Catálogo de publicaciones - libros

There is a lot of evidence indicating that the semantic web is growing very rapidly. For example, an IBM report published last year indicated a 300% increase between 2003 and 2004 and pointed out that the rate of growth of the semantic web is mirroring that of the web in its early days. Indeed, repositories such as Swoogle already contain thousands of ontologies and hundreds of millions of RDF triples. Thus, a large scale semantic web is rapidly becoming a reality and therefore we are quickly reaching the point where we can start thinking about a new generation of intelligent applications, capable of exploiting such large scale semantic markup. Of course, while the semantic web provides an exciting opportunity, it also introduces very complex challenges. For example, the available semantic markup is extremely heterogeneous both with respect to its ontological profile and also to its degree of quality and to the level of trust that can be assigned to it. These features of the semantic web introduce an element of complexity, which was absent from traditional knowledge-based systems, where data quality was under the control of the developers, and provenance and heterogeneity did not apply. In my talk I will discuss these issues in some detail, and in particular I will describe the emerging semantic landscape and highlight some of the distinctive features characterizing the new generation of applications, which will be enabled by a large scale semantic web. In my presentation I will also present some concrete initial prototypes of this new generation of semantic applications, which exploit available large scale web semantics, to provide new ways to support question answering, information extraction and web browsing.

Efficient and effective Knowledge Acquisition and Sharing are of vital importance for large organizations. Complex human and technical aspects make them a complex issue. In this talk I will describe and discuss requirements, tools and methodologies for Knowledge Acquisition and Sharing in large organizations. I will use examples from the aerospace and automotive domain to describe and discuss real world requirements, including those related to human factors and technical issues. Then I will describe techniques from the field of Human Language Technologies and the Semantic Web that can be used for addressing those requirements. I will then present examples of real world applications developed in the aerospace domain. Finally, I will discuss some future trends.

For a multiagent system (MAS), coordination is the assumption that agents are able to adapt their behavior according to those of the other agents. The principle of Property Based Coordination (PBC) is to represent each entity composing the MAS by its observable properties, and to organize their perception by the agents. The main result is to enable the agents to have contextual behaviors. In this paper, we instantiate the PBC principle by a model, called EASI -Environment as Active Support of Interaction-, which is inspired from the Symbolic Data Analysis theory. It enables to build up an interaction as a connection point between the needs of the initiator, those of the receptor(s) and a given context. We demonstrate that thanks to PBC, EASI is expressive enough to instantiate other solutions to the connection problem. Our proposition has been used in the traveler information domain to develop an Agent Information Server dynamically parameterized by its users.

In this paper, we propose a general and abstract formal setting for argumentative dialogue protocols. We identify a minimal set of basic parameters that characterize dialogue protocols. By combining three parameters, namely the possibility or not of backtracking, the number of moves per turn and the turn-taking of the agents, we identify eight classes of protocols. We show that those classes can be reduced to three main classes: a ‘rigid’ class, an ‘intermediary’ one and a ‘flexible’ one. Although different proposals have been made for characterizing dialogue protocols, they usually take place in particular settings, where the locutions uttered and the commitments taken by the agents during dialogues and even the argumentation system that is involved are fixed. The present approach only assumes a minimal specification of the notion of dialogue essentially based on its external structure. This allows for protocol comparison and ensures the generality of the results.

Nowadays, more and more artificial agents integrate emotional abilities, for different purposes: expressivity, adaptability, believability... Designers mainly use Ortony et al.’s typology of emotions, that provides a formalization of twenty-two emotions based on psychological theories. But most of them restrain their agents to a few emotions among these twenty-two ones, and are more or less faithful to their definition. In this paper we propose to extend standard BDI (belief, desire, intention) logics to account for more emotions while trying to respect their definitions as exactly as possible.

We investigate in this work a generalization of the known representation which allows an efficient Boolean encoding for n-ary CSPs. We show that the space complexity of the Boolean encoding is identical to the one of the classical CSP representation and introduce a new inference rule whose application until saturation achieves arc-consistency in a linear time complexity for n-ary CSPs expressed in the Boolean encoding. Two enumerative methods for the Boolean encoding are studied: the first one (equivalent to MAC in CSPs) maintains full arc-consistency on each node of the search tree while the second (equivalent to FC in CSPs) performs partial arc-consistency on each node. Both methods are experimented and compared on some instances of the Ramsey problem and randomly generated 3-ary CSPs and promising results are obtained.

In this paper, we focus on the resolution of Crew Pairing Optimization problem that is very visible and economically significant. Its objective is to find the best schedule, i.e., a collection of crew rotations such that each airline flight is covered by exactly one rotation and the costs are reduced to the minimum. We try to solve it with Ant Colony Optimization algorithms and Hybridizations of Ant Colony Optimization with Constraint Programming techniques. We give an illustrative example about the difficulty of pure Ant Algorithms solving strongly constrained problems. Therefore, we explore the addition of Constraint Programming mechanisms in the construction phase of the ants, so they can complete their solutions. Computational results solving some test instances of Airline Flight Crew Scheduling taken from NorthWest Airlines database are presented showing the advantages of using this kind of hybridization.

In Constraint Programming, enumeration strategies (selection of a variable and a value of its domain) are crucial for resolution performances. We propose to use Local Search for guiding enumeration: we extend the common variable selection strategies of constraint programming and we achieve the value selection based on a Local Search. The experimental results are rather promising.

This paper describes two approaches to integrating standalone information processing techniques into a semantic application capable of acquiring and maintaining knowledge, which we conducted using our open Semantic Web framework of Magpie. We distinguish between integration through aggregation and through choreographing, and argue that the latter is not only simpler to realize but also provides greater benefits. The benefits were, in our experiment, related to developing a capability of maintaining and validating knowledge through an integration of down- and upstream knowledge processing tools. We describe the principles of integration and relate them to pragmatic challenges for the semantic web and to strategic directions of its evolution.

Automatic authorship identification offers a valuable tool for supporting crime investigation and security. It can be seen as a multi-class, single-label text categorization task. Character n-grams are a very successful approach to represent text for stylistic purposes since they are able to capture nuances in lexical, syntactical, and structural level. So far, character n-grams of fixed length have been used for authorship identification. In this paper, we propose a variable-length n-gram approach inspired by previous work for selecting variable-length word sequences. Using a subset of the new Reuters corpus, consisting of texts on the same topic by 50 different authors, we show that the proposed approach is at least as effective as information gain for selecting the most significant n-grams although the feature sets produced by the two methods have few common members. Moreover, we explore the significance of digits for distinguishing between authors showing that an increase in performance can be achieved using simple text pre-processing.