相似性(几何)
计算机科学
自然语言处理
心理学
语音识别
人工智能
模式识别(心理学)
图像(数学)
标识
DOI:10.1080/23273798.2022.2157028
摘要
Language is constantly evolving through speakers’ natural innovations. However, our understanding of how subtle linguistic innovations are processed is still surprisingly limited. To date, most studies investigating processing of non-conventional language have focused on metaphors. Using both brain event-related potentials (ERPs) and offline judgements, the present study investigated the acceptability, as well as the temporal dynamics, associated with processing novel and conventional multiword units. By manipulating both the degree of similarity and degree of conventionality, the results revealed that the acceptability of novel items hinges on their similarity to known multiword units. ERPs revealed that similarity modulated a late positive component (LPC) 550–750 ms after stimulus presentation; this LPC was significantly correlated with acceptability. Additionally, processing of novel multiword units was not modulated by exposure to related multiword units, indicating that similarity-based processing was not dependent on recent prior exposure.
科研通智能强力驱动
Strongly Powered by AbleSci AI