Attended the L&T Network talk: 'Feeling numb with monotony' (31st of March 2021)
Mirjam Brady talked about her reflections on blended teaching in the second semester.
The title of my thesis is "Argumentation Techniques for Existential Rules". It is an original research in the field of Knowledge Representation and Reasoning, one of the main sub-domains in AI.
I was supervised by Madalina Croitoru, Srdjan Vesic and Rallou Thomopoulos. I was also supported by my friend and co-worker, Pierre Bisquert.
During the whole duration of my PhD, I was part of the INRIA GraphIK team at the LIRMM laboratory. I defended my thesis on the 11th of July 2019.
I studied reasoning techniques with argumentation graphs generated from inconsistent knowledge bases expressed in the existential rules language. The three main results are the following. First, we give a structural study of argumentation graphs obtained from knowledge bases expressed in existential rules. Second, we propose and analyse an argumentation framework with sets of attacking arguments for existential rules. Third, we studied argumentation techniques based on ranking-based approaches in both the context of query answering and argumentation reasoning.
I followed a Master program called "MIT" at the University of Montpelier. This program offered courses about Theoretical Computer Science. An non exhaustive list of the topics covered are:
As a first year undergraduate, I studied Computer Science and Mathematics at the University of French Polynesia. In 2012, I moved to Montpelier and entered a course of study focused on Mathematics as a second year undergraduate. An non exhaustive list of the topics covered are:
Bruno Yun, Madalina Croitoru, Pierre Bisquert and Srdjan Vesic
DAGGER is a generator for logic based argumentation frameworks built on inconsistent knowledge bases expressed using existential rules (i.e. Datalog +/-). The tool uses an input in .dlgp format and allows to visualise the corresponding argumentation graph. Furthermore, the argumentation framework can also be downloaded in the Aspartix format (.apx). The user can start by constructing any knowledge base or loading a predefined knowledge base, or both. DAGGER is programmed in a way that it captures the deductive argumentation framework of Croitoru et al. (2013) and the maximal consistent sets (or repairs) of the knowledge base, well known to be equivalent to the preferred and stable semantics in this setting. DAGGER enables a User (or an Expert) to simply construct the argumentation graph corresponding to an inconsistent Datalog+/- knowledge base and get a preview of its structure. The layered architecture is shown in the figure below and it is detailed as follows:
As depicted, the User interacts with the GUI by providing a knowledge base and a computation order. Then, the GUI communicates with the GRAAL library which possesses the toolkit for handling existential rules knowledge bases. Then, the repairs are generated and used for the argument generation. The latter enables the argument filtration process and the attack generation. Next, the argumentation graph is displayed graphically using the GrahStream library or textually using the Aspartix format, or both.
Please find here the Demo submitted to AAMAS 2018 about DAGGER. The paper is entitled "DAGGER: Datalog+/- Argumentation Graph GEneRator".
The git repository is accessible online at the LIRMM's gitlabIf you use the DAGGER tool in your academic work please cite: DAGGER: Datalog+/- Argumentation Graph GEneRator, Bruno Yun, Madalina Croitoru, Pierre Bisquert and Srdjan Vesic, AAMAS 2018.
Bruno Yun, Srdjan Vesic and Madalina Croitoru
In this section, we introduce NAKED: a generator for logic-based argumentation hypergraphs built on inconsistent knowledge bases expressed using existential rules (i.e. Datalog +/-). The tool uses an input in .dlgp format and allows to visualise the corresponding argumentation graph. We offer an interactive method for observing specific arguments and the possibility to export the graph in the DOT format. The user can start by constructing any knowledge base or loading a predefined knowledge base, or both. NAKED builds the arguments and attacks in a novel way: Arguments are either facts or constructed upon other arguments (à la ASPIC+) and sets of consistent arguments can attack a single other argument. We also provide the maximal consistent sets (or repairs) of the knowledge base, well known to be equivalent to the preferred and stable semantics in this setting. DAGGER enables a User (or an Expert) to simply construct the argumentation graph corresponding to an inconsistent Datalog+/- knowledge base and get a preview of its structure. The User interacts with the GUI by providing a knowledge base and a computation order. Then, the GUI communicates with the GRAAL library which possesses the toolkit for handling existential rules knowledge bases. Then, the repairs are computed using an efficient algorithm and used for the argument generation. Then, the argumentation graph is displayed graphically using the GrahStream library or textually using the DOT format, or both.
