Projets novateurs réalisés

Traitement du langage et Résumé automatique de documents

Un résumé est un texte qui décrit de façon synthétique la forme, le contenu, et la thèse principale d’un ensemble de document. Une société comme la Fédération des caisses Desjardins reçoit énormément d’informations provenant de ses membres et clients, notamment via les sondages de satisfaction. Grâce à différentes méthodes de traitement automatique du langage nous pouvons extraire quels sont les opinions les plus souvent évoqués dans ces commentaires. Une fois ces informations identifiées nous pouvons mettre en place des algorithmes permettant de créer automatiquement ces résumés. Ceci a pour avantages de permettre aux conseillers de la fédération de connaitre l’opinion de nos membres sur différents produits beaucoup plus rapidement et efficacement.

Improving the Reliability of AI Systems from a Software Engineering Perspective

Artificial Intelligence techniques have been widely applied to solve real-world challenges, from autonomous driving cars, to detecting diseases. With the popularity of 5G wireless network, more and more AI systems are being developed to provide convenient services to everyone. It is important to ensure the reliability and quality of AI systems from every phase in software development cycle, i.e., development, integration, deployment and monitoring. In this collaboration with Ericsson GAIA, we will propose techniques to systematically improve the quality and reliability of AI systems. In particular, we will explore new testing techniques to test AI systems, especially the underlying AI models. Moreover, we will adapt and improve explainable AI models for the purpose of debugging and optimizing AI models. Automated techniques to support the development, debugging, deployment and monitoring of AI systems will enrich the ecosystem of AI systems in the era of 5G.

Industrial application of genomics derived biomarkers of salmon performance

The Atlantic salmon aquaculture industry is becoming a strategic economic sector for Canada. Aquaculture in Canada employs approximately 25,000 people with a total economic impact of over $5B. While farmed salmon is already Canada’s top aquaculture export, salmon aquaculture has significant capacity for growth in this country. Atlantic salmon farmed in sea cages on the Canadian coasts face multiple environmental stressors that can impair their growth performance and immune status. Aquafeeds can be formulated to improve fish resistance to disease (i.e., health-promoting) or healing after adverse environmental conditions (e.g., harmful algal blooms, hypoxia). The main goal of this research is to develop and apply diagnostic molecular tools such as multiplex PCR assays to evaluate these feeds. Through this collaborative research between a Canadian university (MUN) and a world-leading company in the production of aquafeeds [Cargill Animal Nutrition (CAN)], the intern will gain industry-relevant experience while ensuring the right to publish their research to help in building an academic career.

Economics Reference Guide to Music-Use Copyright Tariffs in Canada

There is a growing body of international research demonstrating that the economic contribution

copyright based industries is becoming increasingly important. Specifically, there is a growing

need in the Canadian music and music-use related industry to analyze the copyright process applied

towards different uses of music in Canada. The objective of the proposed “Reference Guide” is to

provide a detailed overview of the evolution of all major music-use related copyright tariffs in

Canada, including the economic rationale and supporting evidence used to set and modify each

tariff over time. The Reference Guide will focus on economic rather than legal matters in the case

of each of the copyright tariffs considered. The Guide will be available in a hard copy as well as

electronic form with key word search capability. It will fill an important information gap and

provide a research tool to copyright professionals.

Wireless Network Testbed for Network Slicing and Data-driven System Parameter Optimization

The goal of this project is to design and develop a wireless network testbed at the University of British Columbia (UBC) for Rogers Communications Canada Inc. to support different use cases for the fifth generation (5G) wireless networks. We will study the concept of self-organizing network (SON) and design a deep learning-based algorithm for our testbed to determine the optimal network parameters based on network traffic data and key performance indicator (KPI) statistics. We will also design a network traffic forecasting algorithm by capturing the mobility patterns of users. We will employ the proposed forecasting method to equip our testbed with mobility management algorithms for frequent handover operations. We will leverage SDN and network functions virtualization (NFV) to enable our testbed to support network slicing. We will design algorithms for dynamic radio resource allocation within each slice to satisfy the service requirements in each use case.

Wide-baseline Novel Scene Synthesis from a Single Image

Novel view synthesis is the process of generating new images from an unseen perspective, given at least one image of a scene. There may be more than one probable novel view associated with each unseen perspective, an assumption made by existing methods. This simplifying assumption prevents these methods from being applied to more difficult novel views where the set of probable novel views is highly varied. This project proposes to investigate a new approach to generate a wide variety novel views from a single image, and can produce multiple probable outputs. The proposed project will also evaluate the efficacy of the approach with existing approaches on a wide range of camera poses. Major components of this project utilize methods that can be applied to a wider range of domains than computer vision. RBC is a major Canadian bank that can leverage developed methods as part of their products and services.

