Agents hold knowledge, expressed formally, about their environments and the society they live in. This knowledge evolves with agent experiences. We aim at establishing experimentally how this evolution occurs and what are its consequences for knowledge (e.g., agreement, consistency, growth) and for agents (e.g., survival, knowledge increase, efficiency).
In this perspective, we use and adapt techniques developed by Luc Steels and his colleagues for studying the cultural evolution of language [Steels, 2012], i.e., the way language features can be selected or created by agents through constantly attempting to communicate. Instead of focussing on language (the communication vehicle), we consider knowledge (the hidden content of communication). So, we are interested in how agent knowledge, described formally as ontologies and alignments between these ontologies, can evolve as agents try to communicate. This knowledge may be proper to agents (like personal ontologies) or shared (like alignments between ontologies).
This raises questions such as:
This can be experimented by simulating this communication activity through language games that can reach successful communication or failure. In case of failure, agents will perform a repair action in order to improve further communication. In our case, the action may be to adapt the ontology used by the agent or modify their common alignments. The merits of repair strategies in this context are judged by the evolution of the success rate in a population over the number of games played.
We have applied this approach to ontology alignment repair, i.e., the improvement of incorrect alignments [Euzenat, 2014; 2017]. For that purpose, we performed experiments in which agents react to mistakes in alignments. Agents only know about their ontologies and alignments with others and they act in a fully decentralised way. We showed that this cultural repair approach is able to converge towards successful communication through improving the objective correctness of alignments. The obtained results are on par with a baseline of state-of-the-art alignment repair algorithms and agents can start with empty knowledge at first.
This is part of an ambitious program towards what we call cultural knowledge evolution. Many aspects of these experiments may be systematically developed. For instance, we may have agents arbitrate between maintaining alignments or adopting the ontologies of other agents. We may have agents choosing among several repair operators. We may want to change the environment in which agents live so that they have to evolve their ontologies We may want to have homogeneous populations of agents to encounter other populations.
The main goal of the position is to contribute experimenting with cultural evolution techniques in the context of distributed knowledge representation. In particular, two main lines of actions will be performed by the successful candidate:
[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), LNCS 8798:115-130, 2014 ftp://ftp.inrialpes.fr/pub/exmo/publications/euzenat2014c.pdf
[Euzenat, 2017] Jérôme Euzenat, Communication-driven ontology alignment repair and expansion, in: Proc. 26th International joint conference on artificial intelligence (IJCAI), Melbourne (AU), pp185-191, 2017 ftp://ftp.inrialpes.fr/pub/moex/papers/euzenat2017a.pdf
[Steels, 2012] Luc Steels (ed.), Experiments in cultural language evolution, John Benjamins, Amsterdam (NL), 2012
Qualification: PhD or equivalent.
Place of work: The position is located at INRIA Grenoble Rhône-Alpes, Montbonnot (near Grenoble, France) a main computer science research lab, in a stimulating research environment.
Environnement: The work will be carried out in the mOeX team common to INRIA & Laboratoire d'Informatique de Grenoble. mOeX is dedicated to study knowledge evolution through adaptation. It gather permanent researchers from the Exmo team which has taken an active part these past 15 years in the development of the semantic web and more specifically ontology matching.
Hiring date: November 2018.
Duration: 12-18 months (hiring opportunities afterwards)
Salary: From 2125 EUR/month (benefits included, net before income tax), i.e., 2653 EUR/month gross.
Contact: For further information, contact Jerome:Euzenat#inria:fr.
Contact us (do not wait).
Visit INRIA's presentation (including FAQ and forms). Apply to this position.
File: Provide Vitæ, motivation letter, 2 relevant or significant publications and reference letters.