Towards a model for the prediction of Chinese novel verbs

As previous word adoption models, though proposing potential factors that influence the survival of neologisms, receive little empirical examination, this corpus-based study compares the performance of two such models by providing clear operational criteria for each factor in the models and, consequently, proposes a hybrid model that improves the previous results. We focus on seventy-seven Chinese novel verbs that appeared about ten years ago, defining their survival/failure in the real world, and examine the accurate prediction ratio of the two models. Both models display an overall accuracy of about 60 percent. However, as certain factors, e.g., unobtrusiveness, appear to be invalid predictors for the Chinese data, we attempt to improve the results by deleting inappropriate factors and by adjusting the weightings. As the overall accuracy was improved to about 70 percent, we suggest that this study would shed light on the potential factors that influence the survival of Chinese novel verbs.