Papers by Fabrice Muhlenbach
Artificial Intelligence and Ethics: An Approach to Building Ethical by Design Intelligent Applications
Chapman and Hall/CRC eBooks, Oct 13, 2022
Combinations of statistical and semantic approaches applied to scientific digital libraries for the promotion of multidisciplinary research
The knowledge of all science domains is now available on digital libraries. The problem is that t... more The knowledge of all science domains is now available on digital libraries. The problem is that the papers belonging to different research communities do not use the same vocabulary to talk about the same subject. Access to relevant documents with information retrieval tools, search engines or research-paper recommender systems will fail if these methods do not consider this linguistic variability. In this work, we present strategies for using artificial intelligence technologies to successfully expand the literature search to bring diversity to the recommended results, thereby promoting multidisciplinary research.
Methodology for Creating a Community Corpus Using a Wikibase Knowledge Graph
Communications in computer and information science, 2022
Proceedings of the 15th International Joint Conference on Biomedical Engineering Systems and Technologies, 2022
The lack of readily available disability data is a major barrier for disability advocacy globally... more The lack of readily available disability data is a major barrier for disability advocacy globally. The collection and access to disability data is crucial to address social inequities, discrimination, and human rights violations within the disability community. The Disability Wiki project intends to use AI techniques such as Machine Learning and Semantic Web to extract and store existing disability-related data into one platform (Wikibase) and to provide a multilingual natural language enabled search engine and a screen-reader-accessible for its users. a
Quelles difficultés?: Approche des difficultés de l'analyse empirique du contentieux
Fair Recommendations Through Diversity Promotion
We address the problem of overspecialization in streaming platform recommender systems. The perso... more We address the problem of overspecialization in streaming platform recommender systems. The personalization of web pages by delivering content to users is a challenging task in data mining. But it has been proved that beside optimizing the relevance accuracy such systems should also rely on other factors like diversity or novelty. In this paper we focus on modeling users’ boundary area of interest by selecting the most diverse items they liked in the past. We apply diversification while building the top-N list of recommendations. We select the items we want to recommend from an area where we consider a user will find item different from what she or he likes in the past. We evaluate our approach in offline analysis on two datasets, showing that our approach brings diversity and is competitive against implicit state-of-the-art method.
Studies in Health Technology and Informatics, 2021
Human rights monitoring for people with disabilities is in urgent need for disability data that i... more Human rights monitoring for people with disabilities is in urgent need for disability data that is shared and available for local and international disability stakeholders (e.g., advocacy groups). Our aim is to use a Wikibase for editing, integrating, storing structured disability related data and to develop a Natural Language Processing (NLP) enabled multilingual search engine to tap into the wikibase data. In this paper, we explain the project first phase.
International Journal of Law and Information Technology, 1994
This paper examines coding applied by seven different review groups on the same set of twenty eig... more This paper examines coding applied by seven different review groups on the same set of twenty eight thousand documents. The results indicate that the level of agreement between the reviewer groups is much lower than might be suspected based on the general level of confidence on the part of the legal profession in the accuracy and consistency of document review by humans. Each document from a set of twenty eight thousand documents was reviewed for responsiveness, privilege and relevance to specific issues by seven independent review teams. Examination of the seven sets of coding tags for responsiveness revealed an inter-reviewer agreement of 43% for either responsive or non-responsive determinations. The agreement on the responsive determination alone was 9% and on the non-responsive determination was 34% of the total document family count. Pair-wise analysis of the seven groups of reviewers provided higher rates, however no pairing of the teams indicated that there is an unequivocally 1 Thomas I. Barnett is the leader of the e-Discovery, records and information management consulting division of Iron Mountain, Inc.; Svetlana Godjevac is a senior consultant at Iron Mountain, Inc. 2 superior assessment of the dataset by any of the teams. This paper considers the ramifications of low agreement of human manual review in the legal domain and the need for industry benchmarks and standards. Suggestions are offered for improving the quality of human manual review using statistical quality control (QC) measures and machine-learning tools for pre-assessment and document categorization. Group A B C D E F G Non-Responsive
Dans les corpus de textes scientifiques, certains articles issus de communautes de chercheurs dif... more Dans les corpus de textes scientifiques, certains articles issus de communautes de chercheurs differentes peuvent ne pas etre decrits par les memes mots-cles alors qu'ils partagent la meme thematique. Ce phenomene cause des problemes dans la recherche d'information, ces articles etant mal indexes, et limite les echanges potentiellement fructueux entre disciplines scientifiques. Notre modele permet d'attribuer automatiquement une etiquette thematique aux articles au moyen d'un apprentissage des representations semantiques d'articles du corpus deja etiquetes. Passant bien a l'echelle, cette methode a pu etre testee sur une bibliotheque numerique d'articles scientifiques comportant des millions de documents. Nous utilisons un reseau semantique de synonymes pour extraire davantage d'articles semantiquement similaires et nous les fusionnons avec ceux obtenus par un modele de classement thematique. Cette methode combinee presente de meilleurs taux de rappel...
