Abstract

Lexicographers make many choices when compiling dictionary entries. Data-driven approaches can better inform such decisions. This paper proposes integration of two such sources, corpus data and perceptual survey data. A survey was distributed to 100 native speakers of English. Participants were asked to provide a definition and example sentence for 18 polysemous English words. After providing definitions, participants were presented with five definitions and asked to rank them by importance. Patterns in the rankings were analyzed to see what factors predicted higher ratings. Frequency was found to predict rankings, while concreteness was also shown to be moderate a predictor and have a potentially strong effect on certain words. Part of speech was not found to be a significant predictor of rank. Results suggest that crowdsourcing may be useful for certain linguistic tasks and that both corpus and perceptual data can be useful in making lexicographic choices.

中文翻译：

---

通过语料库和感知数据告知词典选择

摘要

词典编辑人员在编译词典条目时会做出许多选择。数据驱动的方法可以更好地为此类决策提供依据。本文提出了两种这样的资源的整合：语料库数据和感知调查数据。已向100位以英语为母语的人进行了一项调查。要求参与者提供18个多义英语单词的定义和例句。提供定义后，向参与者展示了五个定义，并要求他们按重要性进行排名。分析排名中的模式，以查看哪些因素预测了较高的评分。频率被发现可以预测排名，而具体性也被证明是适度的预测器，并且对某些单词可能具有强大的影响。言语的一部分没有被认为是等级的重要预测因子。