'Dare the boist'rous main': the role of the Belfast Newsletter in the process of emigration from Ulster to North America 1760-1800

Before the mass migrations from Ireland in the nineteenth century, earlier waves of migration in the eighteenth century saw significant numbers of people leave Ireland, predominantly from Ulster, to settle in North America. This article, using as its principal data source the Belfast News Letter ( BNL), its letters, advertisements and reports, focuses firstly on reconstructing the late eighteenth-century migration process and voyage, highlighting the barriers represented by the Atlantic Ocean. In addition to the challenges of the sea, there were problems with the ships, the ever-present danger of disease and also threats from other vessels, from privateers to press gangs. The voyage was recognized as a ‘universal dread’, and the risks taken to ‘dare the boist’rous main’ were perhaps not minimized in the pages of the BNL, whose editorial stance was antipathetic to the migration for the potential harm it caused to Ulster by removing so many of its industrious young. The second part of this article goes on to consider the newspaper’s and others’ vested interests in the emigration process, demonstrates how these were manifested in the press and sets the coverage of this very significant early emigration flow within the context of contemporary religious and colonial discourses at a period of very lively transatlantic interactions.

FPGA Implementation of a Pipelined Gaussian Calculation for HMM-Based Large Vocabulary Speech Recognition

A scalable large vocabulary, speaker independent speech recognition system is being developed using Hidden Markov Models (HMMs) for acoustic modeling and a Weighted Finite State Transducer (WFST) to compile sentence, word, and phoneme models. The system comprises a software backend search and an FPGA-based Gaussian calculation which are covered here. In this paper, we present an efficient pipelined design implemented both as an embedded peripheral and as a scalable, parallel hardware accelerator. Both architectures have been implemented on an Alpha Data XRC-5T1, reconfigurable computer housing a Virtex 5 SX95T FPGA. The core has been tested and is capable of calculating a full set of Gaussian results from 3825 acoustic models in 9.03 ms which coupled with a backend search of 5000 words has provided an accuracy of over 80%. Parallel implementations have been designed with up to 32 cores and have been successfully implemented with a clock frequency of 133?MHz.

Optimization of Weighted Finite State Transducer for Speech Recognition

There is considerable interest in creating embedded, speech recognition hardware using the weighted finite state transducer (WFST) technique but there are performance and memory usage challenges. Two system optimization techniques are presented to address this; one approach improves token propagation by removing the WFST epsilon input arcs; another one-pass, adaptive pruning algorithm gives a dramatic reduction in active nodes to be computed. Results for memory and bandwidth are given for a 5,000 word vocabulary giving a better practical performance than conventional WFST; this is then exploited in an adaptive pruning algorithm that reduces the active nodes from 30,000 down to 4,000 with only a 2 percent sacrifice in speech recognition accuracy; these optimizations lead to a more simplified design with deterministic performance.