Contrastive analysis of English and Spanish intonation using computer corpora - a preliminary study.

The thesis presents an account of the design, construction and analysis of a machine-readable corpus of transcribed spoken Spanish. The corpus was compiled from transcriptions of broadcast and conversational speech and was transcribed with prosodic marks by the researcher. Syllable boundaries were also marked. The design was aimed at compatibility with the Lancaster Spoken English Corpus, which already exists, and the primary objective of the research was to discover comparative information about differences between Spanish and English prosody. Analysis by computer showed differences between the two languages in terms of mean tone-unit lengths and in the frequency of occurrence of different tones. An experiment to investigate the degree to which trained phoneticians (including the researcher) agree in transcribing pitch movements by drawing "pitch curves" showed a reasonable degree of agreement as measured by calculating correlation coefficients, though agreement with computer-extracted fundamental frequency curves was less clear-cut. The thesis discusses the possibility of storing such fundamental frequency information along with the "manual" transcription in the corpus in future development of the work.