Suche

Automatisation des commentaires de dégustation fournis par les consommateurs

Rolle: Hauptgesuchsteller/in

Financement: HES-SO Rectorat

Description du projet : Soutien-Contrepartie au dépôt déposée à la session de novembre 2018 du fonds interdomaine HES-SO, en regard du projet "Soon : Social Network of Machines"

Forschungsteam innerhalb von HES-SO: Ghorbel Hatem , Carrino Stefano , Brun-Cosme Aline , Ben Hamida Lamia

Partenaires académiques: Analyse de données; Management des villes et du territoire; Ghorbel Hatem, Analyse de données

Durée du projet: 28.02.2019 - 29.02.2020

Montant global du projet: 33'936 CHF

Statut: Abgeschlossen

mCARS -Système mobile de recommandation d'événements urbains

Rolle: Mitgesuchsteller/in

Requérant(e)s: 425 - Oenologie, Deneulin Pascale, 425 - Oenologie

Financement: HES-SO Rectorat

Description du projet : L'avis des consommateurs intéresse de plus en plus les entreprises agroalimentaires (moins cher et plus représentatif qu'un panel expert). Toutefois, l'analyse automatique de commentaires reste encore peu explorée et se fait aujourd'hui manuellement. Ce projet pluridisciplinaire vise à trouver un processus d'automatisation intégrant des difficultés telles que le traitement de phrases comme «vin très acide mais peu amer» ou «vin présentant une belle acidité et une amertume rafraichissante». Le traitement automatique des négations et la reconnaissance de la polarité des termes en fonction de leur environnement apparaissent comme les principaux défis à relever pour proposer un outil utile dans le domaine du marketing et de l'analyse sensorielle. Plusieurs étapes ont été mises en place durant le temps du projet. 1) Un grand nombre de commentaires de dégustation issus de professionnels et de consommateurs ont été recueillis via la plate-forme Mondovino (Coop) et diverses bases de données 2) Deux ontologies du vin regroupant plus de 2'225 termes distincts en français et 425 termes distincts en anglais ont été créées afin de hiérarchiser et standardiser les termes et concepts du vin 3) Les méthodes de langage naturel ont été appliquées et développées au moyen de plusieurs algorithmes d'apprentissage profond (deep machine learning) pour analyser l'information contenu dans les commentaires recueillis. 4) L'approche développée a été entrainée et testée sur plus de 1000 commentaires de dégustation recueillis lors de tests consommateurs et labélisés par nos soins. L'optimisation des algorithmes de machine learning appliqués au traitement de langage naturel (NLP) a montré une nette amélioration quant à l'extraction des concepts contenus dans les commentaires de dégustation, résultats supportés par des métriques classiques de machine learning. De même, la création d'une ontologie du vin conséquente et son implémentation continue permet à la fois d'améliorer l'extraction automatique mais également de diminuer la variabilité d'un traitement manuel. Si aujourd'hui l'ontologie du vin est conséquente en français, celle en anglais mériterait d'être complétée faute de matériel disponible à la fin de ce projet. Bien que spécifique au secteur du vin, son processus de construction et l'utilisation seront transposables à tout autre domaine touchant l'avis de consommateurs (p.ex. tourisme, agro-alimentaire). La valorisation de ce projet va se poursuivre dans les prochains mois avec la publication d'articles. Une plateforme de démonstration a été mise à disposition en ligne pour valoriser les résultats.

