MOSIG Master 2ND
ADVISOR: Jérôme Euzenat and Manuel Atencia
TEL: +33 (0)476 61 53 66
TEAM AND LAB: Exmo team, INRIA & LIG
MASTER PROFILE: Artificial intelligence and the web
Reference number: Proposal n°1925
Work on cultural language evolution [Steels 2012] investigates experimentally how a population of agents can evolve towards a stable language. It provides a population of agents with interaction games that are played randomly. It is possible to test hypotheses by precisely crafting the rules used by agents in games and observing the consequences. Experiments have shown how agents can converge towards a common vocabulary or grammar.
Our ambition is to adapt this successful approach to the evolution of the way agents represent knowledge, and, in particular, ontologies of their environment. We have adapated this approach to the evolution of alignments [Euzenat & Shvaiko 2013] that agents maintain across their ontologies. We obtained promising results with respect to the convergence of alignments towards a stable and correct set of alignments [Euzenat 2014]. This means that agents with heterogeneous ontologies are able to correct alignments while they communicate.
The goal of this master topic is to investigate how populations of agents can reach such a state. The main questions can be stated as follows:
It is expected that the candidate picks up one of these problems, designs experiments for testing hypotheses with respect to the evolution of agent ontologies, and studies the influence of the setting on other problems.
For that purpose, new game protocols will have to be designed and, in particular, protocols involving several populations of agents (which will provide context). As a first approximation, it could be possible to use the "magic box" tought experiment for defining context [Ghidini & Giunchiglia 2001] or to classify objects based on different features (agents seeing colors, agent seeing form, etc.) and thus generating different ontologies. We have already a framework avalable for describing such type of situations and running experiments. It will have to be extended.
This ambitious topic will not be closed within a few months and it is expected that the work be pursued in PhD covering more globally the questions raised above and new questions arising during the course of the study.
[Euzenat & Shvaiko 2013] Jérôme Euzenat, Pavel
matching, 2nd edition, Springer, Heildelberg (DE), 2013
[Euzenat 2014] Jérôme Euzenat, First experiments in cultural alignment repair (extended version), in: Proc. 3rd ESWC workshop on Debugging ontologies and ontology mappings (WoDOOM), Hersounisos (GR), Lecture notes in computer science 8798:115-130, 2014 https://exmo.inria.fr/files/publications/euzenat2014c.pdf
[Ghidini & Giunchiglia 2001] Chiara Ghidini, Fausto Giunchiglia, Local Models Semantics, or Contextual Reasoning = Locality + Compatibility, Artificial intelligence 127(2):221–259, 2001
[Steels 2012] Luc Steels (ed.), Experiments in cultural language evolution, John Benjamins, Amsterdam (NL), 2012