计算机与数字工程2025，Vol.53Issue(2)：499-504,6.DOI:10.3969/j.issn.1672-9722.2025.02.035
基于注意力胶囊网络的口语理解联合模型
A Joint Model for Spoken Language Understanding Based on Attention Capsule Networks
摘要
Abstract
Intent recognition and slot filling are two important tasks in spoken language understanding systems,and it has be-come a trend to jointly learn the two tasks.However,the existing joint models sequentially annotate the slots while obtaining the sen-tence intent,and do not explicitly preserve the hierarchical relationship among chars,words,slots and intents.This paper designs a joint model for spoken language understanding based on attention capsule network.The model dynamically fuses the input char infor-mation and word information,fully considering the importance of char information and word information in spoken language under-standing.Through self-attention routing and rerouting,the bidirectional information flow of intent and semantic slots is realized.Ex-periments show that the model has achieved good results on the CAIS and ECDT-NLU datasets,with an intent recognition accuracy rate of 94.82%on CAIS,a slot filling F1 score of 88.36%,and intent recognition accuracy on ECDT-NLU reaches 79.94%,and the slot filling F1 score reaches 49.62%,which achieves better performance than other models.关键词
对话系统/口语理解/意图识别/语义槽填充/注意力胶囊网络Key words
dialogue system/spoken language understanding/intent recognition/slot filling/attention capsule network分类
计算机与自动化引用本文复制引用
李维乾,杨卓琳,蒋良..基于注意力胶囊网络的口语理解联合模型[J].计算机与数字工程,2025,53(2):499-504,6.基金项目
智能网络与网络安全教育重点实验室开放基金项目(编号:NS202118901)资助. （编号:NS202118901）
