Simple conceptual graphs can be seen as a very basic description logic, allowing however for answ... more Simple conceptual graphs can be seen as a very basic description logic, allowing however for answering conjunctive queries. In the first part of this paper, we translate some results obtained for conceptual graph rules of form "if A then B" into an equivalent DL-based formalism. Then we show that, our algorithms can automatically decide in some cases whether a given DL has the FOL-reducibility property, provided that the semantics of its constructors can be expressed by rules.
This paper answers the SCG-1 initiative. The room allocation problem provided has been solved in ... more This paper answers the SCG-1 initiative. The room allocation problem provided has been solved in a generic and automatic way. The solution is based on a totally declarative formal model. Basic constructs are simple graphs and the fundamental operation for doing reasonings is the graph morphism known as projection. The other formal constructs are rules and constraints defined in terms of simple graphs. The modeling framework built upon the formal model allows one to describe a problem with asserted facts, rules representing implicit knowledge about the domain, validity constraints and rules transforming the world. A prototype implementing this framework has been built upon the tool CoGITaNT. It has been used to test our modelization of the room allocation problem.
We consider existential rules (also called Tuple-Generating Dependencies or Datalog+/-rules). The... more We consider existential rules (also called Tuple-Generating Dependencies or Datalog+/-rules). These rules are particularly well-suited to the timely ontological query answering problem, which consists of querying data while taking terminological knowledge into account. Since this problem is not decidable in general, various conditions ensuring decidability have been proposed in the literature. In this paper, we focus on conditions that restrict the way rules may interact to ensure that the forward chaining mechanism is finite. After a review of existing proposals, we propose a generalization of the notion of rule dependency, namely k-dependency, that allows to enlarge halting cases. It can also be used to compile the rule base, which leads to improve query answering algorithms.
Griwes is an initiative to develop a common model and an opensource freeware platform shared by d... more Griwes is an initiative to develop a common model and an opensource freeware platform shared by different graph-based frameworks. We provide an overview of its objectives, architecture and specifications. We detail some of the basic mathematical structures that are used to characterize the primitives for graph-based knowledge representation. We then propose to factorize recurrent knowledge representation primitives that can be shared across specific graph-based languages and we provide a proof of concept by showing how two languages (Simple Conceptual Graphs and RDF) can be described in this framework.
This paper reports a collective reflection led in our team about conceptual graph benchmarks. We ... more This paper reports a collective reflection led in our team about conceptual graph benchmarks. We tackle four issues for which agreement should be obtained before benchmarks can be built: what are the fragments of CGs considered? How is information exchanged? What are the problems to be solved? What kinds of tool properties are eval- uated by the benchmarks? We define a basic building block built upon simple conceptual graphs. Finally we propose to provide a first bench- mark adapted from an industrial case study. This benchmark is composed on very simple structures and should allow to focus on interoperability issues.
We present a purely declarative graph-based formal model, which could be the basis of a generic m... more We present a purely declarative graph-based formal model, which could be the basis of a generic modeling framework. Indeed, all kinds of knowledge, be it facts, queries, rules or constraints are represented by labelled graphs. Furthermore, the originality of this model is that reasonings are performed by simple to understand graph operations. Thus, not only knowledge is represented in an easily readable format, but the problem solving can be followed step-by-step by the user in the same graphical format. We show this graph-based approach at work in modeling a problem solving process (the Sisyphus-I case study), and explain why in our opinion this model is suited to knowledge acquisition and specifically to knowledge engineering.
Principles of Knowledge Representation and Reasoning, 2010
We consider positive rules in which the conclusion may contain existentially quantified variables... more We consider positive rules in which the conclusion may contain existentially quantified variables, which makes reasoning tasks (such as Deduction) undecidable. These rules have the same logical form as TGD (tuple-generating dependencies) in databases and as conceptual graph rules. The aim of this paper is to provide a clearer picture of the frontier between decidability and non-decidability of reasoning with these rules. We show that Deduction remains undecidable with a single rule; then we show that none of the known abstract decidable classes is recognizable. Turning our attention to concrete decidable classes, we provide new classes and classify all known classes by inclusion. Finally, we study, in a systematic way, the question "given two decidable sets of rules, is their union decidable?", and provide an answer for all known decidable cases except one.
International Joint Conference on Artificial Intelligence, 2009
In ∀∃-rules, the conclusion may contain existen- tially quantified variables, which makes reason-... more In ∀∃-rules, the conclusion may contain existen- tially quantified variables, which makes reason- ing tasks (as deduction) non-decidable. These rules have the same logical form as TGD (tuple- generating dependencies) in databases and as con- ceptual graph rules. We extend known decidable cases by combining backward and forward chain- ing schemes, in association with a graph that cap- tures exactly
Uploads
Papers by J. Baget