The first progress report: 27/2/98 (project start date: October 1997)

 


Methodology

We have developed our experimental methodology. The materials for recording are ready, and we have decided on our criteria for subject choice. Moreover, a first version of the prosodic transcription system which we will use in the intonational analysis of the data is ready.


Materials

Our experimental materials consist of:

(1) a set of sentences intended to elicit basic contours such as the intonation patterns on simple declaratives, wh-questions, questions without morphosyntactic markers etc.

(2) a reading passage the content of which the experimental subjects are likely to be familiar with; a version of the fairy tale 'Cinderella'

(3) the fairy tale is then retold by the subjects

(4) a set of maps for the 'map task' (each speaker gets a map, one has a route, but the other one has not. The speaker who's got the route describes this route to the other speaker, so that that speaker can draw the route on his or her map. However, the maps aren't exactly the same, and this sparks off discussion; for more info on the map task, see Anderson et.al (1991). The HCRC Map Task Corpus. Language and Speech 34 (4).

(5) a topic for free conversation (smoking and tobacco advertising)



The IViE system for prosodic labelling

'IViE' stands for 'Intonation Variation in English'. The labelling system has 4 tiers:

(1) an orthographic tier
(2) a rhythmic tier
(3) an pitch movement tier (to note pitch accent realisation effects)
(4) a phonological tier

On the orthographic tier (1), the words spoken are transcribed, and on the phonological tier (4), a phonological transcription of intonation is entered. So far, the IViE system resembles ToBI. The other two tiers (rhythmic and pitch movement) are new. Labellers using IViE transcribe the speech data on the rhythmic and the phonetic tier before they enter a phonological transcription. In a sense, these tiers 'unpack' the phonological transcription, and the phonological transcription is arrived at step-by-step.

The rhythmic tier allows for marking of rhythmically prominent syllables. These are the syllables significant pitch movements are anchored to.

The pitch movement tier keeps track of the pitch movements transcribers see in the F0 trace and hear in conjunction with the stressed syllables. This tier will allow us to compile a record of the mapping between particular pitch accents and their phonetic realisations in different varieties of English spoken in the British Isles (i.e. a pitch accent H*+L ('falling accent') can be realised in a number of ways which do not affect its phonological status).

The IViE system is based on comparative labelling system proposed by Grabe (1998a, reference on home page), and we're currently writing a labelling guide. The IViE system may be used in conjunction with xwaves under UNIX, similarly to the ToBI system for prosodic labelling (and this is what we use IViE). Alternatively, labelling can be carried out by hand, as long as a signal processing package is available which can carry out basic acoustic analysis.



Recordings

The first set of data (Cambridge English) has been recorded, and a subsection of these data \has been labelled. Everything but our first version of the map task worked very well. We found that the items on the two maps were too similar, and subjects assumed that we had made mistakes when designing the map (the items were very similar because we were hoping to elicit spontaneous speech data illustrating segmental effects on F0). A second version of the map task was designed, and the differences were made more obvious. The map task data was re-recorded and worked well.


Speaker choice

Our subject groups are 16-17 years of age, and they attend the same schools. We aim to record speakers who were born in Leeds/Dublin/Newcastle etc, and who grew up there, and we concentrate speakers who can be described as being 'lower middle class'. 12 speakers from each variety are recorded; six male speakers and six female speakers.