Tommaso Raso

Validação estatística dos critérios de segmentação da fala espontânea no corpus C-ORAL-BRASIL

Este artigo apresenta o processo empregado no corpus C-ORALBRASIL,bem como os resultados estatisticos da segmentacao dafala espontânea encontrados nesse corpus. Exploramos especialmentea validacao estatistica dos criterios para a segmentacao da fala emenunciados e unidades tonais com base na Teoria da Lingua em Ato.A validacao teve por objetivo assegurar que o produto final tivesse amaior uniformidade possivel quanto a segmentacao da fala. Aposum ano, realizamos novos testes de validacao, para reavaliacao dogrupo principal de transcritores no momento da revisao do corpus.Os resultados da validacao estatistica ao final da fase de treinamentoindicam alta confiabilidade quanto a segmentacao do corpus, e areavaliacao indicou um grau ainda maior de acordo entre ostranscritores. O principal objetivo deste processo, a confiabilidade euniformidade das transcricoes, foi obtido.

Tópico e apêndice no português do Brasil: algumas considerações

O artigo, com base na Teoria da Lingua em Ato (CRESTI, 2000), analisa as unidades informacionais de topico e de apendice de comentario em uma amostra de tres textos do portugues do Brasil: um texto dialogico e dois textos monologicos. O objetivo e mostrarcomo as duas unidades tem comportamentos diferentes tanto no nivel prosodico quanto nos niveis distribucional, morfossintatico e cognitivo. A frequencia delas nos diferentes tipos de textos mostra tambem diferentes papeis na complexizacao dos enunciados, com uma diferente sensibilidade ao tipo de interacao. A tese e que se trata de unidades informacionais diferentes e nao da mesma unidade em posicao especular.

Estudo contrastivo do uso de alocutivos em português brasileiro e italiano

O artigo estuda a unidade informacional de alocutivo em chave contrastivaitaliano-portugues do Brasil, com base na Teoria da Lingua em Ato. Oestudo se baseia na analise de 15 textos para cada lingua (divididos paritariamenteem dialogos, monologos e conversacoes) de, em media, 1500 palavras,retirados de dois corpora de fala espontânea comparaveis: o C-ORAL-ROMitaliano e o C-ORAL-BRASIL. Foram estudados: a frequencia da unidade;suas caracteristicas acusticas; sua distribuicao; sua variacao lexical. Os resultadosapontam para uma maior frequencia de alocutivos no portugues doBrasil (3,32 vezes o numero das unidades em italiano) e tambem para a suamaior variacao lexical (2,71 vezes o numero de types do italiano). A conclusaosugere uma explicacao de carater socio-linguistico para esses resultados.

Uma aplicação da teoria da língua em ato ao português do Brasil

In this article we present the first application of the theory of Languagein Act (CRESTI, 2000b) to Brazilian Portuguese. After a summaryof the theory is presented, we show how a text can be divided inutterances and how the utterance can be divided in tone units, throughthe perception of terminal and non terminal prosodic breaks. Basedon dedicated entonational profiles, we show the informational taggingof the tone units, which is motivated by prosodic, functional anddistributional criteria. We conclude by showing how the prosodicand informational frame clearly change, depending on turn size andevidence of the illocutionary principle.

A segmentação da fala espontânea: aspectos prosódicos, funcionais e aplicações para a tecnologia / Spontaneous Speech Segmentation: Functional and Prosodic Aspects with Applications for Automatic Segmentation

This issue of Revista de Estudos da Linguagem is dedicated to a theme addressed in several other initiatives promoted by its guest editors, along with colleagues from the international community. The theme, which in recent years has played an increasingly important role in the disciplines that study speech production and perception, is the segmentation of speech into smaller units addressed from both formal and functional perspectives, fundamentally under a theoretical approach coupled with an empirical focus. Among the main initiatives, we mention:

Automatic Detection of Prosodic Boundaries in Brazilian Portuguese Spontaneous Speech.

This paper presents some models based on multiple phonetic-acoustic parameters for the automatic detection of prosodic boundaries in spontaneous speech. A sample with seven excerpts of monologic Brazilian Portuguese spontaneous speech was segmented into prosodic units by 14 trained annotators. The perceived prosodic boundaries were annotated as terminal or non-terminal prosodic boundaries. A Praat script was prepared in order to extract a set of acoustic parameters during the speech signal. Two statistical classifiers, namely Random Forest e Linear Discriminant Analysis, were used to generate models of subgroups of acoustic parameters that could work as predictors of prosodic boundaries in comparison with the human annotators. The initial evaluation of the classifiers showed that both present relative success in detecting boundaries. The LDA performed better in predicting boundaries and therefore its models were refined. The final model for terminal boundaries showed 80% of agreement with human annotators. As for non-terminal boundaries, three models were obtained. The sum of boundaries identified by the three models together corresponds to an agreement of 98% with the human annotators.