Forschungsteam innerhalb von HES-SO: Rébénaque-Martinez Pierrick , Van Gysel Laure , Punceva Magdalena , Visinand Steve , Ghorbel Hatem , Danthe Eve , Deneulin Pascale , Albertetti Fabrizio

Partenaires académiques: 425 - Oenologie; Info. Technique; Deneulin Pascale, 425 - Oenologie

Durée du projet: 01.11.2015 - 31.05.2018

Montant global du projet: 249'780 CHF

Statut: Abgeschlossen

Rolle: Hauptgesuchsteller/in

Financement: HES-SO Rectorat

Description du projet : L'idée de ce projet est de répondre au besoin d'un groupe de personnes (famille, amis, collègues, etc.) qui organisent une sortie collective en ville (sortie de famille, de classe, ou de labo par exemple) et qui souhaitent partager un moment autour d'une activité culturelle ou récréative (cinéma, théâtre, musée, manifestation, restaurant, festival, etc.). Afin d'éviter de longues discussions avant de décider de la destination, nous proposons dans ce projet un système de recommandation qui recommandera les activités culturelles les plus attractives en prenant en compte les informations contextuelles, en plus des goûts et préférences de chacun des membres du groupe

Forschungsteam innerhalb von HES-SO: Punceva Magdalena , Ghorbel Hatem , Abou Khaled Omar

Partenaires académiques: FR - EIA - Institut HumanTech; Analyse de données; Ghorbel Hatem, Analyse de données

Durée du projet: 01.01.2016 - 31.08.2017

Montant global du projet: 142'000 CHF

Statut: Abgeschlossen

iNUIT-2014: CrowdVision

Memoria-Mea : Stocker, classifier, rechercher et mémoriser ses informations multimédias personnelles

Rolle: Mitarbeiter

Requérant(e)s: Imagerie, Rizzotti Didier, Imagerie

Financement: HES-SO Rectorat

Description du projet : Real-time crowd monitoring during public events. Le but de ce projet est de fournir des informations sur les mouvements et le comportement des foules, afin d'aider les organisateurs d'événements à dimensionner les infrastructures et les services de sécurité, et à réagir rapidement en cas de mouvements dangereux (panique, bagarre). Les sources d'information principales seront les échanges sur les réseaux sociaux et les images des caméras de surveillance. Un accent particulier serra mis sur la sécurité. Pour ce projet cela signifie deux choses : 1. Sécurité physique, en permettant un bon dimensionnement des services de sécurité et une réaction rapide en cas de besoins. 2. Respect de la sphère privée, en garantissant l'anonymisation des données traitées. Ce projet s'insère dans le programme de recherche iNUIT (Internet of Things for Urban Innovation) de la HES-SO, et s'appuie sur différents composants et services développés dans le cadre d'autres projets de ce programme

Forschungsteam innerhalb von HES-SO: Rizzotti Didier , Punceva Magdalena , Ghorbel Hatem , Chabbi Houda , Tièche François

Partenaires académiques: FR - EIA - Institut iCoSys; Imagerie; Rizzotti Didier, Imagerie

Durée du projet: 01.02.2014 - 01.07.2016

Montant global du projet: 194'900 CHF

Statut: Abgeschlossen

Rolle: Mitgesuchsteller/in

Financement: HES-SO Valais-Wallis; HES-SO Rectorat; EIG - INIT; FR - EIA - ITIC - Multimedia Information Systems; IICT; ARC Ingénierie; FR - EIA - ITIC - Multimedia Information Systems

Description du projet : Le but de Memoria-Mea est la mise en place d'une plateforme PIM (Personnal Information management) ce qui correspondant à un organisateur informationnel de notre mémoire numérique personnelle. Le foisonnement de l'information numérique, la multitude de logiciels nous permettant à chaque fois de n'en classifier qu'une partie fait qu'à long terme, il nous est très difficile de retrouver d'une manière simple l'information dont nous avons besoin. En effet, il faut se plier à des logiques propres à chaque outil utilisé alors que l'idéal serait d'utiliser son propre mécanisme de recherche et de classement. Memoria-Mea vise à offrir un système permettant à une personne d'organiser, de classifier, et de rechercher tout type d'information multimédia numérique avec laquelle elle a interagit pendant ses activités quotidiennes (mails, documents, agenda, émissions TV ou radio, photos'). Les processus de recherche et de classification se baseront sur la logique de l'utilisateur, ses préférences, les aspects contextuels afin de fournir des modalités personnalisées de gestion et d'accès à l'information.

