Catálogo de publicaciones - libros

Services and Systems must include a set of features to remain competent and future conform: intelligent behaviour, personalisation, adaptivity, scalability, manageability, ease of use and user friendliness, security, and self-healing capabilities. As a consequence, new architectural models are needed, which provide the users with access to a cognitive behaviour aspect of the system, and which may draw inspiration from the brain sciences. On the other hand, we have to use knowledge representation and semantic modeling, e.g., ontologies for representing our environment or basic properties of services and systems. This would naturally involve Agent Technology, AI, and Software Technology. So, approaches from many different disciplines have to work in integration.

With this talk we want to pay tribute to the late Professor Gerd Veenker who deserves the historic credit of initiating the formation of the German AI community. We present a summary of his scientific contributions in the context of the early approaches to theorem proving and, against this background, we point out future perspectives of Automated Deduction.

The newly established cluster of excellence investigates the realization of cognitive capabilities such as perception, learning, reasoning, planning, and execution for technical systems including humanoid robots, flexible manufacturing systems, and autonomous vehicles. In this paper we describe cognitive technical systems using a sensor-equipped kitchen with a robotic assistant as an example. We will particularly consider the role of Artificial Intelligence in the research enterprise.

Key research foci of Artificial Intelligence research in include (∘) symbolic representations grounded in perception and action, (∘) first-order probabilistic representations of actions, objects, and situations, (∘) reasoning about objects and situations in the context of everyday manipulation tasks, and (∘) the representation and revision of robot plans for everyday activity.

Artificial Intelligence is about designing and constructing artefacts, normally not about explaining human intelligence. So, why should we care about natural intelligence when talking about AI? There are several important more or less recent findings in brain science as well as ethology which require a deeper rethinking on the AI side. Based on them, the hypothesis in this talk is: The rising complexity of the behaviour system and of personalized social relationships was one of the major reasons for developing intelligence - contrary to the huge resource consumption that intelligence costs an individual. The most important result of this development was the capability of forecasting the behaviour of conspecifics for survival in a complex social environment. This capability was also useful for other purposes, including forecasting behaviour of individuals of other species and nature itself.

The Early Detection, Alert and Response (eDare) system is aimed at purifying Web traffic propagating via the premises of Network Service Providers (NSP) from malicious code. To achieve this goal, the system employs powerful network traffic scanners capable of cleaning traffic from known malicious code. The remaining traffic is monitored and Machine Learning (ML) algorithms are invoked in an attempt to pinpoint unknown malicious code exhibiting suspicious morphological patterns. Decision trees, Neural Networks and Bayesian Networks are used for static code analysis in order to determine whether a suspicious executable file actually inhabits malicious code. These algorithms are being evaluated and preliminary results are encouraging.

Knowledge of a person’s location provides important context information for many applications, ranging from services such as E911 to personal guidance systems that help cognitively-impaired individuals move safely through their community. Location information is also extremely helpful for estimating a person’s high-level activities. In this talk we show how Bayesian filtering and conditional random fields can be applied to estimate the location and activity of a person using sensors such as GPS or WiFi. The techniques track a person on graph structures that represent a street map or a skeleton of the free space in a building. We also show how to learn a user’s significant places and daily movements through the community. Our models use multiple levels of abstraction so as to bridge the gap between raw GPS measurements and high level information such as a user’s mode of transportation, her current goal, and her significant places (e.g. home or work place). Finally, we will discuss recent work on using a multi-sensor board so as to better estimate a person’s activities.

Axiom pinpointing has been introduced in description logics (DLs) to help the user understand the reasons why consequences hold by computing minimal subsets of the knowledge base that have the consequence in question. Until now, the pinpointing approach has only been applied to the DL and some of its extensions. This paper considers axiom pinpointing in the less expressive DL , for which subsumption can be decided in polynomial time. More precisely, we consider an extension of the pinpointing problem where the knowledge base is divided into a part, which is always present, and a part, of which subsets are taken. We describe an extension of the subsumption algorithm for that can be used to compute all minimal subsets of (the refutable part of) a given TBox that imply a certain subsumption relationship. The worst-case complexity of this algorithm turns out to be exponential. This is not surprising since we can show that a given TBox may have exponentially many such minimal subsets. However, we can also show that the problem is not even output polynomial, i.e., unless P=NP, there cannot be an algorithm computing all such minimal sets that is polynomial in the size of its input . In addition, we show that finding out whether there is such a minimal subset within a given cardinality bound is an NP-complete problem. In contrast to these negative results, we also show that one such minimal subset can be computed in polynomial time. Finally, we provide some encouraging experimental results regarding the performance of a practical algorithm that computes one (small, but not necessarily minimal) subset that has a given subsumption relation as consequence.

General action languages, like e.g. the Situation Calculus, use full classical logic to represent knowledge of actions and their effects in dynamic domains. Description Logics, on the other hand, have been developed to represent static knowledge with the help of decidable subsets of first order logic. In this paper, we show how to use Description Logic as the basis for a decidable yet still expressive action formalism. To this end, we use ABoxes as decidable state descriptions in the basic Fluent Calculus. As a second contribution, we thus obtain an independent semantics – based on a general action formalism – for a recent method for ABox-Update.

The W3C standard OWL provides a decidable language for representing ontologies. While its use is rapidly spreading, efforts are being made by researchers worldwide to augment OWL with additional expressive features or by interlacing it with other forms of knowledge representation, in order to make it applicable for even further purposes. In this paper, we integrate OWL with one of the most successful and most widely used forms of knowledge representation, namely Prolog, and present a hybrid approach which layers Prolog on top of OWL in such a way that the open-world semantics of OWL becomes directly accessible within the Prolog system.

Automatically discovering semantic relations between ontologies is an important task with respect to overcoming semantic heterogeneity on the semantic web. Ontology matching systems, however, often produce erroneous mappings. In this paper we propose a method for optimizing precision and recall of existing matching systems. The principle of this method is based on the idea that it is possible to infer logical constraints by comparing subsumption relations between concepts of the ontologies to be matched. In order to verify this principle we implemented a system that uses our method as basis for optimizing mappings. We generated a set of synthetic ontologies and corresponding defective mappings and studied the behavior of our method with respect to the properties of the matching problem. The results show that our strategy actually improves the quality of the generated mappings.