Abstract& Highlighting relevant information in a discourse context is a major aim of spoken language communication. Prosodic cues such as focal prominences are used to fulfill this aim through the pragmatic function of prosody. To determine whether listeners make on-line use of focal prominences to build coherent representations of the informational structure of the utterances, we used the brain event-related potential (ERP) method. Short dialogues composed of a question and an answer were presented auditorily. The design of the experiment allowed us to examine precisely the time course of the processing of prosodic patterns of sentence-medial or -final words in the answer. These patterns were either congruous or incongruous with regard to the pragmatic context introduced by the question. Furthermore, the ERP effects were compared for words with or without focal prominences. Results showed that pragmatically congruous and incongruous prosodic patterns elicit clear differences in the ERPs, which were largely modulated in latency and polarity by their position within the answer. By showing that prosodic patterns are processed on-line by listeners in order to understand the informational structure of the message, the present results demonstrate the psychobiological validity of the pragmatic concept of focus, expressed via prosodic cues.

No abstract

The aim of this paper is to present a tool developed in order to generate French rhythmical structure semi-automatically, without taking grammatical cues into account. On the basis of a phonemic alignment, the software first locates prominent syllables by considering basic acoustic features such as F0, duration and silent pause. It then assigns a degree of prominence to each syllable identified. The estimation of this degree results from a computation of the values of silent pause, relative duration and height averages used for prominence detection in the first step. The second part of the article presents an experiment conducted in order to validate the algorithm's performances, by comparing the predictions of the software with a continuous manual coding carried out by four annotators on a 4-minute stretch of corpus (788 syllables) involving read aloud speech, map task and spontaneous dialogue. The performance of the algorithm is encouraging: a Fleiss' kappa calculation estimates the rate at 0.8, and a correlation agreement calculation at 91%, in the best cases.

The notion of sentence-as it is defined in syntactic, semantic, graphic and prosodic terms-is not a suitable maximal unit for the prosodic and syntactic annotation of spoken corpora. Still, this notion is taken as a reference in many syntactic and prosodic annotation systems. We present here the modular approach we adopted for the annotation of the Rhapsodie corpus of spoken French, which led us to distinguish three types of elementary units operating in discourse (government units, illocutionary units, and intonational periods) and to annotate them separately. We describe the types of interactions identified among these various levels of cohesion. On this basis we propose a reappraisal of the traditional notion of sentence and we define two additional types of discourse units that we consider as the minimal and the maximal span for the notion of sentence.