Abstarct Besides explicit inference of word meanings, associating words with diverse contexts may be a key mechanism underlying vocabulary learning through reading. Drawing from distributional semantic theory, we developed a text modification method called reflash to facilitate both word-context association and explicit inference. Using a set of left and right arrows, learners can jump to a target word’s previous or subsequent occurrences in digital books to synthesize clues across contexts. Participants read stories with target words modified by reflash-only, gloss-only, gloss + reflash, or unmodified. Learning outcomes were measured via Vocabulary Knowledge Scale and a researcher-developed interview to probe word-context association. We modeled the learning trajectories of words across five weeks among three adolescent L2 English learners (113 word-learner pairings) using Bayesian multilevel models. We found that reflash-only words made more gains than words in other conditions on both outcomes, controlling for key covariates such as types of existing knowledge. Our analysis also revealed that context synthesis may be particularly useful for learning specific types of words like homonyms, which has significant pedagogical implications.

中文翻译：

---

上下文综合通过阅读加速词汇学习：分布语义理论对二语词汇研究的启示

摘要 除了词义的显式推断外，将单词与不同的上下文相关联可能是通过阅读学习词汇的一个关键机制。借鉴分布式语义理论，我们开发了一种称为 reflash 的文本修改方法，以促进单词上下文关联和显式推理。使用一组向左和向右箭头，学习者可以跳转到目标词在数字书籍中先前或随后出现的地方，以综合上下文中的线索。参与者阅读目标词被仅重述、仅注释、注释 + 重述或未修改的目标词的故事。学习成果是通过词汇知识量表和研究人员开发的调查词上下文关联的访谈来衡量的。我们使用贝叶斯多层次模型对三名青少年 L2 英语学习者（113 个单词-学习者配对）在五周内的单词学习轨迹进行了建模。我们发现，在控制现有知识类型等关键协变量的情况下，仅重现单词比其他条件下的单词获得更多收益。我们的分析还表明，上下文综合可能对学习特定类型的单词（如同音异义词）特别有用，这具有重要的教学意义。