In this paper we describe a parallel implementation of a preconditioned conjugate gradient algorithm on the Meiko CS The parallel implemen tation is based on the use of a global reduction operator this allows to obtain a simple and e cient code whose performance is easily and pre cisly predictable We describe and validate a performance model and we show and comment experimental results

The paper reports on the lexico semantic annotation level of the Italian Treebank the rst Italian corpus with a multi level anno tation morpho syntactic syntactic and lexico semantic The strategy of annotation and the reference lexical resource are described and the results achieved too

The typological literature generally assumes that semantic maps and the underlying conceptual spaces have mental reality, that is, they correspond to a universal arrangement of the relevant conceptual situations in a speaker's mind, based on perceived relations of similarity between these conceptual situations (Croft 2001 and 2003, Haspelmath 2003, among others). Croft (2003) goes far as arguin...

Abstracts are listed in alphabetical order by presenting author and include information as provided.s are listed in alphabetical order by presenting author and include information as provided.

In this paper, we considered two representative speaker adaptation (SA) approaches – MAP and MLLR techni -ques – for our Korean isolated word recognition task. In addition, we proposed a new speaker adaptation alg orithm to improve the performance of MAP technique. It is based on least squares method between MAP ada pted mean vectors and the corresponding SI mean vect ors. The results of our ex...

ABSTRACT We present a method for lustering do uments and extra ting topi words of ea h luster using a probabilisti graphi al model. We maximize the likelihood of the model with the Expe tation Maximization algorithm. Our experiments demonstrate that the latent variables of the model an be seen as lusters of do uments and terms.

This paper addresses the problem of segmentation of image sequences into moving objects. The segmen-tation (label) eld is modeled by applying a Markov Random Field on regions provided by a watershed algorithm. The algorithm iterates between a motion estimation and a segmentation step. Temporal consistency is introduced by tracking the segmentation in time.

This paper deals with the use of some morphological tools for video coding at very low bit rates Rather than describing a complete coding algorithm the purpose of this paper is to focus on morphological connected operators and segmen tation tools that have recently proved to be attractive for compression