Melody learning and long-distance phonotactics in tone
- PDF / 1,191,883 Bytes
- 51 Pages / 439.37 x 666.142 pts Page_size
- 57 Downloads / 191 Views
Melody learning and long-distance phonotactics in tone Adam Jardine1
Received: 27 November 2018 / Accepted: 27 February 2020 © Springer Nature B.V. 2020
Abstract This paper presents evidence that tone well-formedness patterns share a property of melody-locality, and shows how patterns with this property can be learned. Essentially, a melody-local pattern is one in which constraints over an autosegmental melody operate independently of constraints over the string of tonebearing units. This includes a range of local tone patterns, long-distance tonepatterns, and their interactions. These results are obtained from the perspective of formal language theory and grammatical inference, which focus on the structural properties of patterns, but the implications extend to other learning frameworks. In particular, a melody-local learner can induce attested tone patterns that cannot be learned by the tier projection learners that have formed the basis of work on learning long-distance phonology. Thus, melody-local learning is a necessary property for learning tone. It is also shown how melody-local learners are more restrictive than learning directly over autosegmental representations. Keywords Tone · Learnability · Computational phonology · Representation
1 Introduction 1.1 The proposal Long-distance generalizations pose a challenge for learning, because a learner must discover dependencies that may hold over an unbounded distance. Work on learning phonotactics has made gains towards solving this problem by adopting representational mechanisms that allow a learner to discover long-distance dependencies. Specifically, previously proposed phonotactic learners have either made reference to
B A. Jardine
[email protected]
1
Department of Linguistics, Rutgers University, 18 Seminary Pl, New Brunswick, NJ, USA
A. Jardine
precedence relations (Heinz 2010a), tier projection (Hayes and Wilson 2008; Goldsmith and Riggle 2012; Jardine and Heinz 2016; McMullin and Hansson 2019; Jardine and McMullin 2017; Gallagher and Wilson 2018), or a generalization of the two (Graf 2017). However, the specific problem of learning tonal patterns has not been yet studied in this way. This paper fills this gap by studying the learnability of long-distance phonotactic patterns in tone. This reveals that previously proposed representational mechanisms for learning phonotactic patterns are insufficient for learning the types of patterns that are found in tone. This article instead argues that long-distance phonotactic patterns in tone must be learned using a distinct notion of a melody, or string of unique components of surface tone. Evidence is given below that tone patterns share a property of melody-locality, meaning that they are describable by the intersection of a set of local constraints over strings of tone-bearing units (TBUs) and a set of local constraints over a melody. For example, in Prinmi (Tibeto-Burman, China; Ding 2006; Hyman 2009), words must have exactly one span of H tones, which can be either one (1a) or two (1b) TBUs lo
Data Loading...