Process Knowledge Model For Facilitating Industrial Components' Manufacturing
Computing the semantic relatedness between two entities has many applications domains. In this pa... more Computing the semantic relatedness between two entities has many applications domains. In this paper, we show a new way to compute the semantic relatedness between two resources using semantic web data. Moreover, we show how this measure can be used to compute the semantic relatedness between music genres which can be used for music recommendation systems. We first describe how to build a vector representations for resources in an ontology. Subsequently we show how these vector representations can be used to compute the semantic relatedness of two resources. Finally, as an application, we show that our measure can be used to compute the semantic relatedness of music genres. CCS Concepts •Information systems → Similarity measures; Language models;
Recommender Systems
Design Structure Matrix (DSM) Symmetric matrix that indicates the links/interfaces between decomp... more Design Structure Matrix (DSM) Symmetric matrix that indicates the links/interfaces between decomposed product components Hierarchical Decomposition Methods to decompose products in to components and subcomponents following product hierarchies Systematic Variation Method that refers to the search for and combination of solutions to design subproblems Satisficing Method that refers to the evaluation and selection of alternative solutions and the understanding that searches should not be focused on finding the optimal solution Discursiveness Method that refers to a step-bystep, yet iterative, approach to the product development process Lead User Person who are ahead of trends and develop and/or modify for their own benefit new products and processes Definition
Résumé. Dans les systèmes de recommandation, l’approche du filtrage sur le contenu est revenue en... more Résumé. Dans les systèmes de recommandation, l’approche du filtrage sur le contenu est revenue en force face à celle du filtrage collaboratif grâce à l’arrivée du paradigme de l’apprentissage profond et des techniques de word embedding. Dans cette même veine, l’avènement des folksonomies et du web sémantique a apporté une meilleure compréhension des profils des utilisateurs et des caractéristiques des articles à recommander. Dans cet article, nous nous intéressons au domaine musical et nous introduisons un nouveau calcul de mesure de préférence intégrée dans un système de recommandations basées sur le contenu. En testant notre approche sur le jeu de données Last.fm, nous montrons que l’utilisation de termes issus d’une folksonomie associés à des informations issues du web sémantique permet d’améliorer le processus de recommandation musicale.
The advent of machine learning techniques has made it possible to obtain predictive systems that ... more The advent of machine learning techniques has made it possible to obtain predictive systems that have overturned traditional legal practices. However, rather than leading to systems seeking to replace humans, the search for the determinants in a court decision makes it possible to give a better understanding of the decision mechanisms carried out by the judge. By using a large amount of court decisions in matters of divorce produced by French jurisdictions and by looking at the variables that allow to allocate an alimony or not, and to define its amount, we seek to identify if there may be extra-legal factors in the decisions taken by the judges. From this perspective, we present an explainable AI model designed in this purpose by combining a classification with random forest and a regression model, as a complementary tool to existing decision-making scales or guidelines created by practitioners.
