A challenging aspect of transcribing prosody is that there is a substantial level of variability. For example, an f0 that is high in a speaker's range on a salient syllable can mark a pitch accent but so can a lower but still high-for-this-speaker f0 on another salient syllable. Moreover, a high f0 might mark a pitch accented word or it could be a phrase or boundary tone. In order to determine the appropriate ToBI transcription, the entire utterance must be parsed, that is, understood as a whole to determine how the high and low tones are implemented. This tutorial will begin by introducing utterances with unambiguous ToBI annotations and will gradually include utterances for which even experienced labellers might, in part, disagree. ToBI labelling is a new endeavor compared to, say, phonetic transcription. This should not be disturbing if one recognizes that ambiguous cues are present in other types of speech annotation. For example, the acoustic cues for a canonical /t/ might not be present, e.g. in some renditions of the word "butter". However, the existence of a well established lexicon allows annotators to recognize this as a variation of a single lexical item and label consistently. At this writing, the extent of variations of ToBI labels is not fully understood and this results in some reasonable disagreement in prosodic parses in some renditions of some utterances, particularly in spontaneous speech.

Play armani10.wav: - download armani10.TextGrid


<armani10.wav> with accompanying Praat TextGrid.

The rate of vibration of the vocal folds is what we hear as the pitch, and this is represented in the second window as a semi-continuous blue line superimposed on a different representation of the speech signal, the spectrogram (see "What is the spectrogram?" below). When you play this utterance, notice where you hear a higher pitch. Do these words and syllables correspond to the places where the F0 contour (also called the pitch track or the f0 track, although f0 and perceived pitch are not exactly the same thing) is higher, as indicated by the blue lines?

Underneath the spectrogram with the blue pitch contour is the set of four thinner white boxes that make up the four tiers in the ToBI labeling window. Unlike the speech displays, these boxes are text writeable and make up the tiers where the labeller types in the ToBI transcription labels. The top white box is the Tone tier, and the third box is the Break Index tier. These two tiers represent the core prosodic analysis. The Tone tier is the part of the transcription that corresponds most closely to a phonological analysis of the utterance's intonation pattern. It consists of labels for distinctive pitch events, transcribed as a sequence of high (H) and low (L) tones marked with diacritics indicating their intonational function. Tones function either as pitch accents, parts of pitch accents, or as phrase accents marking the edges of two types of intonationally marked prosodic units (phrases). These categories are based on the work of Janet Pierrehumbert (1980) and joint work by Mary Beckman and Janet Pierrehumbert (1986, 1988).

The Break-Index tier captures the prosodic grouping of the words in an utterance by labelling the end of each word for the subjective strength of its association with the next word, on a scale from 0 (for the strongest perceived conjoining/ weakest perceived boundary, as in doncha for don't you) to 4 (for the most disjoint, i.e. at the end of an intonationally marked major phrase). These categories of association strength or 'break indices' are based on work by Mari Ostendorf, Patti Price, Stefanie Shattuck-Hufnagel, and their associates (Price et al., 1991). The two highest break indices (3 and 4) are equated with two kinds of prosodic groupings (or phrases) that are marked intonationally; groupings of Intonational Phrases are not marked in the ToBI system.

The Orthographic tier is the third white box. It contains a straightforward transcription of all of the words in the utterance, in ordinary English orthography. The word transcriptions are aligned with their locations in the speech waveform. For labellers using Praat or a similar labelling computer application, the convention is to place the orthographic label for a word between two marks that delineate the approximate time interval in the signal that corresponds to the utterance of that word and placing to mark silence between words, if any. The orthographic tier is arguably not part of any core prosodic analysis, except inasmuch as the labels on this tier can be used to interface the transcription to dictionary entries which do indicate such things as which syllable is likely to be most stressed in each word, prosodic information which is not otherwise included in the ToBI system (more on this below).

The Miscellaneous tier is the bottom white box in this display. It is essentially a 'comment' tier that can be used to mark events such as coughs, laughter, long silences and other non-speech events and is traditionally marked with angle brackets (e.g. cough< ... >cough). Like the orthographic tier, it can include many events that are arguably not part of prosody per se. However, many events that are typically marked on the Miscellaneous tier are important for interpreting the analyses on the Tone tier and Break-Index tier, because they disrupt the smooth rhythm of the utterance or interrupt the intonation contour. Labels on this tier usually come in pairs, to mark the beginning and end of each event interval; one exception is the label 'disfl', which often stands alone to flag the occurrence of a perceived disfluency of some type.

[The next two paragraphs contain further discussion of what is not captured by a ToBI transcription; read them if you are curious about this question. Otherwise, skip directly to Chapter 2.0.]

An example of the non-categorical aspects of prosody which we leave out (in accordance with the second principle) is the local tempo of each word in the utterance, which we feel could be more accurately and directly captured by some quantitative measure such as normalized segment duration (e.g., Campbell, 1992) than by any symbolic transcription such as an arbitrary division into, say, categories '1', '2', and '3' (for 'slow', 'medium', and 'fast' tempi).

A categorical aspect of prosody which we leave out (in accordance with the third principle) because it should be fairly predictable is the marking of the stressed and unstressed syllables within each word. By this level of stress we mean the word-internal alternation between more stressed and less stressed syllables, where the relative prominence of any pair of syllables is fairly fixed and can be thought of as inherent to the word's dictionary entry.