You are here

Publications

Primary tabs

  1. C. Brunelli,
    Progetto e implementazione di un sistema di ragionamento basato sull`utilizzo di teorie multiple del primo ordine,
  2. Gianpaolo Avancini; Roberto Flor,
    Sicurezza dei sistemi di calcolo ITC,
  3. P. Giongo,
    Un approccio basato su Alberi di Classificazione Binari alla Comprensione del Parlato,
  4. R. Palmarin,
    Utilizzo di nozioni morfo-sintattiche in un sistema di riconoscimento del parlato,
  5. Gianpaolo Avancini; H. de Rosa; Roberto Flor,
    L`Automazione d`ufficio nell`Istituto Trentino di Cultura. La situazione attuale e le prospettive per il futuro.,
    La presente relazione ha lo scopo di illustrare la situazione attuale dell`Istituto Trentino di Cultura per quanto riguarda il software e l`hardware utilizzato per le procedure di automazione d`ufficio. Di seguito verrà descritto un progetto che, modificando l`orientamento attuale dell`Istituto in questo settore, ha l`obiettivo di portare dei benefici in termini di produttività, sicurezza e integrità delle informazioni,
  6. F. Ciravegna; Alberto Lavelli; D. Petrelli; Fabio Pianesi,
    The GePpeTto Development Environment,
  7. F. Fignoni; Stefano Messelodi; Carla Maria Modena,
    Review of the State of the Art in Optical Character Recognition. Part 2: Hand Printed Documents,
    The automatic recognition of hand written text is an important practical problem with a great variety of potential applications: fax reading, form reading, help to blind people, signature verification, check reading, code recognition, address reading, document database population, and many more. Therefore, the interest of academic groups and commercial associations in hand written recognition is high. In this state of the art review, we address the field of the off-line recognition of unconstrained hand printed characters. We illustrate the results of two comparative tests of ICR systems conducted by NIST: the first deals with the recognition of isolated hand printed characters and the second with the reading of hand printed words from U.S. census forms. Starting from these results and from our experience, we focus our attention on the available commercial products we consider the most reliable. Finally, we present recent trends of the academic research in this area of machine vision, aiming at identifying the most challenging fields for future research,
  8. Emanuele Pianta; Elena Not,
    Designing a Text Planning Architecture for a Multilingual Generation System,
  9. Edmondo Trentin; Diego Giuliani,
    Mixtures of Recurrent Neural Networks for Speaker Adaptation,
    This work introduces a multiple connectionist architecture based on a mixture of Recurrent Neural Networks to approach the problem of speaker adaptation in the acoustic feature domain. Adaptation in the feature space is accomplished by means of a suitable acoustic feature transformation. The aim is the reduction of differences between the acoustic space of a new speaker and the training acoustic space of a given recognizer, in order to increase recognition performance. The transformation has to be estimated from a small amount of speech signal, sampled from the new speaker, and is applied at recognition state to preprocess input speech data, before feeding them into the recognizer. In this work, recognition experiments with continuous speech and a large vocabulary have been carried out using speaker-dependent (SD) and speaker-independent (SI) recognition systems, based on continuous density hidden Markov models. Different connectionist approaches to speaker adaptation are discussed. At first, an extended Multi-Layer Perceptron (MLP) is trained to realize the required multivariate non-linear regression. Various directions along which to extend the feed-forward model with the introduction of recurrent connections are discussed. Experiments with the SD recognizers show a remarkable 56% reduction of the word error rate with respect to the baseline (speech recognition without adaptation) when the recurrent network is used as an acoustic front-end to the recognizer, outperforming the standard linear regression approach. In the SI case, on the other side, it is more difficult to reach a significant recognition improvement with respect to the hidden Markov model alone. This leads to the development of a more effective regression technique based on combined neural networks. A mixture of MLPs, as well as a technique for combining recurrent nets, are used. Experimental results show that the proposed architecture consistently improves recognition performance yielding a 21% reduction of the word error rate,
  10. Luciano Serafini; Chiara Ghidini,
    Local Semantics for Federated Databases,
    In many applications which need for a large amount of information, knowledge is partitioned and represented in a set of databases (DB) integrated in a federated database (FDB). A FDB is a collection of distributed, partial, redundant, and partially autonomous DBs. Distribution, redundancy, partiality and autonomy generate many problems in the management of a FDB. The most important are semantic etherogenity, update propagation, inter-schema dependencies, and query distribution which need for a formal treatment. Several approaches have been proposed in the past. However these formalisms are inadequate as they partially represent some of the aspects, but they fail to represent all of them. The goal of this paper is to develop a formal semantics called local semantics, for FDB, which explicitly represents distributed, redundant, partial, autonomous DBs. We substantiate the above adequancy claim by formalizing three motivating examples which involves all these aspects,