Forschungsteam innerhalb von HES-SO: Rudaz Patrice , Tscherrig Julien , Raileanu Laura Elena , Rizzotti Didier , Fatemi Nastaran , Gabioud Dominique , Ghorbel Hatem , Abou Khaled Omar , Mugellini Elena , Crausaz Didier

Durée du projet: 01.03.2007 - 31.12.2012

Montant global du projet: 390'000 CHF

Statut: Abgeschlossen

2022

SOON :

Wissenschaftlicher Artikel ArODES

social network of machines solution for predictive maintenance of electrical drive in industry 4.0

Laszlo Barna Iantovics, Adrian Gligor, Vicente Rodriguez Montequin, Zoltan Balogh, Ivana Budinska, Emil Gatial, Stefano Carrino, Hatem Ghorbel, Jonathan Dreyer

Acta Marisiensis. Seria Technologica, 2022, vol. 19, no. 2, pp. 12-19

Zusammenfassung:

Predictive methods represent techniques commonly met in Industry 4.0 that offer a way to early predict or detect faults of machines, devices or tools. This is useful to anticipate failures with the main goal of improving maintenance planning. Making such predictions could decrease the unexpected malfunction operation or manufacturing downtime and consequently the overall maintenance costs. In this paper we present the basis of the architecture designed for predictive maintenance in the project Social Network of Machines (SOON) under the paradigm of Industry 4.0, as well as a brief literature state of-the-art survey of the topic. A particular implementation of this architecture, a testbed for electrical motors failure detection, is shown and evaluated.

2021

CITIZEN PARTICIPATION & DIGITAL TOOLS TO IMPROVE PEDESTRIAN MOBILITY IN CITIES

Wissenschaftlicher Artikel

Sandoz Romain, Ertz Olivier, Scius-Bertrand Anna, Fischer Andreas, Hüsser Olivier, Ghorbel Hatem

The International Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences, 2021 , vol. XLVI-4/W1-2021, pp. 29-34

Zusammenfassung:

In this work, we present a framework supported by mobile and web apps and able to propose personalized pedestrian routes that match
user mobility profile considering mobility impediments factors. We explain how these later have been defined using a pedestrian-
centric approach based on travel experiences as perceived in the field by senior citizens. Through workshops, six main factors that may
influence pedestrian route choices were revealed: passability, obstacle in path, surface problem, security, sidewalk width, slope. These
categories were used to build digital tools and guide a citizen participatory approach to collect geolocated points of obstacle documented
with walkability information (picture, category, impact score, free comment). We also involved citizens to evaluate these information
and especially senior referents for validation. Finally we present how we connect these points of obstacle with a pedestrian network
based on OpenStreetMap to configure a routing cost function. The framework has been partially deployed in 2020 with limited people
due to the pandemic. Nonetheless, we share lessons learned from interaction with citizens in the design of such a framework whose
underlying workflow is reproducible. We plan to further assess its relevance and sustainability in the future.

2019

Smart adaptive run parameterization (SArP) :

Wissenschaftlicher Artikel ArODES

enhancement of user manual selection of running parameters in fluid dynamic simulations using bio-inspired and machine-learning techniques

Hatem Ghorbel, Nicolas Zannini, Salma Cherif, Florian Sauser, David Grunenwald, William Droz, Mahamadou Baradji, Djamel Lakehal

Soft Computing, 2019, first online

Zusammenfassung:

