A variety of emerging computer applications require large amounts of knowledge about everyday world, encoded in a format accessible to computers.
To collect such knowledge, one promising approach which we've been exploring is to turn to volunteer contributors. As in Wiki projects, the issues of reliability and objectivity of the contributed information often come up. This talk will present the empirically collected data on the acceptability of the information collected. We will also present the mechanisms we have developed for ensuring quality of the data (validation by multiple contributors etc). In addition, the talk will overview design principles which we found useful for engaging contributors in the contribution process (immediate feedback, high contributor lists, etc). Finally, we will comment on the issue of when it may be more appropriate to extract knowledge from existing sources (eg is-a relations may be extractable from the text already on the web), vs. when it may be more appropriate to construct a knowledge resource from scratch, by turning to volunteers.
.gif)
GNU Free Documentation License (details)
