A weighted constraint grammar analysis of word-medial syllabification in English

다국어 초록 (Multilingual Abstract)

This paper proposes a weighted constraint grammar analysis that best models English syllabification as reported in Eddington et al. (2013). After we examine the factors governing the syllabification of medial consonants in two-syllable words, a set of constraints are formulated to address these factors. Based on these constraints, an initial grammar is constructed, to which we submit distribution information about medial syllabifications obtained from Eddington et al.’s (2013) syllabification survey data. We then perform learning simulations using Exponential Noisy Harmonic Grammar, a weighted constraint grammar designed especially to deal with constraints with negative weight (Pater 2009). As a result of the learning, an output grammar is produced in which each constraint is assigned a numerical weight. From the output grammar, we generate predicted syllabifications, which are then compared with observed syllabifications to evaluate the success of the analysis. We measure the predictive performance of the proposed analysis in terms of Root Mean Square Error and R², both of which provide a positive assessment.

목차 (Table of Contents)

• 1. Introduction 2. Word-medial syllabification in English 3. English medial syllabification in a weighted constraint grammar 4. Conclusions and discussion References