摘要： 针对海量网络数据的描述与管理问题，提出音视频语义信息抽取系统。根据网络音视频流数据包的特征，采用正则表达式制定协议包识别规则。在此基础上，通过分析音视频传输协议结构，提取音视频数据从网络接口层、网络层、传输层到应用层的多层语义信息，并形成UCL标签。实验结果表明，在保证流完整性的前提下，该算法的召回率和准确率均达到80%以上。

关键词: 网络音视频数据, 多层语义理解, 语义抽取, 正则表达式, UCL语义标签

Abstract: In allusion to the problems of describing and managing massive network data, this paper presents a semantic information extraction system for audio/video. According to the characteristics of network audio/video streams, it formulates protocol identification rules using regular expression. On this basis, multilayer semantic information from network interface layer, network layer, transport layer and application layer is extracted by analyzing the structure of audio/video transmission protocol. Meanwhile a Uniform Content Locator(UCL) label is constituted. Experimental results show that the proposed algorithm has recall and accuracy both over 80% under the requirement of complete flows.

Key words: network audio and video data, multilayer semantic comprehension, semantic extraction, regular expression, Uniform Content Locator(UCL) semantic label

中图分类号: 

• TN912.3