Compilation of Corpora for the Study of the Information Structure–Prosody Interface

Compilation of Corpora for the Study of the Information Structure–Prosody Interface M´onica Dom´ınguez1, Alicia Burga1, Mireia Farr´us1, Leo Wanner2,1 University Pompeu Fabra1, Catalan Institute for Research and Advanced Studies (ICREA)2 C. Roc Boronat, 138 Barcelona, Spain {monica.dominguez, alicia.burga, mireia.farrus, leo.wanner}@upf.edu Abstract Theoretical studies on the Information Structure–prosody interface argue that the content packaged in terms of theme and rheme correlates with the intonation of the corresponding sentence. However, there are few empirical studies that support this argument and even fewer resources that promote reproducibility and scalability of experiments. In this paper, we introduce a methodology for the compilation of annotated corpora to study the correspondence between Information Structure and prosody. The application of this methodology is exemplified on a corpus of read speech in English annotated with hierarchical thematicity and automatically extracted prosodic parameters. Keywords: prosody, information structure, theme, rheme, thematicity, language resource, annotated corpora 1. Introduction The interest in the Information Structure–prosody corre- spondence applied to natural language speech generation lies in the derivation of prosody that is communicatively oriented and, therefore, more expressive. Knowing the linguistic mechanisms involved in human communication is pertinent to the achievement of multifaceted speech technologies that can carry out more complex tasks linked to conversational settings. The so-called Information Structure–prosody interface stands out as a solid ground for starting to build up such a communicative model in the computational field. However, empirical approaches to the Information Structure–prosody interface are scarce, studies on a corpus of more than two speakers are uncommon and the availability of corpora is, so to say, exceptional. It is usually one aspect of Information Structure that is studied: thematicity. Thematicity defines how content is packaged in terms of “what is being talked about”, i.e., the ‘theme’ and “what is being said”, i.e., the ‘rheme’. Most of the approaches draw upon a binary flat thematic division and established a one-to-one correspondence between theme–rheme and rising–falling intonation patterns respec- tively; see, e.g., (Steedman, 2000; Haji-Abdolhosseini and M¨uller, 2003; B¨uring, 2003). A different view on thematicity is that advocated by I. Mel’ˇcuk in the context of the MTT (Mel’ˇcuk, 2001). Compared to the traditional theme–rheme dichotomy, thematicity in the MTT introduces two key features that enhance the scope of the theme–rheme span division, namely: (i) the notion of specifier, which sets up the context of the sentence, and (ii) the fact that thematicity is defined over propositions, rather than over sentences. This second feature implies that thematicity is per se hierarchical: if a proposition is embedded, its thematicity will be embedded as well. Previous studies proved that hierarchical thematicity corresponds to a wider range of intonation patterns, and is, therefore, a more adequate representation than binary approaches, especially for long syntactically complex sentences; see, e.g., (Dom´ınguez et al., 2016a). In this paper we present a methodology for the com- pilation of a corpus for research on the Information Structure–prosody interface from an empirical perspective. This methodology is based on the formal description of information (or communicative) structure by Mel’ˇcuk (2001), which has …

Content truncated to 500 words