Catálogo de publicaciones - libros

Large document collections, such as those delivered by Internet search engines, are difficult and time-consuming for users to read and analyse. The detection of common and distinctive topics within a document set, together with the generation of multi-document summaries, can greatly ease the burden of information management. We show how this can be achieved with a clustering algorithm based on fuzzy set theory, which (i) is easy to implement and integrate into a personal information system, (ii) generates a highly flexible data structure for topic analysis and summarization, and (iii) also delivers excellent performance.

We present a system capable of modeling human newspaper readers. It is based on the extraction of reported speech, which is subsequently converted into a fuzzy theory-based representation of single statements. A domain analysis then assigns statements to topics. A number of fuzzy set operators, including fuzzy belief revision, are applied to model different belief strategies. At the end, our system holds certain beliefs while rejecting others.

Complex Nominals (CNs) have simple syntactic structure that conceals non-trivial semantic characteristics. While speakers of natural languages combine noun(s)/adjective(s) with a head noun to indicate existing or novel concepts with ease, formalizing such a semantic process, however, has proven to be a daunting task. In this paper, we present a unified semantic approach for constructions involving a head-noun and modifier(s), i.e., adjective(s)/noun(s). Based on a rigorous typing system, this approach uses set intersection as the only underlying semantic rule. We argue that this novel approach is compositional and warrants consistent inferences.

In many domains, automated speech recognition (ASR) demands highly robust and accurate recognition software. Unfortunately, in such domains, even a 99% accurate recognizer is inadequate, and other methods for increasing the reliability and performance of ASR must be considered. As a possible solution to this problem, post-speech-recognition error detection can assist in proofreading more efficiently. To this end, we have developed a multi-heuristic algorithm using natural language processing to detect recognition errors. As a proof of concept, we have applied this algorithm to the radiology domain. The results are encouraging, showing a 22% increase in the recall performance, and a 6% increase in the precision performance, over the best individual technique.

DEC-POMDPs provide formal models of many cooperative multiagent problems, but their complexity is NEXP-complete in general. We investigate a sub-class of DEC-POMDPs termed . A typical instance consists of an area populated by mobile agents. Agents have no prior knowledge of the area, have limited sensing and communication, and effects of their actions are uncertain. Success relies on planing actions that result in high accumulated rewards. We solve an instance of multiagent expedition based on collaborative design network, a decision theoretic multiagent graphical model. We present a number of techniques employed in knowledge representation and demonstrate the superior performance of our system in comparison to greedy agents experimentally.

Pathfinding on large maps is time-consuming. Classical search algorithms such as Dijkstra’s and A* algorithms may solve difficult problems in polynomial time. However, in real-world pathfinding examples where the search space increases dramatically, these algorithms are not appropriate. Hierarchical pathfinding algorithms that provide abstract plans of future routing, such as HPA* and PRA*, have been explored by previous researchers based on classical ones. Although the two hierarchical algorithms show improvement in efficiency, they only obtain near optimal solutions. In this paper, we introduce the Hierarchical Shortest Path algorithm (HSP) and a hybrid of the HSP and A* (HSPA*) algorithms, which find optimal solutions in logarithmic time for numerous examples. Our empirical study shows that HSP and HSPA* are superior to the classical algorithms on realistic examples, and our experimental results illustrate the efficiency of the two algorithms. We also demonstrate their applicability by providing an overview of our Route Planner project that applies the two algorithms proposed in this paper.