摘要
本文针对广播电视节目的多模态语料库标注存在的问题,采用深度学习框架,结合音视频识别和自然语言分析等技术创建多模态智能分析模型,实现智能化标注和语料库构建。最后,本文将该模型应用到IPTV播出的英语口语会话修补多模态语料库(MCCECSER)的建库实践,说明多模态智能分析模型具备一定技术创新性和实用性。
Aiming at the problems of multimodal corpus tagging of radio and television programs,this paper uses a deep learning framework,combined with audio and video recognition and natural language analysis technology to create a multimodal intelligent analysis model to achieve intelligent tagging and corpus construction.Finally,this paper applies the model to the practice of building a multimodal corpus of English spoken conversation repair(MCCECSER)broadcasted on IPTV,which shows that the multimodal intelligent analysis model has certain technological innovation and practicality.
作者
胡志兵
姚剑鹏
Hu Zhibing;Yao Jianpeng(Ningbo Radio and Television Group,Zhejiang 315000,China;Ningbo University of Technology,Zhejiang 315211,China)
出处
《广播与电视技术》
2022年第8期43-46,共4页
Radio & TV Broadcast Engineering
基金
国家社科基金项目“英语学习者课堂会话自我修补多模态语料库建设与应用研究”(项目编号:19BYY094)资助,为该项目阶段性研究成果。