Modeling Human Morphological Competence.

Yohei Oseki,Alec Marantz

doi:10.3389/fpsyg.2020.513740

Abstract

One of the central debates in the cognitive science of language has revolved around the nature of human linguistic competence. Whether syntactic competence should be characterized by abstract hierarchical structures or reduced to surface linear strings has been actively debated, but the nature of morphological competence has been insufficiently appreciated despite the parallel question in the cognitive science literature. In this paper, in order to investigate whether morphological competence should be characterized by abstract hierarchical structures, we conducted a crowdsourced acceptability judgment experiment on morphologically complex words and evaluated five computational models of morphological competence against human acceptability judgments: Character Markov Models (Character), Syllable Markov Models (Syllable), Morpheme Markov Models (Morpheme), Hidden Markov Models (HMM), and Probabilistic Context-Free Grammars (PCFG). Our psycholinguistic experimentation and computational modeling demonstrated that “morphous” computational models with morpheme units outperformed “amorphous” computational models without morpheme units and, importantly, PCFG with hierarchical structures most accurately explained human acceptability judgments on several evaluation metrics, especially for morphologically complex words with nested morphological structures. Those results strongly suggest that human morphological competence should be characterized by abstract hierarchical structures internally generated by the grammar, not reduced to surface linear strings externally attested in large corpora.

Highlights

Chomsky (1957) seminally argued that the grammar categorically generates grammatical sentences of the language, while speakers gradiently judge acceptable sentences of the language, as summarized below:“The fundamental aim in the linguistic analysis of a language L is to separate the grammatical sequences which are the sentences of L from the ungrammatical sequences which are not sentences of L and to study the structure of the grammatical sequences
Syntactic competence should be reduced to surface linear strings externally attested in large corpora, where grammaticality and acceptability are isomorphic
To recapitulate, going back to the original research question, the results of our psycholinguistic experimentation and computational modeling converged on the conclusion that human morphological competence should be characterized by abstract hierarchical structures, and cannot be reduced to surface linear strings

Summary

Introduction

Chomsky (1957) seminally argued that the grammar categorically generates grammatical sentences of the language, while speakers gradiently judge acceptable sentences of the language, as summarized below:“The fundamental aim in the linguistic analysis of a language L is to separate the grammatical sequences which are the sentences of L from the ungrammatical sequences which are not sentences of L and to study the structure of the grammatical sequences. On this internalist view, syntactic competence should be characterized by abstract hierarchical structures internally generated by the grammar (Everaert et al, 2015; Ott, 2017), where grammaticality and acceptability correspond to linguistic representation and processing, respectively, the familiar competence-performance distinction. Lau et al (2016) recently claimed that the grammar gradiently determines grammatical sentences of the language through probabilities of linear strings without hierarchical structures. On this externalist view, syntactic competence should be reduced to surface linear strings externally attested in large corpora, where grammaticality and acceptability are isomorphic. Whether syntactic competence should be characterized by hierarchical structures or reduced to linear strings has been actively debated in the cognitive science literature

Methods

Results

Discussion

Conclusion