6.1.1 InteGraal

• Name:
  InteGraal : Knowledge-Representation and Reasoning for Data Integration
• Keywords:
  Knowledge Bases, Data integration, Knowledge representation, Automated Reasoning, Heterogeneous Data, Knowledge Graphs
• Scientific Description:
  InteGraal is a tool for integrating and reasoning on heterogeneous and federated data. The tool embodies algorithms and techniques developed at the crossroads between the fields of knowledge representation and reasoning and data management. From the historic point of view, this tool is the result of a complete re-engineering of the Graal tool, whose API and functionalities have been completely updated. Also, with respect to Graal, the tool is very much oriented towards data integration.
• Functional Description:
  InteGraal has been designed in a modular way, in order to facilitate software reuse and extension. It should make it easy to test new scenarios and techniques, in particular by combining algorithms. The main features of Graal are the following: (1) internal storage to store data by using a SQL or RDF representation (Postgres, MySQL, HSQL, SQLite, Remote SPARQL endpoints, Local in-memory triplestores) as well as a native in-memory representation (2) data-integration capabilities for exploiting federated heterogeneous data-sources through mappings able to target systems such as SQL, RDF, and black-box (eg. Web-APIs) (3) algorithms for query-answering over heterogeneous and federated data based on query rewriting and/or forward chaining (or chase)
• Release Contributions:

  2024. Added reasoning with stratified negation. Advanced usability and features for mappings for integrating heterogeneous data. Extended the command line interface.

  2023. Mappings for integrating heterogeneous data. Compilation-based query rewriting. Command line interface.

  2022: First release, software deposit with Apache 2 licence.

  2021: Functional specification, design and development of a major improved version of the tool. Started refactoring of the API, and of several modules for knowledge base representation, data storage, query answering and forward-chaining reasoning (chase). Started the development of new modules for handling heterogeneous data: mappings and federations.

• News of the Year:
  This year we continued working on the major module including mappings for integrating heterogeneous data. We also added the support for reasoning with negation. Finally, we also improved our command line interface to interact with the platform. Moreover, the tool has been used in several internship and thesis of the team, as well as for collaborations with our research partners.
• URL:
  https://­gitlab.­inria.­fr/­rules/­integraal
• Publication:
  https://hal.science/INRIA2/lirmm-04304601v1
• Contact:
  Federico Ulliana
• Participants:
  Jean-Francois Baget, Pierre Bisquert, Guillaume Perution-Kihli, Michel Leclère, Marie-Laure Mugnier, Florent Tornil, Federico Ulliana

6.1.2 TreeForce

• Keywords:
  JSon, Databases, Knowledge Bases, Automated Reasoning, Rewriting, NoSQL, Data integration, Knowledge representation, Heterogeneous Data
• Scientific Description:
  TreeForce is a java tool for reasoning on tree data. It leverages on query rewriting techniques and NoSQL document oriented key-value stores. This library can be seen as a general toolbox for implementing reasoning techniques tailored for tree-shaped data and rules. It is composed of two main modules. The first includes generic data structures and algorithms for trees and tree-automata. The second includes automata-based query rewriting techniques as well as efficient evaluation techniques for large sets of rewritings.
• Functional Description:
  TreeForce is a java tool for reasoning on tree data. It leverages on query rewriting techniques and NoSQL document oriented key-value stores. This library can be seen as a general toolbox for implementing reasoning techniques tailored for tree-shaped data and rules. It is composed of two main modules. The first includes generic data structures and algorithms for trees and tree-automata. The second includes automata-based query rewriting techniques as well as efficient evaluation techniques for large sets of rewritings.
• Release Contributions:
  2023. ArangoDB wrapper. Code improvement. 2022. Novel instance-aware rewriting and evaluation algorithms. Introduced summarization, partitioning and parallelization techniques. 2021: First version of TreeForce. Automata for unordered tree languages. Automata-based query-rewriting algorithms. MongoDB wrapper.
• Contact:
  Federico Ulliana
• Participants:
  Olivier Rodriguez, Federico Ulliana

