Knowledge Representation, Sharing and Retrieval on the Web

Abstract

By “knowledge retrieval”, we refer to the automatic retrieval of statements permitting a tool to make logical inferences and answer queries precisely and correctly, as opposed to retrieving documents or statements “related to” the queries. Given the ambiguity of natural language and our current inability to make computers “understand” it, the knowledge has to be manually encoded and structured using a formal graphic/textual language and ontologies (structured catalogs of categories and associated constraints of use). TheWeb currently contains a lot of data, more and more structured data (databases, structured documents) and simple metadata but very little knowledge as defined above, i.e. very few knowledge representations. Moreover, this knowledge has been encoded using various languages and unconnected or loosely connected ontologies, and following different representation conventions. Hence, currently, not only knowledge sources are rare but each require the development of a special wrapper for their knowledge to be interpreted and hence retrieved, combined or exploited. This article reviews various projects concerning knowledge representation, sharing and retrieval on the Web, then details requirements for a “Semantic Web” and illustrates them with notations, conventions and cooperation rules from our own tool, WebKB-2. Knowledge retrieval mechanisms and interfaces used in WebKB-2 are also given as illustrations.