Machine learning of event segmentation for news on demand
Stanley Boykin, Andrew Merlino · Communications of the ACM · 2000
The article focuses on deconstructing broadcast news using all sources of input from the multimedia stream Biblioteca de Ciencias y Tecnología Normal Biblioteca de Ciencias y Tecnología 2 1 2006-05-24T21:46:00Z 2006-05-24T21:46:00Z 1 143 792 UCLA 6 1 934 11.6568 Clean Clean 21 false false false MicrosoftInternetExplorer4 /* Style Definitions */ table.MsoNormalTable {mso-style-name:"Tabla normal"; mso-tstyle-rowband-size:0; mso-tstyle-colband-size:0; mso-style-noshow:yes; mso-style-parent:""; mso-padding-alt:0cm 5.4pt 0cm 5.4pt; mso-para-margin:0cm; mso-para-margin-bottom:.0001pt; mso-pagination:widow-orphan; font-size:10.0pt; font-family:"Times New Roman"; mso-ansi-language:#0400; mso-fareast-language:#0400; mso-bidi-language:#0400;} The article focuses on deconstructing broadcast news using all sources of input from the multimedia stream. In order for the topic, event or activity retrieval effort to be useful to the end user, multimedia processing systems depend upon correct story segmentation, tracking and detection. In this article, only story segmentation is discussed. The segmentation addressed in this article, unlike the topic detection and tracking segmentation effort, is based on using all of sources of input from the multimedia stream. The segmentation will be performed on broadcast news sources. The MITRE-developed News on Demand system Broadcast News Navigator (BNN) contains a Finite State Machine (FSM) based story segmentation routine. The segmentation routine allows the end user to view information by stories. The article compares FSM of BNN segmentation system with an automatically induced segmentation system using hidden Markov models.