The role of Peyer’s patches in mycotoxicity

Mycotoxins are toxic metabolites from mold that commonly contaminate cereal and livestock feed crops around the world. The consumption of mycotoxins contributes to substantial economic losses to the livestock industry due to reduced productivity and health. There are over 500 different classes of mycotoxins, many of which have unknown mechanisms of action (MOA). Adding mycotoxin binders to livestock feeds is an important remediation strategy for reducing risk of mycotoxicosis, since they can reduce bioavailability of certain mycotoxins, reducing absorption across the gut epithelium. When testing the toxicity of mycotoxins in different species, and the efficacy of mycotoxin binders for reducing mycotoxin bioavailability, in vitro studies are critical for initial screening. Previous studies have demonstrated the utility of a bovine macrophage cell line (BoMAC) for carrying out such studies; however, BoMACs alone do not morphologically reflect the intestinal barrier where most mycotoxin exposure occurs. It is physiologically more relevant to carry out in vitro exposure studies using intestinal epithelial cells (IEC), or even better, co-cultured IEC + macrophages (IEC+BoMAC).

Impact de la réalité augmentée sur l’expérience de consommation et les intentions d’achat de produits industriels

L’objet de cette recherche est de mieux comprendre l’utilité des expériences virtuelles, ici la réalité augmentée et la visualisation 3D, dans le cadre de la promotion de produits industriels personnalisables. En effet, l’utilisation de ces outils par les entreprises offrent de nombreux avantages, y compris dans le cadre de la personnalisation de masse, tendance essentielle de ces dernières années (Bonetti, Warnaby, & Quinn, 2018; Deloitte, 2019). L’entreprise partenaire propose ainsi aux entreprises des solutions de visualisation 3D et de réalité augmentée qui permettent de mettre en valeur leur produit et de présenter les différentes options de personnalisation possibles. En ce sens, on s’intéresse ici à une meilleure compréhension des facteurs d’adoptions, des freins et de l’impact sur le processus de décision des différents acteurs du centre d’achat associés à l’utilisation des technologies d’expériences virtuelles ainsi qu’à leur impact sur le ROI des entreprises qui mettent en place l’utilisation de telles technologies.

Modeling of wall-pressure fluctuations induced by pressure gradient turbulent boundary layer

The wall-pressure fluctuations induced by turbulent boundary layers (TBL) can be an important source of noise and vibration in different applications. These fluctuations yield the trailing-edge noise, dominant source of wind turbine or marine propeller noise, and others. A correct modeling of the generated wall-pressure fluctuations is the first issue in analyzing this phenomenon. The project applicant and its supervising professor have largely studied TBL noise, publishing innovative solutions for certain flow conditions. The use of empirical model fit to
experimental data or based on analytical solutions are considered the best tool to estimate the turbulent noise, which allows handling the problem easily making it industrially viable. This project intends to improve and extend the existing approaches by combining the expertise of the applicant and the supervisor in collaboration with Mecanum. Furthermore, the novel model will be applied to investigate the design of acoustic liners for Urban Air Vehicles.

Usability evaluation of a runway testing device interface

Various forms of usability testing can be used to optimize interface design and maximize human-computer interaction principles . A well-integrated, intuitive interface has the capacity to improve human efficiency, mitigate errors or lapses and improve situational awareness. Usability methodologies such as heuristic evaluations and cognitive walkthroughs can be performed at any point in the design process to identify areas for improvement. Additionally, in-field usability testing with users of the technology can further pinpoint inadequacies in software or hardware design. The proposed project will evaluate an interface design for an aviation braking availability tester (BAT) device by completing a systematic evaluation by an HCI expert. Outcomes to the partner organization include delination of any problem areas according to human factors principles, suggestions on improvements to the BAT system operation, and usage guidelines to be considered for future training applications.

Effet de la structure spatiale intra-peuplement sur la modélisation de la succession en forêt boréale mixte

Afin de gérer de façon durable les ressources forestières, il importe de prévoir comment une forêt sera régénérée après un scénario de coupe donné. Des modèles de simulation permettent de faire ces prévisions à moyen et long terme, mais leur précision dépend notamment de la qualité des données qu’on y entre. Dans ce projet, nous voulons déterminer l’importance de connaître la disposition détaillée des différentes essences dans une forêt afin de prédire comment la composition de cette forêt changera après plusieurs décennies. Cette recherche aidera les entreprises forestières, dont notre partenaire Rayonier A.M., à établir de meilleurs plans d’aménagement durable pour répondre aux exigences des normes de certification forestière.

Developpement d’un materiau ignifuge non toxique pour Ie procede derotomoulage et d’un protocole de test permettant d’evaluer les caracteristiques decombustion de ces resines.

Le projet de recherche concerne le developpemen d’un composite polymere pouvant etre moule dane le procede do roromoulage. Tout d’abord, des tests de feu devront etre prepares affin de s’assurer de la qualite des formulations de resines developpees sans avoir a les faire analyser dans un laboratoire accredite (tres cher). Les tests a developper seront similaires aux tests standards: ASTM E162, ASTM E662 et ombardier SMP 800-C. Par la suite, le developpement du composite polymere sera le principlal defi. Afin d’obtenir les specifications demanes, divers polymeres, materiaux et additifs devront etre testes dans les formulations.