart ambiance. Smart ambiance can be used with any cost based planner. This thesis demonstrates dierent aspects of smart ambiance by causing an industry standard action planner to select more contextually appropriate behaviours than it otherwise would have without the smart ambiance
- …
