Automatic transcription of conversational telephone speech

This paper discusses the Cambridge University HTK (CU-HTK) system for the automatic transcription of conversational telephone speech. A detailed discussion of the most important techniques in front-end processing, acoustic modeling and model training, language and pronunciation modeling are presented. These include the use of conversation side based cepstral normalization, vocal tract length normalization, heteroscedastic linear discriminant analysis for feature projection, minimum phone error training and speaker adaptive training, lattice-based model adaptation, confusion network based decoding and confidence score estimation, pronunciation selection, language model interpolation, and class based language models. The transcription system developed for participation in the 2002 NIST Rich Transcription evaluations of English conversational telephone speech data is presented in detail. In this evaluation the CU-HTK system gave an overall word error rate of 23.9%, which was the best performance by a statistically significant margin. Further details on the derivation of faster systems with moderate performance degradation are discussed in the context of the 2002 CU-HTK 10 × RT conversational speech transcription system. © 2005 IEEE.

On Mice and Men and De ratones y hombres: Cultural transfer of substandars characters from literary discourse to film discourse

Eguíluz, Federico; Merino, Raquel; Olsen, Vickie; Pajares, Eterio; Santamaría, José Miguel (eds.)

A study of populations of the anchoveta, Cetengraulis mysticetus, based on meristic characters

ENGLISH: This study was undertaken to determine whether meristic characters indicate that more than one major population of anchovetas occurs in the range of the species from Mexico to Peru. Interest in this species lies in the fact that it is the principal bait fish used to catch yellowfin and skipjack tunas in the Eastern Pacific. Specimens examined were from collections made by California tuna fishing vessels at six major baiting localities covering nearly the entire range of the species, namely, Almejas Bay on the outer coast of Baja California, Guaymas and Ahome Point in the Gulf of California, Gulf of Fonseca, Gulf of Panama, and Gulf of Guayaquil. Four meristic characters were selected for study: vertebrae, dorsal fin rays, anal fin rays, and gill rakers on the first gill arch. Vertebral counts, using X-ray film, were taken from a total of 1,500 fish, 250 each from each of the six localities. For the other characters, 125 anchovetas were examined from each locality for a total of 750, the counts being made with the aid of a binocular microscope. Specimens were between 80 and 165 mm. standard length. SPANISH: Este estudio ha sido hecho con el propósito de determinar si los caracteres numéricos de las anchovetas indican que existe más de una población de este pez en la zona en que se encuentra la especie, comprendida entre México y Perú. El interés en dicha especie radica en el hecho de que éste es el pez de carnada usado principalmente para la pesca de los atunes "aleta amarilla" y "barrilete" en el Pacífico Oriental. Los especímenes que han sido examinados, se tomaron de las muestras recogidas por los barcos atuneros de California en seis de las mejores localidades en que se pesca la anchoveta, las cuales comprenden casi toda la zona en donde se encuentra la especie, a saber, Bahía de Almejas en la costa exterior de Baja California, Guaymas y Punta Ahorne en el Golfo de California, el Golfo de Fonseca, el Golfo de Panamá y el Golfo de Guayaquil. Cuatro caracteres numéricos fueron escogidos para su estudio: los que presentan 1) las vértebras, 2) los radios de la aleta dorsal, 3) los radios de la aleta anal y 4) las branquispinas del primer arco branquial. Mediante el uso de películas con rayos X, se contaron las vértebras en un total de 1,500 peces, es decir, 250 de cada una de las seis mencionadas localidades. En relación con los otros caracteres, se examinaron 125 anchovetas de cada área, o sea, un total de 750 ejemplares, habiendo sido hecho el conteo por medio de un microscopio binocular. Los especímenes tenían un largo standard entre 80 y 165 milímetros. (PDF contains 24 pages.)

Studies on the biology of Gymnarchus niloticus in Lake Chad: Age determination and growth; meristic and morphometric characters

The meristic and morphometric characteristics of Gymnarchus niloticus are described and linear equations relating various parts of the body to the head length or total length are given. The age of G. niloticus in Lake Chad (Nigeria) was determined from growth marks on the opercular bones. The mean lengths for age, and mean weights for age obtained for the first five years of life are given. The assymptotic length and the von Betarlanffy growth parameters for the males and females combined are given

An evaluation of geographic and annual variation in morphometric characters and gill-raker counts of yellowfin tuna, Thunnus albacares, from the Pacific Ocean

ENGLISH: Samples of yellowfin tuna, Thunnus albacares, collected from five areas of the Pacific Ocean (Mexico, Ecuador, Australia, Japan, and Hawaii) between January and May of 1988 and 1990 were examined for spatiotemporal variation in morphometric characters and gill-raker counts. 'Iwo-factor analysis of variance, with area and year treated as grouping factors, indicated a significant difference in the means of the total gill-raker counts among fish from different areas, but no significant difference between fish caught in different years. The morphometric data were adjusted by allometric formulae to remove size effects. The correct classification rates for the five groups, using discriminant function analysis, based on adjusted morphometric characters, were 77.60/0 for the samples from 1988 and 74.40/0 for those from 1990. These are 72.00/0 and 68.00/0 (Cohen's kappa statistic) better than would have occurred chance. The pattern of geographic variability, however, is unstable for these two years, thus requiring separate discriminant functions for each year. Although there is annual variability in the morphometric characters, these results demonstrate that the stocks examined are morphometrically distinguishable and that their phenetic relationships reflect their geographic origin. SPANISH: Se examinaron muestras de atún aleta amarilla, Thunnus albacares, tomadas de cinco áreas del Océano Pacífico (México, Ecuador, Australia, Japón, y Hawaii) entre enero y mayo de 1988 y 1990, para descubrir variaciones espaciotemporales en las características morfométricas y los conteos de branquiespinas. Un análisis de varianza de dos factores, con área y año como factores de agrupación, indicó una diferencia significativa en los promedios de los conteos de branquiespinas totales entre peces de distintas áreas, pero ninguna entre peces capturados en distintos años. Se ajustaron los datos morfométricos con fórmulas alométricas para eliminar los efectos de la talla del pez. En un análisis de función discriminante, las tasas de clasificación correcta de los cinco grupos, basadas en características morfométricas ajustadas, fueron 77.60/0 para las muestras de 1988 y 74.40/0 para aquellas de 1990. Estas cifras son 72.00/0 y 68.00/0 (estadístico de kappa de Cohen) mejores de lo que se hubiera obtenido al azar. Sin embargo, la variabilidad geográfica es inestable en estos dos años, requiriendo por lo tanto funciones discriminantes separadas para cada año. Aunque existe variabilidad anual en las características morfométricas, estos resultados demuestran que los stocks examinados son morfamétricamente distinguibles, y que su relación fenética refleja su origen geográfico. (PDF contains 31 pages.)