Computational fluid dynamic (CFD) simulations present numerous challenges in the domain of artificial intelligence. Computational time, resources and cost that can reach disproportional size before leading a simulation to its fully converged solution are one of the central issues in this domain. In this paper, we propose a novel algorithm that finds optimal parameter settings for the numerical solvers of CFD software. Indeed, this research proposes an alternative approach; rather than going deeper in reducing the mathematical complexity, it suggests taking advantage of the history of previous runs in order to estimate the best parameters for numerical equation resolution. In fact, our approach is bio-inspired and based on a genetic algorithm (GA) and evolutionary strategies enhanced with surrogate functions based on machine-learning meta-models. Our research method was tested on 11 different use cases using various configurations of the GA and algorithms of machine learning such as regression trees extra trees regressors and random forest regressors. Our approach has achieved better runtime performance and higher convergence quality (an improvement varying between 8 and 40%) in all of the test cases when compared to a basic approach which requires manually selecting the parameters. Moreover, our approach outperforms in some cases manual selection of parameters by reaching convergent solutions that couldn’t otherwise be achieved manually.

2018

Sequential dialogue act recognition for Arabic argumentative debates :

Wissenschaftlicher Artikel ArODES

Reconocimiento de acto de diálogo secuencial para debates argumentativos árabes

Samira Ben Dbabis, Hatem Ghorbel, Lamia Hadrich Belguith

Procesamniento del lenguaje, 2018, n° 60, pp. 53-60

Beurret Stéphane, Ghorbel Hatem, Neuenschwander Patrick

Zusammenfassung:

Dialogue act recognition remains a primordial task that helps user to automatically identify participants’ intentions. In this paper, we propose a sequential approach consisting of segmentation followed by annotation process to identify dialogue acts within Arabic politic debates.To perform DA recognition, we used the CARD corpus labeled using the SADA annotation schema. Segmentation and annotation tasks were then carried out using Conditional Random Fields probabilistic models as they prove high performance in segmenting and labeling sequential data. Learning results are notably important for the segmentation task (F-score=97.9%) and relatively reliable within the annotation process (fscore=63.4%) given the complexity of identifying argumentative tags and the presence of disfluencies in spoken conversations.

2016

Joint forces to create technology expertise for Industry 4.0: Innovative solutions for Industry 4.0 from TORNOS SA and HE-Arc

Professioneller Artikel

Innovative solutions for Industry 4.0 from TORNOS SA and HE-Arc

Magazine «berncapitalarea», 2016 , no 1

Zusammenfassung:

TORNOS SA can already supply interested customers with software solutions designed specifically for the 4th Industrial Revolution. The Bernese-Jura-based machine tool manufacturer has developed and launched the innovative TISIS software in close collaboration with Haute Ecole Arc Ingénierie (HE-Arc).

2015

Sentiment classification at discourse segment level :

Wissenschaftlicher Artikel ArODES

experiments on multi-domain Arabic corpus

Amine Bayoudhi, Hatem Ghorbel, Houssem Koubaa, Lamia Hadrich Belguith

The Journal for Language Technology and Computational Linguisitcs (JLCL), 2015, vol. 30, n°1, pp. 1-25

Zusammenfassung:

Sentiment classification aims to determine whether the semantic orientation of a text is positive, negative or neutral. It can be tackled at several levels of granularity: expression or phrase level, sentence level, and document level. In the scope of this research, we are interested in the sentence and sub-sentential level classification which can provide very useful trends for information retrieval and extraction applications, Question Answering systems and summarization tasks. In the context of our work, we address the problem of Arabic sentiment classification at sub-sentential level by (i) building a high coverage sentiment lexicon with semi-automatic approach; (ii) creating a large multi-domain annotated sentiment corpus segmented into discourse segments in order to evaluate our sentiment approach; and (iii) applying a lexicon-based approach with an aggregation model taking into account advanced linguistic phenomena such as negation and intensification. The results that we obtained are considered good and close to state of the art results in English language.

