• Digital Humanities and Cultural Heritage
• Publications close
• Contribution for newspaper or weekl... close

This evidence based practice paper presents preliminary results in using an artificial intelligence classifier to mark student assignments in a large class setting. The assessment task consists of an approximately 2000 word reflective essay that is produced under examination conditions and submitted electronically. The marking is a simple pass/fail determination, and no explicit feedback beyond the pass/fail grade is provided to the students. Each year around 1500 students complete this assignment, which places a significant and time-constrained marking load upon the teaching faculty.This paper presents a Natural Language Process (NLP) framework/tool for developing a machine learning based binary classifier for automated assessment of these assignments. The classifier allocates each assignment a score representing the probability that the assignment would receive a passing grade from a human marker. The effectiveness and performance of the classifier is measured by investigating the accuracy of those predictions.Several iterations and statistical analyses were carried out to determine operational thresholds that balance the risks of false positives and false negatives with the required quantity of human marking to assess the assignment.The resulting classifier was able to provide accuracy levels that are potentially feasible in an operational context, and the potential for significant overall reductions in the human marking load for this assignment.

Dr Erik von Willebrand first described a family with bleeding symptoms in a Finnish publication in 1926. A closer look at this landmark publication sheds light on some pathophysiological aspects of von Willebrand disease that may be applicable even in the current era. We attempt to relay in this article how the teachings from this original description may provide a benchmark for further research in this condition. Non peer reviewed

The emergence of fan fiction websites, where fans write their own storied about a topic/genre, has resulted in serious content rating issues. The websites are accessible to general audiences but often includes explicit content. The authors can rate their own fan fiction stories but this is not required and many stories are un- rated. This motivates automatically predicting the content rating using recent natural languages processing techniques. The length of the fan fiction text, ambiguity in ratings schemes, self-annotated (weak) labels, and style of writing all make automatic content rating prediction very difficult. In this paper, we propose several embedding techniques and classification models to address these problem. Based on a dataset from a popular fan fiction website, we show that binary classification is better than multiclass classification and can achieve nearly 70% accuracy using a transformer-based model. When computation is considered, we show that a traditional word embedding technique and Logistic Regression produce the best results with 66% accuracy and 0.1 seconds computation (approximately 15,000 times faster than DistilBERT). We further show that many of the labels are not correct and require subsequent preprocessing techniques to correct the labels. We propose an Active Learning approach, that while the results are not conclusive, suggest further work to address.

We present first results from the use of XGBoost, a highly effective machine learning (ML) method, within the Bristol Betting Exchange (BBE), an open-source agent-based model (ABM) designed to simulate a contemporary sports-betting exchange with in-play betting during track-racing events such as horse races. We use the BBE ABM and its array of minimally-simple bettor-agents as a synthetic data generator which feeds into our XGBoost ML system, with the intention that XGBoost discovers profitable dynamic betting strategies by learning from the more profitable bets made by the BBE bettor-agents. After this XGBoost training, which results in one or more decision trees, a bettor-agent with a betting strategy determined by the XGBoost-learned decision tree(s) is added to the BBE ABM and made to bet on a sequence of races under various conditions and betting-market scenarios, with profitability serving as the primary metric of comparison and evaluation. Our initial findings presented here show that XGBoost trained in this way can indeed learn profitable betting strategies, and can generalise to learn strategies that outperform each of the set of strategies used for creation of the training data. To foster further research and enhancements, the complete version of our extended BBE, including the XGBoost integration, has been made freely available as an open-source release on GitHub. Comment: Accepted for presentation/publication at the 16th International Conference on Agents and Artificial Intelligence (ICAART2024); Rome, Italy, 24-26 February 2024. 13 pages; 18 figures

Standardized laboratory testing is normally used for characterization of building materials and components. In real life, especially when assessing buildings in terms of waterdamaged building materials, built-in moisture, rising damp or other problems, information on the hygrothermal state of the constructions may be relevant, but is often sparse. This paper presents a study where application of a newly developed non-destructive experimental method is used to measure the relative humidity (RH) at the surface of four types of building materials. Knowing the RH at the surface, a risk assessment of fungal growth can be made. The materials include concrete, brick, aerated concrete, and gypsum, with and without a surface treatment. The method is based on the idea of establishing a water vapor transport towards equilibrium within a small air volume in contact with the surface of interest. In the laboratory the specimens were conditioned at respectively 65 % RH and 90 % RH in a climate chamber, and then transferred to conditions with 50 % RH. The open side of a Petri dish, equipped with a temperature and relative humidity sensor, was attached to the materials using reusable adhesive. The development of the relative humidity over time within the petri dish was monitored and parameterized mathematically, solving an exponential rate equation. The parameterization and certain sets of assumptions allow among other things a calculation of time to equilibrium which makes it possible to assess the reproducibility and comparison in between materials. As the materials were tested with and without surface treatment, it was possible to estimate the effect of a typical surface treatment on the moisture transport. The measurements compared well to simulations made by a model, though the model predicts slightly faster achievement of equilibrium. Despite the limited number of cases in the current study, the presented results are promising with regards to enable rough estimates of RH at the surface of building materials outside the laboratory with a very simple, inexpensive, and non-destructive method.

First we recall a method of computing scalar products of eigenfunctions of a Sturm-Liouville operator. This method is then applied to Macdonald and Gegenbauer functions, which are eigenfunctions of the Bessel, resp. Gegenbauer operators. The computed scalar products are well defined only for a limited range of parameters. To extend the obtained formulas to a much larger range of parameters, we introduce the concept of a generalized integral. The (standard as well as generalized) integrals of Macdonald and Gegenbauer functions have important applications to operator theory. Macdonald functions can be used to express the integral kernels of the resolvent (Green functions) of the Laplacian on the Euclidean space in any dimension. Similarly, Gegenbauer functions appear in Green functions of the Laplacian on the sphere and the hyperbolic space. In dimensions 1,2,3 one can perturb these Laplacians with a point potential, obtaining a well defined self-adjoint operator. Standard integrals of Macdonald and Gegenbauer functions appear in the formulas for the corresponding Green functions. In higher dimensions the Laplacian perturbed by point potentials does not exist. However, the corresponding Green function can be generalized to any dimension by using generalized integrals. Comment: 9 pages

Abstract Previous research estimates that building physics and occupant practices equally contribute to the explanation of variations in residential heating consumption. However, the so-called performance gap, where calculated heating demand diverge from actual heating consumption, indicates that the relation between occupant practices and building physics vary across energy efficiency of the building. In this paper, using data from 2019 to 2021, we investigate such interactions 1) by comparing the calculated heating demand with actual energy consumption for residential space heating and domestic hot water (DHW) across energy performance certificates (A to G), and 2) by investigating variations in over- and underconsumption (deviance from calculated to actual) across socio-economic characteristics of households. In line with previous studies, we find that households living in energy inefficient houses tend to have lower heating demand than expected. Moreover, we find that lower-income households and households living in rural area has less overconsumption, indicating that they have more frugal heating practices.