6.1.3 B-Runner

• Name:
  B-Runner
• Keywords:
  Benchmarking, Experimentation, Java, Automated Reasoning, Databases, Knowledge Graphs
• Scientific Description:
  B-Runner is a Java tool for the conduction of experimental analysis.
• Functional Description:
  Runner is a library for collaborative benchmarking on knowledge and rule-based reasoners. The motivation for this project was to systematize testing, both on on InteGraal and other reasoners. The goal of B-Runner is to enable benchmarking for reasoning tools with a small cost, high robustness, and repeatability guarantees. The tool can be used as a best-practice for realizing and communicating on experimental analysis.
• Release Contributions:
  Core module for experiment conduction
• News of the Year:
  First release of the tool.
• URL:
  https://­gitlab.­inria.­fr/­rules/­brunner
• Contact:
  Federico Ulliana
• Participants:
  Federico Ulliana, Pierre Bisquert, Florent Tornil, Renaud Colin, Quentin Yeche, Akira Charoensit

Existential rules.

Existential rules are first-order-logic formulas representing implications of the form $\forall X\forall Y\left[\mathrm{𝐵𝑜𝑑𝑦}(X,Y)\to \exists Z\mathrm{𝐻𝑒𝑎𝑑}(X,Z)\right]$ where Body and Head are positive conjunctions of atoms without functional symbols, and Head can have existentially quantified variables. These rules allow one to model complex relationships over the domains of interest, and at the same time dispose of a value invention mechanism through existentially quantified variables. This makes them suitable for many data and knowledge tasks on both open and closed domains. As a result, existential rules are ubiquitous in many fields. They are used to model dependencies, schema mappings, and expressive queries in databases. They are used as ontological languages as a valid complement to Description Logics, and at the same time as a generalization of so-called Horn Description Logics which lay at the foundations of important Semantic Web standards.

Rule-based query answering

Given a query $Q$, a database $D$, and a set of rules $R$, query answering asks to determine whether $D,R\vDash Q$ (where $\vDash$ denotes standard first-order logic entailment), that is if the query $Q$ is a logical consequence of the knowledge base made by the database $D$ and the rules $R$. In the field of knowledge representation and reasoning, rule-based query answering is studied for rules expressing ontologies and referred as ontology-mediated query answering (OMQA). Formalisms such as Description Logics and Existential Rules (a.k.a, Tuple-Generating-Dependencies, or Datalog$\pm$) are typically targeted for expressing ontologies. Overall, the main emphasis of this topic is in the study of rule languages and the role they play in query answering.

Rule-based query answering over heterogeneous and federated data

In this context, the problem formulation remains similar, however the database $D$ is replaced by a more complex notion of federation$(𝒟,ℳ,𝒮)$ where $𝒟$ is a collection of heterogeneous data sources, $𝒮$ is a global integration schema, and $ℳ$ is a set of mappings linking the datasources in $𝒟$ to the global schema $𝒮$. This framework is at the foundations of data integration (DI) in databases and of ontology-based data access (OBDA) in knowledge representation and reasoning. OBDA focuses on global integration schemes and rules built on ontologies enabling query rewriting, while DI is more concerned with rules representing data-dependencies. Overall, both give more emphasis to heterogeneous and federated data in rule-based query answering.

Reasoning strategies for query answering

The two prominent strategies for rule-based query answering are materialization (also known as saturation, or forward-chaining) and virtualization (also known as query rewriting, or backward chaining). Both can be seen as ways of reducing query answering (which involves reasoning) into classical query evaluation. Materialization amounts to storing the inferences enabled by rules, thereby obtaining an extended database, on which queries are evaluated. Query rewriting amounts to compiling relevant rules into the query, thereby obtaining a rewritten query (usually a union of queries), which is evaluated on the (unaltered) database. Both approaches have their own strengths, and at the basis of this duality is the fact that while materialization is independent of queries, rewriting is independent of the database. Hence, each strategy better suits certain applicative scenarios, and both can possibly be combined thereby resulting in hybrid approaches.

