Learning Strictly Local Subsequential Functions

Chandlee, Jane; Eyraud, Rémi; Heinz, Jeffrey

doi:10.1162/tacl_a_00198

Cited by 38 publications

(29 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Admittedly, there is not yet a deep understanding of the formal and learnability properties of subregular functions that can model string-to-string mappings. However, recent work has defined a number of relevant classes and identified some of their formal properties (e.g., Chandlee, 2014;Chandlee, Eyraud, & Heinz, 2015;Mohri, 1997), including a number of promising results regarding learnability (Chandlee, Eyraud, & Heinz, 2014;Chandlee & Jardine, 2014). Of particular relevance is the finding that subsequential functions (Mohri, 1997) characterize many attested patterns of longdistance consonant dissimilation (Payne, 2014) and consonant harmony (Luo, 2017), while excluding a number of complex pathologies.…”

Section: Further Issuesmentioning

confidence: 99%

Inductive learning of locality relations in segmental phonology

McMullin

Hansson

2019

Laboratory Phonology: Journal of the Association for Laboratory Phonology

View full text Add to dashboard Cite

This paper reports on a series of artificial grammar learning experiments focused on locality relations in patterns of long-distance consonant agreement (harmony) and disagreement (dissimilation). Participants in experimental conditions were exposed to dependencies affecting stem-suffix pairs of liquids at either a short-range (transvocalic, CVCVLV-LV) or medium-range (beyond-transvocalic, CVLVCV-LV) distance. Two experiments used a poverty of stimulus paradigm, offering no information about the other distance level; participants interpreted short-range interaction as a strictly transvocalic dependency but medium-range interaction as unbounded, generalizing to other distances. Two experiments employed a 'rich stimulus' paradigm, where training data unambiguously indicated the absence of any dependency at the other distance; this enabled probing of specific locality patterns, in particular strictly beyond-transvocalic dependencies. The constraint-based Agreement by Correspondence model of non-adjacent consonant interactions predicts such patterns to be possible for dissimilation but not harmony. The results do not support this hypothesis: Participants seem to have serious difficulty learning strictly beyond-transvocalic dependencies of either kind. Our findings are more consistent with recent proposals that the space of learnable phonotactic restrictions is delimited by the Tier-based Strictly 2-Local class of formal languages. Strictly transvocalic and unbounded dependencies lie within this region, whereas strictly beyond-transvocalic dependencies are more complex, falling outside the learner's hypothesis space.Practice CVCV (8) CVCV-li (8) CVCV-ɹu (8) 24Training CVCVLV (96) CVCVlV-li (96) CVCVɹV-li (96) CVCVlV-ɹu (96) CVCVɹV-ɹu (96) 480 CVLVCV (96) CVlVCV-li (96) CVɹVCV-li (96) CVlVCV-ɹu (96) CVɹVCV-ɹu (96) 480 CVCVCV (96) CVCVCV-li (96) CVCVCV-ɹu (96) 288Testing CVCVLV (32) CVCVlV-li (16) CVCVɹV-li (16) CVCVlV-ɹu (16) CVCVɹV-ɹu (16) 96 CVLVCV (32) CVlVCV-li (16) CVɹVCV-li (16) CVlVCV-ɹu (16) CVɹVCV-ɹu (16) 96 LVCVCV (32) lVCVCV-li (16) ɹVCVCV-li (16) lVCVCV-ɹu (16) ɹVCVCV-ɹu (16) 96 Total = 1560

show abstract

Section: Further Issuesmentioning

confidence: 99%

Inductive learning of locality relations in segmental phonology

McMullin

Hansson

2019

Laboratory Phonology: Journal of the Association for Laboratory Phonology

View full text Add to dashboard Cite

show abstract

“…This sort of linear application of morphophonology is reminiscent of recent work in computational phonology implemented using finite state transducers (e.g. Chandlee 2014; Chandlee, Eyraud, and Heinz 2014;Heinz 2018, etc. ), and left-to-right linear structure building has been proposed in the realm of morphosyntax as well, e.g.…”

