Query Evaluation with Asymmetric Web Services

KnowItAll [4], and others have successfully constructed semantic knowledge bases of large scale. Factual knowledge is typically represented in RDF, the W3C standard for Semantic-Web contents. RDF data can be seen as a graph whose nodes are entities These knowledge bases can be queried using the W3C-endorsed SPARQL [33] language. Yet, a knowledge base about entities can never be fully complete or always up to date. With the ANGIE system [23], we have shown that Web services can step in to fill this gap. Web services lend themselves to the extension of knowledge bases, because they deliver structured data. This eliminates the need for noisy information extraction techniques. Furthermore, there are Web services that offer a wide repertoire of data of good quality, well maintained and up to date. This makes Web services an interesting device for complementing knowledge bases. The ANGIE system incorporates Web services as follows: When a user asks a query, ANGIE tries to find the answer in the local knowledge base and resorts to Web services whenever the local knowledge base is not sufficient. ANGIE composes Web services and data from the local knowledge base on the fly, so that the user does not notice that some of the data was not present in the knowledge base before. For example, assume that the user asks for all songs by Canadian singers: