Enabling a Web of Knowledge

A major obstacle impeding progress on the “web of data” is content creation—a difficult, tedious, and timeconsuming task. How do we make human-scalable, user-friendly tools to enable the web of data? Content integrity is also a major concern. How do we engender confidence in results returned from the web of data? Although seemingly unrelated, we show in this paper that it is exactly their relationship that is the key to solving both problems. As we show in this paper, we can semi-automatically derive both data and metadata from data-rich web pages to create a web of data that we then superimpose over these data-rich web pages. We link the web of data to the current web of pages, resulting in a higher-order “web of knowledge.” This web of knowledge provides provenance and thus engenders the confidence necessary to raise the level of the web from “data” to “knowledge.” We focus mainly on two prototype tools we have implemented: (1) TISP—a tool to automatically generate ontologies for data-rich, machinegenerated web pages and annotate these pages with respect to these generated ontologies and (2) FOCIH—a tool to semi-automatically generate user-specified ontologies and annotate web pages with respect to these user-specified ontologies. We also briefly survey a suite of tools we and others are creating that have the potential to further enable this web of knowledge. As a measure of how successful these tools are and potentially can be, we summarize precision and recall results, which indicate the degree of scalability we can hope to achieve in enabling a web of knowledge.