Knowledge engineering

Jump to: navigation, search
Knowledge Engineering Group

Within the BFO team since July 2012, the Knowledge Engineering group, led by Cecilia Zanni-Merk, basically focuses on the design and implementation of formal models for the development of Knowledge Based Systems (KBS). A KBS is software that reproduces the behavior of a human expert performing an intellectual task in a specific area. It is based on the explicit nature of knowledge, which is formalized in different ways. Among these formal models, ontologies are formalized and structured representations of the vocabulary specific to a certain area of study. Ontologies are commonly used with a set of rules which are chained to simulate the reasoning of a human expert.

The originality of the works developed within the Knowledge Engineering group is founded on the proposal of a layered architecture Knowledge, Rules and Experience to manage the complexity of the development of a KBS. The Knowledge layer contains domain ontologies, the Rules layer allows different types of reasoning (monotonic, spatial, temporal, fuzzy, or others) depending on the application; and finally, the Experience layer allows the capitalization and reuse of prior knowledge, either by the use of CBR (Case-Based Reasoning) or SOEKS (Set of Experience knowledge Structure) and DDNA (Decisional DNA) or by other mechanisms.

The main areas of application of these tools are:

  • Analysis of remote sensing images
  • Analysis and diagnosis of SMEs
  • Computer aided inventive design
  • Environment and Sustainable Development

Current Projets and Collaborations

Optimization guided by ontologies and by the capitalization of prior knowledge (in collaboration with the SONIC group of the BFO team

This project seeks to optimize the journeys undertaken by a fleet of vehicles in an area, based on a large number of constraints (traffic conditions, vehicle type, "social" or "union" constraints, etc.). In this context, the structuring of the data needed to properly represent a situation and to take into account the constraints with an approach oriented towards multi-criteria optimization becomes essential. We propose to knowledge engineering techniques, particularly the creation of a domain ontology to formalize the model of the knowledge base.

In order to improve the performance of the genetic algorithm, we are also interested in the formalization of decisional knowledge to guide the solutions to these practical optimization problems, by the capitalization of e already acquired knowledge during precedent uses of the developed optimization algorithm and the obtained "results".

Semantic Region Labeling for Remote Sensing Image Interpretation (in collaboration with the Data Mining group of the BFO team

The increasing availability of High Spatial Resolution satellite images is an opportunity to characterize and identify urban objects. Image analyses methods using object-based approaches based on the use of domain knowledge, are necessary to classify data. A major issue in these approaches is domain knowledge formalization and exploitation. The use of formal ontologies seems a judicious choice to deal with these issues. Therefore, the aim of these works is to highlight the benefits in the use of a thematic ontology for automatic regions labelling.

Description logics (DL) are being used to exploit the knowledge in the ontologies and develop software tools to assist the automatic labelling of satellite images. Aspects to be considered in the final approach include fuzzy logic (to address the symbol anchoring problem between the low level informations available from the image analysis software and the conceptual knowledge that has been formalized as an urban ontology), spatial and temporal reasoning.

Formalization of Theoretical and Experience Knowledge for the Analysis and Diagnosis of SMEs (in collaboration with the Design Engineering Laboratory - EA3998)

This project proposes the development of software for analysis and diagnosis of SMEs. In particular, the project focuses on the articulation of theoretical models coming from management science and specific cases about real companies; because during their development, SMEs find themselves, sometimes in situations that are not fully consistent with the theoretical models.

The theoretical models coming from the management sciences will be formalized as independent domain ontologies that can be, sometimes, contradictory with each other. The establishment of a detailed framework to manage the complexity of these model is needed. Capitalization of the knowledge about previously studied cases is also needed, to improve the quality of the analysis and diagnosis performed.

Semantic Technologies for Computer Aided Inventive Design(in collaboration with the Design Engineering Laboratory - EA 3938)

These works focus on the modelling of the formulation and problem solving processes of TRIZ (Theory of Inventive Problem Solving). The main objective is the description of all the knowledge bases of TRIZ, to complete the model and make it wholly consistent by the definition of the missing semantic links.

This formalization should allow the development of an intelligent manager of these knowledge sources, with the aim of assisting the TRIZ expert during his activity. Indeed, during the processing of a new case, experts are brought to work with various models at different levels of abstraction. The knowledge manager should suggest the experts the use of the relevant knowledge sources, consistent with the level of abstraction of the model they are building. The manager will also be able to complete "automatically" the rest of the models, by exploiting the semantinc links obtained among the different knowledge sources.

Environment and Sustainable Development

We use case-based reasoning to assist in the management of agricultural lands. The cases in question are farms or agricultural parcels on which certain operations (crop assignment, production mode ...) are used. Knowledge collected from farmers and agricultural experts is formalized by ontologies and rule systems for retrieval and adaptation.

This work is also based on models of qualitative spatial reasoning.


The complete list of the publications of the members of the group can be found at this URL.