Polish Consumers’ Understanding of Different Front-of-Package Food Labels: A Randomized Experiment

Dietary practices are a key behavioral factor in chronic disease prevention; one strategy for improving such practices population-wise involves front-of-package labels (FoPL). This online randomized study, conducted in a quota-based sample of 1159 Polish adults (mean age = 40.9 ± 15.4 years), assessed the objective understanding of five FoPL: Health Star Rating, Multiple Traffic Lights, NutriScore, Reference Intakes (RI) and Warning Label. Objective understanding was evaluated by comparing results of two nutritional quality ranking tasks (without/with FoPL) using three food categories (breakfast cereals, cakes, pizza). Associations between FoPL exposure and objective understanding were assessed via multivariable ordinal logistic regression. Compared to RI and across food categories, significant improvement in objective understanding was seen for NutriScore (OR = 2.02; 95% CI: 1.41–2.91) and Warning Label (OR = 1.61; 95% CI: 1.12–2.32). In age-stratified analyses, significant improvement in objective understanding compared to RI emerged mainly among adults aged 18–30 years randomized to NutriScore (all food categories: OR = 3.88; 95% CI: 2.04–7.36; cakes: OR = 6.88; 95% CI: 3.05–15.51). Relative to RI, NutriScore was associated with some improvement in objective understanding of FoPL across and within food categories, especially among young adults. These findings contribute to the ongoing debate about an EU-wide FoPL model.

Neural ranking models for document retrieval

Ranking models are the main components of information retrieval systems. Several approaches to ranking are based on traditional machine learning algorithms using a set of hand-crafted features. Recently, researchers have leveraged deep learning models in information retrieval. These models are trained end-to-end to extract features from the raw data for ranking tasks, so that they overcome the limitations of hand-crafted features. A variety of deep learning models have been proposed, and each model presents a set of neural network components to extract features that are used for ranking. In this paper, we compare the proposed models in the literature along different dimensions in order to understand the major contributions and limitations of each model. In our discussion of the literature, we analyze the promising neural components, and propose future research directions. We also show the analogy between document retrieval and other retrieval tasks where the items to be ranked are structured documents, answers, images and videos.

SWSA: A Hybrid Scientific Workflow Scheduling Algorithm based on Metaheuristic Approach in Cloud Computing Environment

The cloud computing environment with a set of distributed computing resources is a suitable platform for the execution of large-scale applications. One of these applications is scientific workflow applications in which a large set of interrelated tasks are executed for a certain purpose. Scientific workflow scheduling is one of the main challenges in this area, which aims at the optimal assignment of tasks to computational resources. Given the heterogeneity of cloud computing resources, the scientific workflow scheduling is an NP-Complete problem that can be solved by heuristic methods. In this paper, an improved evolutionary algorithm called Scientific Workflow Scheduling Algorithm (SWSA) for scheduling scientific workflows in the cloud will be provided by ranking tasks and improving the initial population of tasks. The objective of this algorithm is to create a balance and an improvement in the parameters of the execution cost and workflow execution completion time. In this proposed approach, a heuristic algorithm is used to rank and generate the initial population, which increases the convergence rate. The experimental results show that SWSA is more efficient in terms of cost and execution time compared with other approaches.

Impact of information presentation on interpretability of spatial hazard information: Lessons from a study in avalanche safety

Avalanche warning services publish avalanche condition reports, often called avalanche bulletins, to help backcountry recreationists make informed risk management decisions about when and where to travel in avalanche terrain. To be successful, the information presented in bulletins must be properly understood and applied prior to entering avalanche terrain. However, few avalanche bulletin elements have been empirically tested for their efficacy in communicating hazard information. The objective of this study is to explicitly test the effectiveness of three different graphics representing the aspect and elevation of avalanche problems on users’ ability to apply the information. To address this question, we conducted an online survey that presented participants with one of three graphic renderings of avalanche problem information and asked them to rank a series of route options in order of their exposure to the described hazard. Following completion of route ranking tasks, users were presented with all three graphics and asked to rate how effective they thought the graphics were. Our analysis dataset included responses from 3,056 backcountry recreationists with a variety of backgrounds and avalanche safety training levels. Using a series of generalized linear mixed effects models, our analysis shows that a graphic format that combines the aspect and elevation information for each avalanche problem is the most effective graphic for helping users understand the avalanche hazard conditions because it resulted in higher success in picking the correct exposure ranking, faster completion times, and was rated by users to be the most effective. These results are consistent with existing research on the impact of graphics on cognitive load and can be applied by avalanche warning services to improve the communication of avalanche hazard to readers of their avalanche bulletins.

