Building the Semantic Web by Mass Collaboration

Though research on the Semantic Web has progressed at a steady pace, its promise is yet to be realized. One major difficulty has been that the Semantic Web requires global consistency and quality, properties that are nearly impossible to achieve in such a large scale, mass collaborative system. We circumvent these problems by adopting instead the notion of local consistency. We propose individual maintenance of beliefs and trusts in other users. This paper defines global properties for combination functions which merge such trusts and beliefs, and defines a class of functions for which merging may be done locally while being equal to having done the merging locally. We show an equivalence between merging trusts and merging beliefs, allowing use of whichever is more efficient. We give examples of specific functions which fit into the framework and apply them to data from the Epinions knowledge-sharing site and our implemented, real-world testbed, the BibServ bibliography server. Experimental results confirm that the methods are robust to maliciousness and noise, and do not require unreasonably high expectations of user quality. We hope that these methods will help move the Semantic Web closer to fulfilling its promise.