COPAS - Contrast and Parallelism in Speech

Project PI
Ana Isabel Mata


The main purpose of this project is to determine how intonational and syntactic resources are deployed by speakers (i) to convey contrast associated with topic, common ground and new information at different hierarchical levels (constituents, sentences, question/answer pairs, adjacent turns) and (ii) to construct parallelism in expositive-argumentative contexts. Thus, COPAS addresses a fundamental problem of linguistic inquiry, the architecture of grammar, insofar as it will work on interfaces among different grammar modules (syntax, prosody, semantics) and between grammar and other modules involved in language use, such as information structure.

Hence, COPAS will focus on a set of concepts from information structure and semantics relevant for the study of the above mentioned interfaces: Contrast, Givenness, Aboutness, Newness, core concepts in the definition of both marked and unmarked Topic and Focus. In order to assess interface relations among information structure, syntax and intonation, COPAS will concentrate on syntactic structures involving “discourse-driven” activation of the peripheries (through Move and Merge) and related intonational structures (pitch accents types/tunes). We will consider different types of (i) left and right dislocations (including topicalizations and antitopics); (ii) clefts (including reduced clefts in fragment answers); (iii) syntactic and prosodic parallel structures.

Although the interaction between intonational and syntactic patterns and the role intonational variation plays in information packaging have received an increasing interest in recent years, and systematic correlations between information structure and intonation have been proposed [Pierrehumbert & Hirschberg, 1990] – in particular, an association between certain pitch accent/contour types and topics/foci has been claimed for different languages [Pierrehumbert, 1980; Steedman, 2000; Büring, 2003] - few empirical studies have explored these links using naturally occurring speech. As for Portuguese, these links remain insufficiently understood as prosodic research based on speech outside a laboratory setting is fairly recent [Mata, 1999] and most of the published work concerns phrasing. Moreover, experiments of ToBI-like prosodic annotation of spontaneous speech corpora are also fairly recent and, although the first proposal Towards a P_ToBI [Viana et al., 2007] is useful for manual annotation, a tool for automatic annotation as AuToBI [Rosenberg, 2010] is not yet available for Portuguese. A major goal of this project is to contribute to fill these gaps.

The study of this type of interactions is theoretically important, as it increases our knowledge of the nature of the language module and it may contribute to answer open questions regarding strategies to convey Contrast and to construct Parallelism in human speech communication. The results of COPAS are expected to contribute to build models integrating intonation patterns in spoken language processing in a more natural way. They are also expected to contribute to the field of Educational Linguistics, since awareness of the interplay of multilevel linguistic resources in spoken discourse is a basis on which to develop sophisticated skills and strategies of language use.

In what concerns adult grammar, the main goals of COPAS are (i) to determine whether Contrast is intonationally encoded in a similar way across different syntactic structures; (ii) to determine whether different syntactic processes, which may generate left and right dislocations and clefts, have distinct intonational correlates; (iii) to determine which syntactic and intonation parallelism strategies speakers use to build utterance sequences with the property of textuality in monologues and dialogues.

COPAS has two major milestones – the first account of intonation-syntax-discourse interactions resulting from naturally occurring speech analyses (M3) and the integrated final account, capitalizing on results from perception and comprehension experiments (M5) – one initial milestone and two intermediate milestones. Research strategies comprise (i) the annotation of a selected working corpus representative of different communication contexts, including dialogues and monologues by adults, teenagers and children, (ii) automatic ToBI prosodic annotation experiments using subsets of the annotated materials, (iii) multilevel (qualitative and quantitative) analyses of intonation-syntax-discourse interactions, (iv) perceptual and comprehension experiments designed to assess the effects of the correlations found.

COPAS is an innovative project on Portuguese and in Portugal, insofar as it brings together a multidisciplinary research team with the resources, knowledge and expertise that allows an integrated approach to intonation-syntax-discourse interactions.

Final workshop

Prosody-syntax-semantics interfaces in Portuguese: exploring spontaneous speech corpora. 2-3 July 2015, Anfiteatro II, Faculdade de Letras da Universidade de Lisboa.