7.5.1 Reasoning with Existential Rules

Ontology-based (or rule-based) query answering is a problem that takes as input an ontology $R$ (i.e., a set of existential rules in our context), a set $D$ of facts, and a Boolean conjunctive query (CQ) $Q$, and asks whether $R,D\vDash Q$, where $\vDash$ denotes standard first-order logic entailment. This problem is undecidable in general, and a widely investigated approach to tackle it in some cases is query rewriting: given some “rule query” $\langle R,Q\rangle$, we compute a Boolean query $Q_R$ such that, for any fact set $D$, it holds that $R,D\vDash Q$ if and only if $D\vDash Q_R$. Insofar, previous work has mostly focused on output queries $Q_R$ expressed as union of Boolean conjunctive queries (UCQs), and an effective algorithm that computes such a query $Q_R$ whenever it exists has been proposed in the literature. However, UCQ-rewritability is not a very general notion and many real-world interesting rule queries do no admit UCQ-rewritings. This raises the question whether such a generic algorithm can be designed for a more expressive target language, such as datalog. We solved this question by the negative, by studying the difference between datalog-expressibility and datalog-rewritability. More precisely, we showed that query answering under datalog-expressible rule queries is undecidable.

• Published at the International Conference on Principles of Knowledge Representation and Reasoning (KR-2024) 13 with Michaël Thomazo (Inria VALDA)

We studied the problem of computing explanations for queries under datalog knowledge bases. One of the major benefits of rules (and symbolic AI methods in general) is explainability. When new knowledge is obtained via a reasoning process, it is possible to determine precisely all elements of the knowledge base that yield this knowledge. Typically, one would use a SAT (Boolean satisfiability problem) solver to compute the explanations. However, SAT-solving is computationally expensive, and as the knowledge base grows, the time required increases exponentially. In our work, we presented a method for filtering a datalog knowledge base to optimize the time used by a SAT solver. This is achieved by creating a hypergraph representing the grounded knowledge base and pruning the nodes that are not reachable from the fact that one wants to explain. The approach proved to be time-effective. Interestingly, one additional benefit of using this hypergraph is that it is possible to encode more information about the rules used in the reasoning process. By using an off-the-shelf group-SAT solver, this extra information allows us to find specific explanations that would be missed if we only considered facts.

• Published at the International Joint Conference on Rules and Reasoning (RuleML+RR 2024) 14.

7.5.2 Non-Monotonic Reasoning

Answer set programming (ASP) is a non-monotonic logic programming formalism used in various areas of artificial intelligence like combinatorial problem solving and knowledge representation and reasoning. It is known that enhancing ASP with function symbols makes basic reasoning problems highly undecidable. However, even in simple cases, state of the art reasoners, specifically those relying on a ground-and-solve approach, fail to produce a result. Therefore, we reconsidered consistency as a basic reasoning problem for ASP. We showed reductions that give an intuition for the high level of undecidability. These insights allow for a more fine-grained analysis where we characterize ASP programs as “frugal” and “non-proliferous”. For such programs, we were not only able to semi-decide consistency but we also proposed a grounding procedure that yields finite groundings on more ASP programs with the concept of “forbidden” facts.

• Published at the International Joint Conferences on Artificial Intelligence (IJCAI 2024) 18 and the International Workshop on Nonmonotonic Reasoning (NMR 2024) 17 with Lukas Gerlach (Technische Universität Dresden) and Markus Hecher (MIT, Boston).

We investigated the evaluation of queries in non-monotonic propositional logic. In order to avoid ambiguity and be efficient, the context in which a query is made can help to better target the relevant pieces of information from the knowledge base to be processed by the inference system. We studied a notion of "dynamical compartmentalization" where the knowledge base that will be used for reasoning is dynamically extracted from the original base. Compartmentalization is a selection of a sub-base which is done according to a function, called "refiner", and depending on this function some properties are satisfied. We introduced a particular "syntactic refiner" that uses a similarity symbol-based distance between a context (a multiset of variable symbols) and a formula of a knowledge base. We proved that the inference operator based on this refiner, called "contextual inference", satisfies a series of desirable axioms.

• Published at the European Conference on Artificial Intelligence (ECAI 2024) 15 with Florence Dupin de Saint-Cyr (IRIT).

