News At Seven is a system that automatically generates a virtual news show. Totally autonomous, it collects, parses, edits and organizes news stories and then passes the formatted content to an artificial anchor for presentation. Using the resources present on the web, the system goes beyond the straight text of the news stories to also retrieve relevant images and blogs with commentary on the topics to be presented.
Once it has assembled and edited its material,
News At Seven presents it to the audience using a graphical game engine and text-to-speech (TTS) technology in a manner similar to the nightly news watched regularly by millions of Americans. The result is a cohesive, compelling performance that successfully combines techniques of modern news programming with features made by possible only by the fact that the system is, at its core, completely virtual.
In this, our first deployment of the system, the show produced is a three-minute daily news update, featuring national, international, and human-interest stories, with commentary from blogs on the national story. After the material has been assembled, the system is ready to present the news using preset scripts. The engine, and our extensions to it, allows us to present believable human-like newscasters as well as more imaginative scenes and sets that are only possible because the show is virtual. We also use techniques to make the generated vocal audio more interesting and believable.
Although News At Seven is a very young system, it is already able to create a compelling, cohesive, on-topic news update performance. As we continue to advance its capabilities in finding interesting stories and displaying them in interesting ways, we fully intend to offer a viable replacement to the typical televised news show, tailored to a user's specific interests.
News At Seven - Project
Linear Mode
