Replication Data for: Predicting Russian aspect by frequency across genres

We ask whether the aspect of individual verbs can be predicted based on the statistical distribution of their inflectional forms and how this is influenced by genre. To address these questions, we present an analysis of the “grammatical profiles” (relative frequency distributions of inflectional forms) of three samples of verbs extracted from the Russian National Corpus, representing three genres: Journalistic prose, Fiction, and Scientific-Technical prose. We find that the aspect of a given verb can be correctly predicted from the distribution of its forms alone with an average accuracy of 92.7%. Remarkably, this accuracy is statistically indistinguishable from the accuracy of prediction of aspect based on morphological marking. We maintain that it would be possible for first language learners to use distributional tendencies, in addition to morphological and other cues (for example semantic and syntactic cues), in acquiring the verbal category of aspect in Russian.

本研究旨在探究两大问题：一是能否通过单个动词的屈折形式（inflectional forms）的统计分布，预测其动词体属性；二是文本体裁对该预测任务的影响机制。为解答上述问题，本研究从俄罗斯国家语料库（Russian National Corpus）中抽取三类动词样本，涵盖新闻散文、虚构类文本与科技散文三种体裁，并对其“语法轮廓（grammatical profiles）”——即屈折形式的相对频率分布——开展分析。研究结果显示，仅依托动词形式的分布特征，即可准确预测目标动词的体属性，平均准确率达92.7%。值得注意的是，该准确率与基于形态标记（morphological marking）预测动词体属性的准确率在统计学上无显著差异。本研究认为，俄语母语学习者在习得动词体范畴时，除可借助形态标记及语义、句法等其他线索外，亦可利用分布倾向性辅助学习。