7.6.1 Benchmarking Reasoning Systems

Conducting experimental analysis on rule reasoners is a mainstream task for validating novel algorithms and systems. Nevertheless, providing robust, verifiable, and reproducible experiments can still raise a sensible challenge. We introduced B-Runner, an open library for collaborative benchmarking focusing on the deployment of articulate tests for knowledge and rule-based systems with low cost and high robustness. B-Runner reduces the benchmarking setup time while guaranteeing experiment repeatability. At the same time, it improves the scrutability of experimental protocols thereby enhancing their robustness as well as fairness of system comparisons.

• Published at the International Joint Conference on Rules and Reasoning (RuleML+RR 2024) 19 and at Bases de Données Avancées (BDA-2024) 20 with Quentin Yeche (Inria EVERGREEN).

7.6.2 Exploring the Application of Ontology Based Data Access in Agroecology

As part of Elie Najm’s PhD thesis (funded by #DigitAg) and in collaboration with the INRAE ABSys research lab in agroecology, we have investigated issues related to the design of sustainable agroecosystems based on agroecology. Agroecology leads to more complex agroecosystems including a higher number of plant species, whether cash crops or service crops (i.e., plants providing various ecosystem services). The design of these new agrosystems requires to integrate various data as well as unstabilized scientific knowledge. We more specifically considered the issue of selecting service plant species according to their potential to provide ecosystem services. To tackle that issue, we adopted an approach based both on a formalized representation of domain knowledge and on the exploitation of available data, collected independently from the targeted application. More specifically, we rely on the one hand on recent scientific results in agronomy linking functional traits (i.e., measurable characteristics of plant species) to ecosystem services, and on the other hand on data about functional traits collected by the research community in ecology. The architecture of our system reprises that of rule-based query answering over heterogeneous and federated data we already presented. We provide a methodology to acquire scientific knowledge in the form of diagrams linked to data sources, as well as a formalization in a logical rule-based language. Importantly, our rules are independent from specific diagrams and data, to ensure better genericity and system’s ability to evolve. We carried out an experimental evaluation of our system on the use case of vine grassing, i.e., installing herbaceous service plants in vineyards, which showed that very satisfactory results can be obtained as long as the proportion of missing data values is not too high.

• Published in Computers and Electronics in Agriculture 12 as a joint work with Christian Gary, Raphael Metral, and Léo Garcia (INRAE).

Following this work, and as part of Guillaume Pérution's postdoc, we have performed an extensive rework of this application with the aim of improving the quality of both the code (bug fixes, extensibility, genericity) and the modeling (declarativity, genericity). In addition, the new implementation essentially relies on Integraal, which has also led to the evolution of Integraal's functionalities.

8.1.1 Visits of international scientists

8.1.2 Visits to international teams

8.1.3 Visits to national teams

CQFD (ANR PRC, Jan. 2019-Dec. 2024)

Participants: Jean-François Baget, Pierre Bisquert, Nofar Carmeli, David Carral, Michel Leclère, Marie-Laure Mugnier, Guillaume Pérution-Kihli, Florent Tornil, Federico Ulliana.

CQFD (Complex ontological Queries over Federated heterogeneous Data), coordinated by Federico Ulliana (BOREAL), involves participants from Inria Saclay (CEDAR team), Inria Paris (VALDA team), Inria Nord Europe (SPIRALS team), IRISA, LIG, LTCI, and LaBRI. The aim of this project is to tackle two crucial challenges in OMQA (Ontology Mediated Query Answering), namely, heterogeneity, that is, the possibility to deal with multiple types of data-sources and database management systems, and federation, that is, the possibility of cross-querying a collection of heterogeneous datasources. By featuring 8 different partners in France, this project aims at consolidating a national community of researchers around the OMQA issue.

www.lirmm.fr/cqfd/

Convergence institute #DigitAg (2017-2026)

Participants: Jean-François Baget, Marie-Laure Mugnier, Federico Ulliana.

Located in Montpellier, #DigitAg (for Digital Agriculture) gathers 17 founding members: research institutes, including Inria, the University of Montpellier and higher-education institutes in agronomy, transfer structures and companies. Its objective is to support the development of digital agriculture. BOREAL is involved in this project on the issues of designing data and knowledge management systems adapted to agricultural information systems, and of developing methods for integrating different types of information and knowledge (generated from data, experts, models). A PhD thesis (Elie Najm, 2019-2022) investigated knowledge representation and reasoning for the design of new agroecological systems, in collaboration with the research laboratory ABSys - Biodiversified Agrosystems (formerly UMR SYSTEM).

www.hdigitag.fr/eng/

9.1.1 Scientific events: organisation

9.1.2 Scientific events: selection

9.1.3 Journal

9.1.4 Scientific expertise

• Marie-Laure Mugnier was a member of the selection committee for a Professorship position at the University of Montpellier.

