Prosodic information

 

The area of word prosody, and, more generally, the description of other prosodic units which have quasi-morphemic functions, is gradually emerging as an important area for spoken language lexica. For present purposes, prosodic properties are defined as properties of word forms which are larger than phonemes.  Further specification in phonetic terms (e.g. F0 patterning ) and in semantic terms (e.g. attitudinal meaning) may also be given but is not essential for present purposes.

One type of lexical information on prosody pertains to phonological or morphological  properties of words, such as Swedish pitch  accents , or stress  positions in words. Some aspects of word prosody are predictable on the basis of the regular phonological and morphological  structure of words, but some are idiosyncratic.  Examples in English where word stress  is significant include the noun-verb alternation type as in export - /eksp:t/ (Noun), /eksp:t/ (Verb). In German, word stress  is significant for instance in distinguishing between compound  separable particle verbs and derived inseparable prefixed  verbs as in übersetzen - /y:bztsn/ (compound)  vs. /y:bztsn/ (derivation).  

It has been shown [Waibel (1988)] that taking word prosody into account in English can produce a significant improvement in recognition rate.

In addition, there is lexical information associated with prosodic units which occur independently of particular words, and therefore may themselves be regarded as lexical signs and be inventarised in a prosodic lexicon  [Aubergé (1992)]. To give a highly simplified example in a basic attribute-value notation, a prosodic lexicon  for an intonation  language might have the following structure.

Terminal_1:
  <phonetics pitch> = fall
  <semantics>       = statement or instruction.

Terminal_2:
  <phonetics pitch> = rise
  <semantics>       = question or polite instruction.

This kind of information, in which prosodic categories function as a kind of morpheme  with an identifiable meaning, is generally not regarded as lexical information, but treated as a separate layer of organisation in language. Intonation  is being taken increasingly into account for prosodic parsing  in two main senses of this term:

1. Analysis of speech signal in respect of the fundamental frequency (F0 , F-zero) trajectory, for speech recognition,  in relation to words, sentences and dialogue units.
2. Analysis of sentence structure for the generation of intonation  patterns in speech synthesis. 

Prosodic representation in the lexicon is in general restricted to the prosodic properties of words, such as stress  position in English, Dutch, and German words, or tonal accent  in Swedish words, or to rhythmically relevant units such as the syllable  and the foot . For spoken language processing in which prosody plays a role, it is also necessary to include an inventory of prosodic forms, and their meanings, which play a role at the sentence level, independently of specific words: i.e. a prosodic lexicon .

It should be borne in mind that in linguistics, ``prosody'' currently has a broader meaning, and covers all properties of pronunciation which are not directly concerned with defining consonants and vowels. Prosody in this sense covers, for example, syllable structure and phonological word phonotactics, as well as the more traditional categories of intonation, accent, and phrasing.

The IPA  defines symbols for representing lexical and non-lexical types of prosody, and a subset (for word prosody) has been encoded in the SAMPA  alphabet. However, the state of knowledge in the area of prosody is less stable than in the area of segmental word structure, and a range of different conventions is available [Bruce (1989)]; in this area, there are SAMPA  ``dialects '', for instance replacing SAMPA  " and % for primary and secondary stress  by the more iconic (single quote) and (two single quotes) or " (double quote).

The ToBI  (Tones and Break Indices) transcription , originally developed for American English, has now been applied to several languages (see also Chapter 5).

In oriented spoken language lexicography within the VERBMOBIL  project, attribute-based formal representations of prosodic features in the lexicon have been developed using the ILEX  (Integrated Lexicon) model and the lexical knowledge representation language DATR  [Bleiching (1992), Gibbon (1991)].

There is an increasing tendency no longer to regard prosodic representations as totally exotic and quite unlike anything else. But there is still insufficient consensus on lexical prosodic features to permit generally valid recommendations to be made for prosodic representations in the lexicon. For most purposes, plain SAMPA  or ToBI  style symbols will be adequate. For covering new ground with extended lexica for use with discourse phenomena at the dialogue level, a lexical knowledge representation language with a more general notation, as illustrated above, may be more appropriate.