Effective and practical neural ranking

Supervised machine learning methods that use neural networks ("deep learning") have yielded substantial improvements to a multitude of Natural Language Processing (NLP) tasks in the past decade. Improvements to Information Retrieval (IR) tasks, such as ad-hoc search, lagged behind those in similar NLP tasks, despite considerable community efforts. Although there are several contributing factors, I argue in this dissertation that early attempts were not more successful because they did not properly consider the unique characteristics of IR tasks when designing and training ranking models. I first demonstrate this by showing how large-scale datasets containing weak relevance labels can successfully replace training on in-domain collections. This technique improves the variety of queries encountered when training and helps mitigate concerns of over-fitting particular test collections. I then show that dataset statistics available in specific IR tasks can be easily incorporated into neural ranking models alongside the textual features, resulting in more effective ranking models. I also demonstrate that contextualized representations, particularly those from transformer-based language models, considerably improve neural ad-hoc ranking performance. I find that this approach is neither limited to the task of ad-hoc ranking (as demonstrated by ranking clinical reports) nor English content (as shown by training effective cross-lingual neural rankers). These efforts demonstrate that neural approaches can be effective for ranking tasks. However, I observe that these techniques are impractical due to their high query-time computational costs. To overcome this, I study approaches for offloading computational cost to index-time, substantially reducing query-time latency. These techniques make neural methods practical for ranking tasks. Finally, I take a deep dive into better understanding the linguistic biases of the methods I propose compared to contemporary and traditional approaches. The findings from this analysis highlight potential pitfalls of recent methods and provide a way to measure progress in this area going forward.

Exploring Change and Continuities in Internationally Mobile Pre-Service Teachers’ Conceptualisations of Effective Teaching

This qualitativestudy investigatesthe changes and continuities in conceptions of teaching and learning from course commencement to course completion for a group of international pre-service teachers undertaking a two-year Masters-level degree in Initial Teacher Education (ITE). Data were collected through a series of graphic elicitation activities and ranking tasks at baseline and endpoint. Findings indicate that there was:a growing emphasis on student engagement and its linkages to student learning; a shift from viewing teaching as the transfer of knowledge to learning as anactive process; and a more developed repertoire of professional language to explain what is valued and why. This study provides valuable insights into international pre-service teachers’ evolving conceptions of teaching and learning. These findings suggest that international pre-service teachersneed many opportunities to interrogate and refine their understanding of teaching and learning and how this appliesto the contexts in which they will teach.

Semi-supervised Multi-task Network For Image Aesthetic Assessment

Image aesthetic assessment has always been regarded as a challenging task because of the variability of subjective preference. Besides, the assessment of a photo is also related to its style, semantic content, etc. Conventionally, the estimations of aesthetic score and style for an image are treated as separate problems. In this paper, we explore the inter-relatedness between the aesthetics and image style, and design a neural network that can jointly categorize image by styles and give an aesthetic score distribution. To this end, we propose a multi-task network (MTNet) with an aesthetic column serving as a score predictor and a style column serving as a style classifier. The angular-softmax loss is applied in training primary style classifiers to maximize the margin among classes in single-label training data; the semi-supervised method is applied to improve the network’s generalization ability iteratively. We combine the regression loss and classification loss in training aesthetic score. Experiments on the AVA dataset show the superiority of our network in both image attributes classification and aesthetic ranking tasks.

Functional MRI Investigation on Paradigmatic and Syntagmatic Lexical Semantic Processing

The word embeddings related to paradigmatic and syntagmatic axes are applied in an fMRI encoding experiment to explore human brain’s activity pattern during story listening. This study proposes the construction of paradigmatic and syntagmatic semantic embeddings respectively by transforming WordNet-alike knowledge bases and subtracting paradigmatic information from a statistical word embedding. It evaluates the semantic embeddings by leveraging word-pair proximity ranking tasks and contrasts voxel encoding models trained with the two types of semantic features to reveal the brain’s spatial pattern for semantic processing. Results indicate that in listening comprehension, paradigmatic and syntagmatic semantic operations both recruit inferior (ITG) and middle temporal gyri (MTG), angular gyrus, superior parietal lobule (SPL), inferior frontal gyrus. A non-continuous voxel line is found in MTG with a predominance of paradigmatic processing. The ITG, middle occipital gyrus and the surrounding primary and associative visual areas are more engaged by syntagmatic processing. The comparison of two semantic axes’ brain map does not suggest a neuroanatomical segregation for paradigmatic and syntagmatic processing. The complex yet regular contrast pattern starting from temporal pole, along MTG to SPL necessitates further investigation.