9.1.5 Research administration

• President of the “Section 27 Commitee” (Computer Science) of the University of Montpellier (since July 2021): Marie-Laure Mugnier
• Member of the Council and Human Ressources Commission of the Scientific Pole MIPS (Mathematics Informatics Physics and Systems) of the University of Montpellier (since its creation): Marie-Laure Mugnier

9.2.1 Teaching

• Michel Leclère and Marie-Laure Mugnier teach at the Computer Science department of the Science Faculty. They are in charge of courses in Programming and Logics (Licence), as well as Symbolic Artificial Intelligence, Semantic Management of Data, and Theory of Data and Knowledge Bases (Master).
• Concerning full-time researchers in 2024, Jean-François Baget, Nofar Carmeli, David Carral, and Federico Ulliana (on secondment from Montpellier University) taught in the Computer Science Master about topics including Theory of Data and Knowledge Bases, Datawarehouses, Big-Data and NoSQL systems.
• Guillaume Pérution-Kihli has been coaching students from University of Montpellier for the SWERC (Southwestern Europe Regional Contest)1 competition since 2020.

9.2.2 Supervision

• Pierre Bisquert, David Carral, and Federico Ullliana continue to co-supervise Akira Charoensit, who is a PhD student based in our group and is about to finish her second year.
• David Carral continues to co-supervise Lucas Larroque with Michaël Thomazo. Lucas is a PhD student based at the ENS Paris who has recently started the second year of his degree.
• Nofar Carmeli and David Carral supervised Noah Collinet, who was an M2 student of the University of Montpellier, during her master's project.
• Michel Leclère and Jean-François Baget supervised Maksym Lytvynenko, who is a master's student at the University of Montpellier. Namely, they worked together on an M1 research project.
• Nofar Carmeli and David Carral supervised François Colin De Verdiere and Clément Rouvroy, who are two L3 students from the ENS of Lyon and Paris, respectively. They visited our group to do a research internship on topics related with database theory that lasted 6 weeks.
• Nofar Carmeli supervised Nicolas Valayannopoulos-Akrivou, who is an L2 student from MIT on a research internship on topics related with database theory.

9.2.3 Juries

9.3.1 Productions (articles, videos, podcasts, serious games, ...)

• Michel Chein wrote the article "Et si certains scientifiques, en Intelligence Artificielle, avaient une responsabilité dans le défiance vis à vis de la raison ?", which appeared in the 54th volume of the Bulletin de l'Académie des Sciences et Lettres de Montpellier. For more info, see this link.
• Michel Chein wrote the article "Penser l'Intelligence Artificielle (les agents conversationnels)", which will appear in the 55th volume of the Bulletin de l'Académie des Sciences et Lettres de Montpellier. For more info, see this link.
• Michel Chein wrote the article "Émergence de l'Informatique à Montpellier" in the brochure for the "45 ans du Centre National de Calcul Scientifique à Montpellier (CINES)". This publication is an extension of his previous article "Sur la science informatique et son installation à Montpellier", which was published in the 48th volumen of the Bulletin de l'Académie des Sciences et Lettres de Montpellier. For more info, see this link.

9.3.2 Participation in Live events

• Marie-Laure Mugnier gave an interview about symbolic and hybrid AI to the magazine LUM of the university of Montpellier. This led to an article in the special issue about AI published on 07/10/2024 (LUM n°22) and to a podcast. For more info, see this link.