Section: Cyclic Spell-outmentioning

confidence: 90%

Seenku argument-head tone sandhi: Allomorph selection in a cyclic grammar

McPherson

2019

Glossa: A Journal of General Linguistics

View full text Add to dashboard Cite

Seenku (Mande, Burkina Faso) displays a complex tone sandhi system, sensitive to phonological, morphological, and syntactic structure. In this paper, I argue that the opaque, phonetically unnatural alternations are best accounted for in an allomorph selection approach, following work by Tsay and Myers (1996), Zhang and Lai (2008), and others on Taiwanese Southern Min sandhi. In this model, lexical entries contain multiple surface allomorphs along with subcategorization frames, and forms that follow the same pattern are abstracted into increasingly general lexical templates or schemas. Morphology is post-syntactic, with syntactic structure matched with lexical entries, along the lines of Distributed Morphology (Halle and Marantz 1993; Embick and Noyer 2007), but with a richer generative lexicon, as in Construction Morphology (Booij 2010b) or analogical approaches to word formation (e.g. Bybee 1995). The domains of application and the interaction of tone sandhi with both itself and other morphotonological processes point to the need for cyclic application, which I implement using phase-based spell out (Uriagereka 1999; Chomsky 2000). This approach allows for productive extension of the patterns without recourse to an overly complicated phonological component.

show abstract

“…Much recent work at the University of Delaware has been to develop a hierarchy for string relations that is analogous to the one for stringsets shown in Figure 2 on page 14. The most notable advance in this regard has been work by Jane Chandlee in her thesis (Chandlee, 2014), and subsequent publications (Chandlee et al, 2014a;Jardine et al, 2014;Chandlee and Heinz, 2014). This work establishes relational counterparts to the Strictly Local stringsets, discusses their significant coverage of empirical phenomena, and explains how they can be learned.…”

Section: Transformationsmentioning

confidence: 99%

“…More recently, Beros and de la Higuera (2014) also show how to generalize subsequential functions in a way that permits a degree of optionality. While subclasses of classes have not been studied the fact that they preserve important aspects of the underlying finite-state transducers and that classes like ISL have automata-theoretic characterizations based on subsequential transducers (Chandlee, 2014;Chandlee et al, 2014a) strongly suggests that subclasses like ISL which permit a degree of optionality are only waiting to be discovered.…”

Section: Non-deterministic Regular Functions and Regular Relationsmentioning

confidence: 99%

“…This is a good argument. However, I have tried to highlight the fact that the subregular regions to which phonological patterns appear to belong also readily lend themselves to a theory of learning (Oncina and Garcia, 1991;Heinz, 2007Heinz, , 2010aChandlee, 2014;Chandlee et al, 2014a;Jardine et al, 2014). Furthermore, if people generalize from data in the way suggested by these learning procedures, it explains the computational nature of the phonological patterns.…”

Section: Organizing Phonological Theory Around These Computational Prmentioning

confidence: 99%

See 1 more Smart Citation

The computational nature of phonological generalizations

Heinz¹

2018

Phonological Typology

Self Cite

108

View full text Add to dashboard Cite

This chapter studies the nature of the typology of phonological markedness constraints and the nature of the typology of the transformation from underlying to surface forms from a computational perspective. It argues that there are strong computational laws that constrain the form of these constraints and transformations. These laws are currently stated most clearly in terms of the so-called subregular hierarchies, which have been established for stringsets (for modeling constraints) and are currently being established for string-to-string maps (for modeling the transformations). It is anticipated that future research will reveal equally powerful laws applicable to non-string-based representations. Finally, this chapter argues that these laws arise as a natural consequence of how humans generalize from data.Wilhelm von Humboldt's phrase "language makes infinite use of finite means" (1836/1999) is oft-cited by Chomsky because not only does it encapsulate an important characteristic of natural language, but it also highlights why generative grammars play an important role in understanding this aspect of language. In brief, the generative grammars are the finite means, but the linguistic knowledge they represent can be applied to unboundedly many linguistic forms. The psychological reality of generative grammars is the powerful scientific hypothesis which underlies all work in generative linguistics.

show abstract

Learning Strictly Local Subsequential Functions

Cited by 38 publications

References 14 publications

Inductive learning of locality relations in segmental phonology

Inductive learning of locality relations in segmental phonology

Seenku argument-head tone sandhi: Allomorph selection in a cyclic grammar

The computational nature of phonological generalizations

Contact Info

Product

Resources

About