An inheritance-based description of Bulgarian noun inflection

abstract morphophonemic representation, DATR This paper discusses an analysis of Bulgarian noun inflection that uses non-monotonic inheritance and abstract morphophonemic representation. The analysis is encoded in the lexical knowledge representation language DATR. In this discussion, we approach Bulgarian noun inflection in terms close to those of Network Morphology (cf. Corbett & Fraser (1993)) and related theories of inflection. Thus, an essential feature of the analysis is that inflectional types are represented as objects within a nonmonotonic inheritance hierarchy that gives an explicit account of regularities, subregularities and exceptions through highly constrained lexical entries, default inflectional class(es) and non-productive class(es). Our goal has been to make a precise and explicit description allowing economical morphophonemic representations to simplify the description of alternations and the morphotactics. For example, the phoneme i has at least three sources: i1, i2 and ы. The i that is not involved in morphophonemic alternations is transliterated as i, while the i that triggers 2 nd palatalization is marked in the lexicon as 2 but surfaces as i in the generated forms 1. This may look like diachronic description but is not; rather our description is synchronic and applicationoriented since the use of deep morphophonemic representations is 1 For details on the transcription cf. the table i