International journals

• 10 articleK.Karl Bringmann, N.Nofar Carmeli and S.Stefan Mengel. Tight Fine-Grained Bounds for Direct Access on Join Queries.ACM Transactions on Database SystemsDecember 2024HALDOIback to text
• 11 articleN.Nofar Carmeli, M.Martin Grohe, B.Benny Kimelfeld, E.Ester Livshits and M.Muhammad Tibi. Database Repairing with Soft Functional Dependencies.ACM Transactions on Database Systems4922024, 1-34/8HALDOIback to text
• 12 articleE.Elie Najm, M.-L.Marie-Laure Mugnier, C.Christian Gary, J.-F.Jean-François Baget, R.Raphael Métral and L.Léo Garcia. Integrating data and knowledge to support the selection of service plant species in agroecology.Computers and Electronics in Agriculture217February 2024, 108594HALDOIback to text

International peer-reviewed conferences

• 13 inproceedingsD.David Carral, L.Lucas Larroque and M.Michaël Thomazo. Ontology-Based Query Answering over Datalog-Expressible Rule Sets is Undecidable.KR 2024 - 21st International Conference on Principles of Knowledge Representation and ReasoningHanoi, VietnamNovember 2024HALback to text
• 14 inproceedingsA.Akira Charoensit, D.David Carral, P.Pierre Bisquert, L.Lucas Rouquette and F.Federico Ulliana. Rule-aware Datalog Fact Explanation Using Group-SAT Solver.RuleML+RR’24: Companion Proceedings of the 8th International Joint Conference on Rules and ReasoningRuleML+RR 2024 - 8th International Joint Conference on Rules and ReasoningBucarest, RomaniaSeptember 2024HALback to text
• 15 inproceedingsF.Florence Dupin de Saint-Cyr and P.Pierre Bisquert. The Form and the Content: Non-Monotonic Reasoning with Syntactic Contextual Filtering.ECAI 2024ECAI 2024 - 27th European Conference on Artificial IntelligenceSantiago de Compostela, SpainIOS PressAugust 2024, à paraîtreHALback to text
• 16 inproceedingsI.Idan Eldar, N.Nofar Carmeli and B.Benny Kimelfeld. Direct Access for Answers to Conjunctive Queries with Aggregation.ICDT 2024 - 27th International Conference on Database TheoryLeibniz International Proceedings in InformaticsPaestum, ItalySchloss Dagstuhl – Leibniz-Zentrum für Informatik2024, 4:1-4:20HALDOIback to text
• 17 inproceedingsL.Lukas Gerlach, D.David Carral and M.Markus Hecher. Finite Groundings for ASP with Functions: A Journey through Consistency (Extended Abstract).CEUR Workshop ProceedingsNMR 2024 - 22nd International Workshop on Nonmonotonic ReasoningCEUR-3835Nonmonotonic Reasoning 2024 : Proceedings of the 22nd International Workshop on Nonmonotonic Reasoning (NMR 2024)Hanoi, VietnamNovember 2024, 183-186HALback to text
• 18 inproceedingsL.Lukas Gerlach, D.David Carral and M.Markus Hecher. Finite Groundings for ASP with Functions: A Journey through Consistency.IJCAI 2024 - 33rd International Joint Conference on Artificial IntelligenceJeju, South Korea2024, 3386-3394HALDOIback to text
• 19 inproceedingsF.Federico Ulliana, P.Pierre Bisquert, A.Akira Charoensit, R.Renaud Colin, F.Florent Tornil and Q.Quentin Yeche. Collaborative Benchmarking Rule-Reasoners with B-Runner.RuleML+RR 2024 - 8th International Joint Conference on Rules and Reasoning15183Lecture Notes in Computer ScienceBucarest, RomaniaSeptember 2024, 21-31HALDOIback to text

National peer-reviewed Conferences

• 20 inproceedingsF.Federico Ulliana, P.Pierre Bisquert, A.Akira Charoensit, R.Renaud Colin, F.Florent Tornil and Q.Quentin Yeche. Collaborative Benchmarking Rule-Reasoners with B-Runner.BDA 2024 - 40e Conférence sur la Gestion de Données – Principes, Technologies et ApplicationsOrleans, FranceOctober 2024HALback to text

Reports & preprints

• 21 miscP.Pierre Bisquert, P.Patrice Buche, A.Abdelraouf Hecham and M.Maksim Koptelov. Justified collective decision-making in agri-food systems: an approach based on computational social choice, argumentation methods and tools (Supplementary Material).June 2024HAL