Predicting Chinese Abbreviations from Definitions: An Empirical Learning Approach Using Support Vector Regression