The area of linguistic word modelling in the ASL project involves the areas syllable phonology, morphophonology and word prosody. This paper is concerned with the area of syllable phonology, in particular the linguistic knowledge component which is required by a syllable parser in order to construct syllable and syllable type hypotheses on the basis of underspecified phonetic data. These are then passed to the morphophonological parser for further analysis. In this report a new approach to the phonotactic description of German is proposed which, starting from recent well-motivated developments in phonology (Goldsmith, 1976,1990; Browman & Goldstein, 1986,1989; Bird & Klein, 1990), utilises autosegmental tiers of parallel phonological events (cf. Carson-Berndsen & Gibbon, 1991). This event-based phonotactics for German will be used in order to reduce the search space for lexical hypotheses. Assuming that at the phonetic level all combinations of sounds are possible, the phonotactics reduces the search space to cover only those combinations which are permissible in the language in question. In the ASL context a further source of information can be made use of in order to guide the morphophonological parser, namely a marking in terms of attested and unattested forms with respect to a corpus or lexicon.
Although directly related to the work presented in Carson-Berndsen (1991), this report is concerned primarily with the phonological domain. That is to say, rather than dealing with the temporal annotations at the phonetics/phonology interface, the phonotactic description assumes an abstraction from absolute time, describing instead temporal relations between phonological events. In section 2, the motivation for an event-based phonotactic description is presented, followed by a discussion in section 3 of adequacy criteria and generalisation strategies for phonotactic description. In section 4 the formal framework for the phonotactic constraints relevant for the syllable parser of a speech recognition system are introduced and in section 5 a complete event-based phonotactics of German syllables is given.