O papel da pausa na segmentação prosódica de corpora de fala

Resumo : O trabalho apresenta uma analise da pausa silenciosa enquanto criterio de segmentacao da fala em unidades comunicativamente autonomas. O objetivo deste trabalho e contextualizado com uma discussao sobre a unidade de referencia da fala e a nocao de pausa. Argumenta-se a favor de uma segmentacao da fala em unidades pragmaticas de natureza acional, delimitadas no fluxo da fala por meio de fronteiras prosodicas. Realizou-se analise estatistica em amostra aleatoria de fala espontânea extraida do corpus C-ORALBRASIL. Por intermedio de um modelo de regressao nao linear, procurou-se identificar duracoes de pausas consistentemente preditivas de fronteiras de unidades de referencia. Os resultados mostram que nao ha nenhum valor de duracao de pausa que possa ser utilizado como criterio confiavel para a segmentacao da fala em unidades comunicativamente autonomas. Palavras chave : fala; corpora; segmentacao; unidade prosodica; ilocucao. Abstract : This work presents an analysis of silent pauses when applied as criterion for the segmentation of speech into communicatively autonomous units. The aim of this work is contextualized in a discussion about the unit of reference for speech and the notion of pause. We argue that speech should be parsed into pragmatic units of actional nature, which are signaled within the speech flow by prosodic boundaries. We performed a statistic analysis on a random spontaneous speech sample extracted from C-ORAL-BRASIL corpus. Non-linear regression models were applied so as to identify pause durations that could consistently predict unit boundaries. Results show that there isn’t any pause duration value that could be used as a reliable criterion for the segmentation of speech into communicatively autonomous units. Keywords : Speech; Corpora; Segmentation; Prosodic Unit; Illocution.

The C-ORAL-BRASIL I: reference corpus for informal spoken brazilian portuguese

The C-ORAL-BRASIL is a Brazilian Portuguese spontaneous speech corpus, representative of the state of Minas Gerais diatopy (primarily from the capital city, Belo Horizonte,metropolitan area). The corpus was compiled following the same architecture and segmentation criteria adopted by the C-ORAL-ROM [1] as well as its alignment software, the WinPitch [2]. The corpus comprises 139 informal speech texts, 208,130 words, 21:08:52 hours of recording (6.1 GB wav files). The mean word number per text is 1,500. The recordings were carried out with high resolution, non-invasive wireless equipment, generally with clip-on, monodirectional microphones, and a mixer whenever there were more than two interactants, in a few occasions omnidirectional microphones were used. The texts are transcribed following the CHAT format [3], implemented for prosodic annotation [4]. The main goals for the corpus architecture are the documentation of the diaphasic and diastratic variations in Brazilian Portuguese speech.

The C-ORAL-BRASIL Corpus: Methodological Basis for the Treatment of Spontaneous Speech

This paper highlights the primary methods employed in the C-ORAL-BRASIL compiling process, i.e, recording, transcribing and segmenting oral texts. The C-ORAL-BRASIL is a Brazilian Portuguese corpus of spontaneous speech, designed for the study of informational structure. It is representative of the diaphasic variation, seeking to cover as many different comunicative situations as possible. This paper presents and exemplifies the processes of transcription and segmentation of speech into prosodic units as employed in our on-going research. It concludes with illustrations of some questions that the corpus will enable us to answer.

Formar formadores para a escrita profissional

This paper presents some results of a ten year research and teaching project for educating educators for professional writing, either in the educational system or in the working environment. Specifically, it presents a curriculum of basic competences that all educators should have, independently of the didactic destination. These competences can be summarized in five points: the conscience of the variation between speaking and writing textual strategies; the ability to identify information within a context, and in using it for textual purposes; the conscience of the parameters that characterize a textual genre, and the capacity of managing them reformulating a text for different purposes; the skills for managing the different aspects of a communicative situation; and, finally, argumentation, with and without documentation. The study privileges textual rather than surface linguistic competences; so, even if elaborated for an Italian speaking environment, it is applicable in any occidental language and culture.