2023

Auction-based job scheduling for smart manufacturing

Konferenz ArODES

Emil Gatial, Zoltán Balogh, Sepideh Hassankhani Dolatabadi, Hatem Ghorbel, Stefano Carrino, Jonathan Dreyer, Vicente Rodriguez Montequin, Adrian Gligor, Laszlo Barna Iantovics

2021

Citizen participation & digital tools to improve pedestrian mobility in cities

Konferenz ArODES

Olivier Ertz, Andreas Fischer, Hatem Ghorbel, Olivier Hüsser, Romain Sandoz, Anna Scius-Bertrand

The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences ; Proceedings of the 6th International Conference on Smart Data and Smart Cities

Zusammenfassung:

In this work, we present a framework supported by mobile and web apps and able to propose personalized pedestrian routes that match user mobility profile considering mobility impediments factors. We explain how these later have been defined using a pedestrian-centric approach based on travel experiences as perceived in the field by senior citizens. Through workshops, six main factors that may influence pedestrian route choices were revealed: passability, obstacle in path, surface problem, security, sidewalk width, slope. These categories were used to build digital tools and guide a citizen participatory approach to collect geolocated points of obstacle documented with walkability information (picture, category, impact score, free comment). We also involved citizens to evaluate these information and especially senior referents for validation. Finally we present how we connect these points of obstacle with a pedestrian network based on OpenStreetMap to configure a routing cost function. The framework has been partially deployed in 2020 with limited people due to the pandemic. Nonetheless, we share lessons learned from interaction with citizens in the design of such a framework whose underlying workflow is reproducible. We plan to further assess its relevance and sustainability in the future.

SOON :

Konferenz ArODES

social network of machines to optimize task scheduling in smart manufacturing

Hatem Ghorbel, Jonathan Dreyer, Farid Abdalla, Vicente Rodriguez Montequin, Zoltan Balogh, Emil Gatial, Ivana Bundinska, Adrian Gligor, Laszlo Barna Iantovics, Stefano Carrino

Proceedings of the 2021 IEEE 32nd Annual International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC), 13-16 September 2021, Helsinki, Finland

Stefano Carrino, Jonathan Guerne, Jonathan Dreyer, Hatem Ghorbel, Alain Schorderet, Raphaël Montavon

Zusammenfassung:

The aim of the Social Network of machines (SOON) project is to investigate the impact of using of autonomous social agents to optimize manufacturing processes in the framework of Industry 4.0. In this article, we present the multi-agent SOON architecture and the built solutions aiming at optimizing the scheduling of tasks. Two different scheduling approaches are proposed. The first approach is based on an ‘auction’ paradigm where the task assignment is decided according to the capability of a machine agent to bid for a task. The second approach is built on a heterarchical agents network where agents learn the acquisition of cooperative tasks. Both solutions are capable of managing and synchronizing the communication between agents while performing their tasks. To describe each approach, two industrial use cases are illustrated: wire rod mill manufacturing and mechanical part manufacturing. Finally, in the heterarchical network, agents are trained with reinforcement learning to maximize the cumulative reward and optimize the manufacturing scheduling. Results show that reinforcement learning allows learning the optimal behavior in multiple scenarios.

2020

Machining quality prediction using acoustic sensors and machine learning

Konferenz ArODES

Proceedings of the 14th International Conference INTER-ENG 2020 Interdisciplinarity in Engineering, 8–9 October 2020, Mures, Romania

Zusammenfassung:

The online automatic estimation of the quality of products manufactured in any machining process without any manual intervention represents an important step toward a more efficient, smarter manufacturing industry. Machine learning and Convolutional Neural Networks (CNN), in particular, were used in this study for the monitoring and prediction of the machining quality conditions in a high-speed milling of stainless steel (AISI 303) using a 3mm tungsten carbide. The quality was predicted using the Acoustic Emission (AE) signals captured during the cutting operations. The spectrograms created from the AE signals were provided to the CNN for a 3-class quality level. A promising average f1-score of 94% was achieved.

2015

Sentiment classification of Arabic documents :

Konferenz ArODES

experiments with multi-type features and ensemble algorithms

Amine Bayoudhi, Lamia Hadrich Belguith, Hatem Ghorbel

Proceedings of the 29th Pacific Asia Conference on Language, Information and Computation, 30 October - 1 November 2015, Shanghai, China