Using keywords entered by the user, the program selects news site RSS feeds and specific stories to focus on. "The stories are edited for length, and changed to make them more colloquial and suited to speech," says Hammond.The project homepage at Northwestern has several demos that showcase the 3D anchors.
Active voice
For example, information contained within brackets is turned into something that can be read aloud. Where possible, sentences are also changed from the passive to the active voice. "An offer was made by George Bush..." would become "George Bush made an offer..."
The next step is to extract further key terms from these reports and use these to search for supplementary images and videos on sites like YouTube and Google Video. Technorati and Google Blog Search are also used find opinionated blog posts related to the topic of the broadcast. "The software looks for words and phrases indicative of emotional impact," Hammond explains.
Finally, the software generates a script from the collated material. Text-to-speech software then lets computer animated characters - taken from the shoot-em-up computer game Half Life - take on the role of news anchors, with the pictures and videos shown on a screen behind them.
Other characters are used, as interviewees, presenting material taken from blogs posts. This is treated slightly differently - emphasis is added to these characters' synthesised speech to get a more personal message across.
Northwestern's infolab research listing has dozens of research papers relating to the "News At Seven" and other cutting-edge graphics and news apps.
No comments:
Post a Comment
All comments will be reviewed before being published. Spam, off-topic or hateful comments will be removed.