Automatic Multimedia Fission, Categorizationand Fusion for Personalized Visualization in.pdf

下载文档

3
0
约1.01万字
约 11页
2017-04-09 发布于江苏
举报
版权申诉
保障服务

Automatic Multimedia Fission, Categorizationand Fusion for Personalized Visualization in.pdf

1、本文档共11页，可阅读全部内容。
2、原创力文档（book118）网站文档一经付费（服务费），不意味着购买了该文档的版权，仅供个人/单位学习、研究之用，不得用于商业用途，未经授权，严禁复制、发行、汇编、翻译或者网络传播等，侵权必究。
3、本站所有内容均由合作方或网友上传，本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺！文档内容仅供研究参考，付费前请自行鉴别。如您付费，意味着您自己接受本站规则且自行承担风险，本站不退款、不进行额外附加服务；查看《如何避免下载的几个坑》。如果您已付费下载过本站文档，您可以点击这里二次下载。
4、如文档侵犯商业秘密、侵犯著作权、侵犯人身权等，请点击“版权申诉”（推荐），也可以打举报电话：400-050-0827(电话支持时间：9:00-18:30)。

Automatic Multimedia Fission, Categorizationand Fusion for Personalized Visualization in

Automatic Multimedia Fission, Categorization and Fusion for Personalized Visualization in Multimedia Information Retrieval 供個人化資料撿索的自動化多媒體分拆、歸類及融合技術 PI: Helen Meng Progress Summary for Shun Hing Institute for Advanced Engineering The Chinese University of Hong Kong March 7, 2006 2 Objectives ? Long-term challenge: Automatic organization of massive amounts of multimedia content to facilitate browsing and retrieval by users ? Research issues – Multimedia fission – division / extraction of information constituents at various levels of granularity, e.g. stories / topics paragraphs, images, key terms, etc. – Multimedia categorization – classification according to semantic content – Multimedia fusion – regroup information constituents to facilitate information presentation to the user in relevant forms 3Progress to date ? Multimedia Fission – Automatic story segmentation based on pause duration – Automatic named entity extraction from Chinese speech recognition transcripts ? Multimedia Fusion – Web-derived answer patterns for fusing information constituents ? Multimedia data collection – Web news text and television news broadcasts – Automatic organization of online news text and images Manpower: 1 research associate, 1 research assistant (part year), 1 research assistant professor (half year) 4 Automatic Audio News Story Segmentation Story Boundary Detection Audio/video News Programs Story Segmentation Audio Extraction Speech Recognition Segmented Audio/video News Programs Initial approach: use of audio pauses Multimedia Fission, Fusion, Data Collection 5Experimental Corpus Oct – Dec May – JuneFeb – Apr Periodsumber of programs EvaluationDevelopmentTraining TDT3TDT2 Data Distribution GB encoded Chinese charactersTranscription and ASR output 1998Published Year Radio News BroadcastsNature MandarinLanguage Voice of America (VOA)Source Topic Detection and Tracking Corpus (from Linguistic Data Consoritum) Multimedia Fission, Fusion, Data Collect