Jorge Baptista

Array hybrid multiplier versus modified booth multiplier: comparing area and power consumption of layout implementations of signed radix-4 architectures

In this paper, we describe the fully automated custom layout implementations of two architectures for signed multiplication. Performance comparisons between the two, namely in terms of their power consumption and area estimation are provided for 8 and 16-bit operands. The first architecture consists of a signed array multiplier that uses a radix-4 hybrid encoding to reduce the partial product lines and switching activity in the data buses. This new arithmetic operand encoding was recently proposed in (Costa et al., 2004), however only results at the logic level were presented. The second architecture implemented was the widely used modified Booth multiplier (Khater et al., 1996). The layout of both multipliers was generated by an automatic layout synthesis tool called TROPIC (MOraes, 1999). We compare the layout implementations in terms of area and power, as well as provide comparisons to first-order area estimates done in the logic design phase. The results show that the new hybrid array multiplier can be significantly more efficient, with close to 30% power savings.

Portuguese Temporal Expressions Recognition: From TE Characterization to an Effective TER Module Implementation

Taking into account the temporal dimension conveyed in texts is a challenge to natural language processing. At the same time this task is of great importance for a wide range of natural language processing applications. The goal of this paper is twofold. First a characterization of Portuguese temporal expressions as they appear in texts is presented. This classification is intended to meet the requirements of high inter-agreement between annotators of temporal expressions. Second, relying on this characterization, an effective temporal expression annotation tool is described. Results from its evaluation are reported.

Building a dictionary of anthroponyms

This paper presents a methodology for building an electronic dictionary of anthroponyms of European Portuguese (DicPRO), which constitutes a useful resource for computational processing, due to the importance of names in the structuring of information in texts. The dictionary has been enriched with morphosyntactic and semantic information. It was then used in the specific task of capitalizing anthroponyms and other proper names on a corpus automatically produced by a broadcast news speech recognition system and manually corrected. The output of this system does not offer clues, such as capitalized words or punctuation. This task expects to contribute in rendering more readable the output of such system. The paper shows that, by combining lexical, contextual (positional) and statistical information, instead of only one of these strategies, better results can be achieved in this task.