초록
영어
Choi, Hye-Won. 2010. Analyzing and Predicting Ordering Choices in Korean with a Logistic Regression Model. Korean Journal of Linguistics 35-1. 265-289. This paper examines word order alternation in dative construction of Korean and proposes a statistical model of logistic regression based on the analysis of corpus data from Sejong Corpus of Modern Korean (Kim 2000). The model is designed to predict whether each sentence is ordered 'dative before accusative' (DA) or 'accusative before dative' (AD), given the predictor variables such as syntactic weight, discourse givenness, semantic class of verbs, animacy, pronominality, and light-verb-headedness. The model actually predicts the observed orders pretty accurately and reliably. This statistical modeling analysis not only verifies valid factors that truly influence word order but also shows the direction and magnitude of the partial effect of each variable. (Ewha Womans University)
목차
1. Introduction
2. Data
2.1 Predictor Variables
2.2 Response Variable
2.3 Data Summary
3. A Logistic Regression Model
3.1 Effects of Predictor Variables on Order
3.2 Partial Effects of Predictors
4. Model Evaluation
4.1 Summary Statistics
4.2 Classification Accuracy
4.3 Cross-validation
4.4 Predicted vs. Observed Probabilities in Groups
5. Conclusion
References