Catálogo de publicaciones - libros

We propose a novel, high-level model of human learning and cognition, based on association forming. The model configures any input data stream featuring a high incidence of repetition into an association network whose node clusters represent data ‘concepts’. It relies on the hypothesis that, irrespective of the high parallelism of the neural structures involved in cognitive processes taking place in the brain cortex, the channel through which the information is conveyed from the real world environment to its final location (in whatever form of neural structure) can transmit only one data item per time unit. Several experiments are performed on the ability of the resulting system to reconstruct a given underlying ‘world graph’ of concepts and to form and eventually maintain a stable, long term core of memory that we call ‘semantic’ memory. The existence of discontinuous, first order phase transitions in the dynamics of the system is supported with experiments. Results on clustering and association are shown as well.

The development of Artificial Intelligence (AI) research has followed mainly two directions: the use of symbolic and connectionist (artificial neural networks) methods. These two approaches have been applied separately in the solution of problems that require tasks of knowledge acquisition and learning. We present the results of implementing a Neuro-Symbolic Hybrid System (NSHS) that allows unifying these two types of knowledge representation. For this, we have developed a compiler or translator of symbolic rules which takes as an input a group of rules of the type IF ... THEN..., converting them into a connectionist representation. Obtained the compiled artificial neural network this is used as an initial neural network in a learning process that will allow the “refinement” of the knowledge. To prove the refinement of the hybrid approach, we carried out a group of tests that show that it is possible to improve in a connectionist way the symbolic knowledge.

In this paper, we extend Moss and Parikh’s bi-modal language for knowledge and effort by an additional modality describing improvement. Like the source language, the new one too has a natural interpretaion in spatial contexts. The result of this is that concepts like the comparison of topologies can be captured within the framework of modal logic now. The main technical issue of the paper is a completeness theorem for the tri-modal system arising from the new language.

In this paper, we propose a new belief revision operator, together with a method of its calculation. Our formalization differs from most of the traditional approaches in two respects. Firstly, we formally distinguish between defeasible observations and indefeasible knowledge about the considered world. In particular, our operator is differently specified depending on whether an input formula is an observation or a piece of knowledge. Secondly, we assume that a new observation, but not a new piece of knowledge, describes exactly what a reasoning agent knows at the moment about the aspect of the world the observation concerns.

A powerful and useful approach for modeling knowledge and qualitative reasoning is the . The background of Cognitive Maps is the research about learning environments carried out by Cognitive Psychology since the nineteenth century. Along the last thirty years, these underlying findings inspired the development of computational models to deal with causal phenomena. So, a Cognitive Map is a structure of concepts of a specific domain that are related through cause-effect relations with the aim to simulate behavior of dynamic systems. In spite of the short life of the causal Cognitive Maps, nowadays there are several branches of development that focus on qualitative, fuzzy and uncertain issues. With this platform wide spectra of applications have been developing in fields like game theory, information analysis and management sciences. Wherefore, with the purpose to promote the use of this kind of tool, in this work is surveyed three branches of Cognitive Maps; and it is outlined one application of the Cognitive Maps for the student modeling that shows a conceptual design of a project in progress.

Interval algebra of Allen [4] propose a set of relations which is particularly interesting on historical annotating tasks [1]. However, finding the feasible relations and consistent scenario has been shown to be NP-complete tasks for interval algebra networks [11,10]. For point algebra networks and a restricted class of interval algebra networks, some works propose efficient algorithms to resolve it. Nevertheless, these sets of relations (made of basic relation disjunctions) are not intuitive for describing historical scenarios. In this paper we propose a set of concrete relations for the annotator, and we formalize it in terms of temporal algebras. We then describe how our model can be matched with other ones to merge calculation efficiency and information suitability.

In this paper, we describe EventNet, a toolkit for inferring temporal relations between Commonsense events. It comprises 10,000 nodes and 30,000 temporal links mined from the Openmind Commonsense Knowledge Base. It enables applications to deduce "obvious" (to people) temporal relations between commonly occurring events, for example: First, you wake up, then you can leave the house in the morning. The temporal relation might be one of cause and effect, of action/goal or prerequisite relations, or simply that they tend to follow each other in a commonly occurring “script”. In addition, the algorithm has some built-in heuristics to infer when its information is not enough. It then finds semantically similar nodes to dynamically search the knowledge base. EventNet has been used in projects such as an intelligent kitchen, and in intelligent interfaces for consumer electronics devices.

This paper describes an ontology-mapping framework in the context of query answering (QA). In order to incorporate uncertainty inherent to the mapping process, the system uses the Dempster-Shafer model for dealing with incomplete and uncertain information produced during the mapping. A novel approach is presented how specialized agents with partial local knowledge of the particular domain achieve ontology mapping without creating global or reference ontology. Our approach is particularly fit for a query-answering scenario, where an answer needs to be created in real time that satisfies the query posed by the user.

Ontology merging is the process of creating a new ontology from two or more existing ontologies with overlapping parts. Currently, there are many domain areas in Computer Science interested in this topic. Federated Databases and Semantic Web are some of them. In this paper we introduce a three level approach that provides a semi-automatic method to ontology merging. It performs some tasks automatically and guides the user in performing other tasks for which his intervention is required.

One of the big challenges of the knowledge management is the active and intelligent dissemination of the know-how to users while executing their tasks, without bothering them with information that is too far from their competences or out of their interest fields. Delivering a new document to the concerned users consists basically on appreciating the semantic relevance of the content of this document (domain ontology) in relation with users’ competences (competences ontology). In this paper, we illustrate the importance, within a documentary dissemination service, of the integration of an ontologies system, essentially based on the domain and the competences, but also on the users, the documents, the processes and the enterprise. The maintenance of these ontologies and basically those related to documents, to domain and to competences is a crucial aspect for the system survival. So, we describe what role the texts analysis can play for the maintenance of these ontologies.