Résumé. Les systèmes de recommandation permettent de présenter à un utilisateur des éléments susc... more Résumé. Les systèmes de recommandation permettent de présenter à un utilisateur des éléments susceptibles de l’intéresser. La mise en place de tels systèmes dans les domaines culturels soulève souvent le questionnement de la place de la diversité, de la nouveauté, et surtout de la découverte. Nous pensons que l’être humain, bien qu’ayant ordinairement une tendance à se placer dans une zone de confort correspondant à ce qu’il connaît, apprécie occasionnellement d’être poussé à des explorations le faisant sortir de sa routine. Nous avons développé dans cette optique une méthode, basée sur la dissimilarité, qui élargit les centres d’intérêt des utilisateurs. Nous avons réussi à délimiter une zone intermédiaire entre des items « trop similaires » et des items « trop différents ». Afin de valider cette hypothèse, nous avons développé une application qui permet de tester et de valider cette méthode. Dans cet article de démonstration, nous expliquons le concept de « zone intermédiaire », n...
Considering the use of artificial intelligence for greater personalization of patient care and be... more Considering the use of artificial intelligence for greater personalization of patient care and better management of human and material resources may seem like an opportunity not to be missed. In order to offer a better humanization of the care pathway, artificial intelligence is a tool that decision-makers in the hospital sector must appropriate by taking care of the new ethical issues and conflicts of values that this technology generates. Envisager le recours a l'intelligence artificielle pour une plus grande personnalisation de la prise en charge du patient et une meilleure gestion des ressources humaines et materielles peut sembler une opportunite a ne pas manquer. Afin de proposer une meilleure humanisation du parcours de soin, l'intelligence artificielle est un outil que les decideurs du milieu hospitalier doivent s'approprier en veillant aux nouveaux enjeux ethiques et conflits de valeurs que cette technologie engendre.
A Virtual Community for Disability Advocacy: Development of a Searchable Artificial Intelligence–Supported Platform (Preprint)
BACKGROUND The lack of availability of disability data has been identified as a major challenge h... more BACKGROUND The lack of availability of disability data has been identified as a major challenge hindering continuous disability equity monitoring. It is important to develop a platform that enables searching for disability data to expose systemic discrimination and social exclusion, which increase vulnerability to inequitable social conditions. OBJECTIVE Our project aims to create an accessible and multilingual pilot disability website that structures and integrates data about people with disabilities and provides data for national and international disability advocacy communities. The platform will be endowed with a document upload function with hybrid (automated and manual) paragraph tagging, while the querying function will involve an intelligent natural language search in the supported languages. METHODS We have designed and implemented a virtual community platform using Wikibase, Semantic Web, machine learning, and web programming tools to enable disability communities to uploa...
2017 IEEE International Conference on Data Mining Workshops (ICDMW)
In this article we address the problem of expanding the set of papers that researchers encounter ... more In this article we address the problem of expanding the set of papers that researchers encounter when conducting bibliographic research on their scientific work. Using classical search engines or recommender systems in digital libraries, some interesting and relevant articles could be missed if they do not contain the same search key-phrases that the researcher is aware of. We propose a novel model that is based on a supervised active learning over a semantic features transformation of all articles of a given digital library. Our model, named Semantic Search-by-Examples (SSbE), shows better evaluation results over a similar purpose existing method, More-Like-This query, based on the feedback annotation of two domain experts in our experimented use-case. We also introduce a new semantic relatedness evaluation measure to avoid the need of human feedback annotation after the active learning process. The results also show higher diversity and overlapping with related scientific topics which we think can better foster transdisciplinary research.
2020 IEEE International Conference on Systems, Man, and Cybernetics (SMC)
The introduction of artificial intelligence into activities traditionally carried out by human be... more The introduction of artificial intelligence into activities traditionally carried out by human beings produces brutal changes. This is not without consequences for human values. This paper is about designing and implementing models of ethical behaviors in AI-based systems, and more specifically it presents a methodology for designing systems that take ethical aspects into account at an early stage while finding an innovative solution to prevent human values from being affected. Two case studies where AI-based innovations complement economic and social proposals with this methodology are presented: one in the field of culture and operated by a private company, the other in the field of scientific research and supported by a state organization.
Uploads
Papers by Fabrice Muhlenbach