Koko - The Internet gorilla.
A search engine for journalists
Jørgen Albretsen, Jack Foster, Niels Mørk
The Danish School of Journalism
Even journalists suffer from information overload. They are usually able to cope with this in traditional newsroom setups.In the present convergence of different types of media, however, they are facing new challenges and must often refrain from using relevant information because of tight deadlines. This calls for products with an immediate benefit to the media world combining theoretical issues from information retrieval with hands-on experience in teaching journalistic research within real-time production conditions. This is done at the Danish School of Journalism.
Koko has been developed at, and is now in use at The Danish School of Journalism. The school receives news feeds from Scandinavian and international news- and photo agencies. Each feed has its own special setup and format. Koko uses filter technology and in-house developed software to group, index, and save this information in a standard off the shelf database. On top of this is put an elaborate system of retrieval software running on servers and accessible on the school's Intranet using standard WWW browsers.
The student and indeed everybody at the school can use all the information in Koko. The name Koko hints at the way journalists normally want information. Powerful and fast, yet precise enough to work from. Without too much learning a user interface, the user can peruse stories, perform searches, and get related material from subject headings and keywords retrieved from the news wire or the captions of photos. Central to Koko is the ability to project the dynamic nature of news feeds onto the user interface. News happen every minute, every hour, all year. Old stories can suddenly become relevant. Koko provides an easy to use retrieval and projection of news wires and photos. An option is to have a news ticker give you the latest on screen.
After a typical search in Koko, the results are presented to the user as shown in Figure 1. Thumbnails of wire photos and headings from ordinary wires are combined and displayed top-down with the latest news at the top. When selecting a particular entry, photos can be downloaded in full size JPEG quality, wires are displayed in fulltext as well as formatted for copy-paste into f.ex. QuarkExpress.
Figure 1. A typical use of Koko
Koko is designed around the hypertext idea that news gathering is a special form of information retrieval, where the jounalist uses relational search patterns to build an opinion of relevant sources.
Our experience shows us that Koko has fulfilled this need for providing up to date information. It is very important that the journalist recognizes his or hers usual working habits in Koko. Journalists base their work on reliable sources. The system described above provides this recognition through its user interface in a fast and reliable way.
The idea is to expand the Koko concept with natural language interface, translation of keywords and concepts from one newsfeed to another. This calls for elements of computational linguistics and a transition towards new World Wide Web standards such as XML. One idea is to develop DTDs that define tagsets for newswires and photos. This work is ongoing at different news providers.
An even more ambitious dimension is to integrate news feeds containing sound and video into Koko. The sound and video files are stored on separate servers and pointed to from Koko, much like the present storage of wire photos. The group behind Koko is currently involved in SMIL-based radio productions at the School of Journalism. The present goal is to have students port their normal radio productions into SMIL adding relevant photos, captions,links, etc. into the player interface, currently Real Player G2. For an example, see (1).
Koko is used for wire feeds to these SMIL productions. We plan to integrate this further. We believe that the concept behind Koko is an ideal frame for developing this special form of information retrieval for a particular group.
(1) Example of SMIL-production at http://rubicon.djh.dk:8080/ramgen/ho/byline/radio/fre11-2/realsystemg2.smi