The GUI below shows NAKED drawing the argumentation graph corresponding to a simple example.
Please find here the Demo submitted to AAMAS 2019 about NAKED. The paper is entitled "NAKED: N-Ary graphs from Knowledge bases Expressed in Datalog±".
The git repository is accessible online at the LIRMM's gitlab
In the ReEntrust project, we aim at rebuilding users' trust on online platforms after a breakdown. In order to achieve this goal, we first have to (1) identify the elements that generate these breakdowns, (2) analyse the effect of several factors such as digital literacy or age range in these processes and (3) gather valuable data on the mental patterns that a user goes through when using these platforms. We first chose to focus on recommender systems applied to the particular case of hotel booking. We designed a fake booking website where users where told to book a hotel room in Paris within a designed budget and for a specific date. This fake booking website was designed so that it would look like a real booking platform with the difference that it has several trust-breaking features:
The git repository is accessible online on my github page.
Jun Zhao, Bruno Yun, Bénédicte Legastelois, Menisha Patel, Helena Webb, Michael Rovatsos, Marina Jirotka
Algorithm Playground is about investigating what makes algorithms more trustworthy. It aims at bringing a human perspective on decisions made by recommender systems. This is a tool developed by the Edinburgh Team involved in the ReEnTrust Project. Algorithm Playground gives users the chance to discover how the algorithms influence their choices when looking for a hotel to book. In this tool, the users can perform several tasks:
Please find below a demo video about Algorithm Playground.
The git repository is accessible online on Github
Mirjam Brady talked about her reflections on blended teaching in the second semester.
I participated to discussion panel about finding new ways to encourage student engagement online.
I participated to the Designing and Tutoring Online Course (DTOC) LITE online event (5 days). This course provided inspiration and guidance on using the tools within MyAberdeen to create an engaging and well structured course area for students.
I presented the two papers the following two papers:
I served as an internal examiner for Rui Mao's thesis: "Computational Metaphor Processing".
I was a program committee member for the 6th International Workshop on Graph Structures for Knowledge Representation and Reasoning.
I was a program chair member for the 25th International Conference on Conceptual Structures in Bolzano, Italy.
I participated to the UKRI Bootcamp. It has been designed to enable researchers to transform their creative ideas into a draft proposal that can be submitted to a UK Research and Innovation (UKRI) Research Council.
I attended the New Supervisor Training. The topics covered were:
I participated to the Trust, Privacy & The Internet of Things Early Career Workshop at the Ardoe House Hotel and Spa.
I was responsible for reviewing 5-10 papers.
I reviewed a paper for the 2019 PRIMA Conference.
Participation to the 3rd Summer School on Argumentation: Computational and Linguistic Perspectives at the Warsaw University of Technology, Poland. Best SSA 2018 paper award for presenting "How Can You Mend A Broken Inconsistent KB in Existential Rules Using Argumentation".
Participation to the 26th Doctiss at the St Priest Campus, Montpelier, France.
Oral presentation about the Pack4Fresh project, for the UMR IATE meeting, at Le Grand Arbre restaurant in Montpellier, France. [slides]
Visit of the CRIL for the APOLONIO (PEPS Project), at the Artois University in Lens, France.
Visit of the CRIL for the GDR IA Project, at the Artois University in Lens, France.
Poster presentation for the Sixteenth International Symposium on Intelligent Data Analysis (IDA 2017), at the Woburn House in London, UK.
Oral presentation for the 10th Meeting of the AMANDES Project, at the University of Paris Descartes, France.
Poster presentation for the 25th Anniversary of the LIRMM at the Château d'Ô Domain of Montpellier, France.
Participation to the Third Nordic Logic Summer School at the Department of Mathematics of Stockholm, Sweden.
Organiser of the 25th Doctiss at the St Priest Campus, Montpellier, France .
Organizing member of the 30th International Workshop on Description Logics, Montpellier, France.
AAMAS 2017 Conference at Sao Paulo - (Student scholarship).
Module ED: Statistique pour expérimentateurs.
Workshop - Intégration de données, connaissances et modèles. Atelier Protégé et Atelier Argumentation.