Abstract: One of the most challenging tasks for infant language learners is to extractmeaning from the speech stream they hear. Linguistic contexts often stand as themain cues for children to infer the semantics of words. For instance, uponhearing the sentence “you broke the window”, to understand the meaning of“broke”, children have to first recognize the two participants “you” and “thewindow” in the utterance, and later establish the causal relation between them.The semantics embedded in linguistic contexts, often referred to as distributionalsemantics, plays an important role in children’s semantic learning (Erickson &Thiessen, 2015; Theakston, Lieven, Pine, & Rowland, 2001; Tomasello, 2003).However, many studies have looked at only specific contextual patterns orselected semantic features to examine semantics in first language acquisition,therefore lacking comprehensiveness in their approach. In this study, weexamined semantics with a computational approach that comprehensivelyaccesses linguistic contexts to generate semantic representations. Using lexicalcausatives as a test case, we processed causative semantics in both Englishchilddirected speech (CDS) and child speech (CS), thereby investigating thesemantic development in parentchild interaction.