Chapter 7
Syntactic segmentation of spoken corpus data
What prosody can contribute
Most corpus-based syntactic segmentation schemes rely on transcriptions alone, which can lead to
segmentation difficulties, especially when analyzing spontaneous conversations. We therefore suggest an approach to
segmentation that complements syntactic segmentation techniques with prosodic analyses and describe correspondences in
syntactic and prosodic segmentation as well as the exact syntactic contexts in which prosodic analyses are necessary
to avoid ambiguities and potential inaccuracies. Using 10 recordings from the Louvain Corpus of Native English
Conversation, utterances are independently and manually segmented and annotated for various linguistic
variables. While the results of our analyses indicate a considerable overlap of intermediate phrases and clausal
units, we also showcase syntactic contexts where prosody is needed for disambiguation (e.g. monologs, discourse
markers, dysfluencies, and adverbials).
Article outline
- 1.Introduction
- 2.Syntactic vs. prosodic units and analyses
- 2.1Comparing basic concepts and definitions
- 2.2Comparing syntactic and prosodic structures of speech
- 2.3Approaches to analyses at the syntax-prosody interface
- 3.Database and methodology
- 3.1Corpus
- 3.2Prosodic segmentation
- 3.3Syntactic segmentation
- 4.Results
- 4.1Correspondence of intonation units and syntactic units
- 4.2Lengths of intonation units and syntactic units
- 4.3Analyzing the necessity of prosody for syntactic segmentation
- 5.Discussion
- 6.Conclusion
-
Notes
-
References
-
Appendix