On the Semantic Web, Completeness does Matter!
0 likes448 views
The document discusses completeness statements about RDF data sources and how they can be used for query answering. It provides examples of completeness statements that capture whether a data source is complete for all actors in a movie, all presidents of a university up to a certain date. It also shows how queries can be analyzed using such completeness statements to determine whether the answers will be fully complete or might be missing some results due to limitations in the data source.
On the Semantic Web, Completeness does Matter!
- 1. On the Semantic Web, completeness does matter! Fariz Darari (fariz.darari@stud-inf.unibz.it) PhD Supervisor: Prof. Werner Nutt … but they are not machine-readable nor formal :( … especially since it employs OWA query is correct? complete query up to when? query is empty due to incompleteness or non-existence? OK let’s have them in that way, then what? Can we check .. Yes, we can check whether a query with negation is correct if the negation part is complete. By inspecting the timestamps of the statements, we can check up to what date query answers are complete. Yes, we can check whether a query returns an empty answer due to non-existence when some no-value statement (completeness statement capturing non-existence) touches query parts. Can we do more, perhaps? Statistical completeness reasoning & completeness reasoning over RDF streams On the Web we already have completeness statements …………………………. Complete for all actors of Reservoir Dogs Compl(resDogs actor ?a) Give me actors played in Armageddon but not in Reservoir Dogs SELECT ?a { armageddon actor ?a NOT EXISTS {resDogs actor ?a} } Analysis The answer of Q is always correct. Suppose a query answer is not correct. Then, it must be an Armageddon actor that in reality was also a Reservoir Dogs actor, but our data source does not know if she was in Reservoir Dogs. However, by our statement, our data source is complete for Reservoir Dogs actors, hence a contradiction! Give me people who have ever been a president in both Union College and MIT SELECT ?p { uCollege everPresident ?p . mit everPresident ?p } Complete for all presidents of Union College up to 2010 Compl(uCollege everPresident ?p, 2010) Complete for all presidents of MIT up to 2014 Compl(mit everPresident ?p, 2014) Analysis Query answers will be complete up to 2010 since the presidents of Union College are complete up to that time and even further for those of MIT. However, from 2011 onwards, the query completeness cannot be guaranteed as we might be missing some Union College presidents after 2010. Elizabeth I has no child No(elizabethI child ?c) Give me all spouses of Elizabeth I’s children SELECT ?c ?s { elizabethI child ?c . ?c spouse ?s } Analysis Q will return an empty answer because of non-existence of the information (i.e., some query part returns nothing, hence the whole query also returns nothing) that is asked, not by the incompleteness of the data source. Fariz Darari, Werner Nutt, Giuseppe Pirrò, Simon Razniewski: Completeness Statements about RDF Data Sources and Their Use for Query Answering. ISWC 2013. Fariz Darari, Simon Razniewski, Werner Nutt: Bridging the Semantic Gap between RDF and SPARQL using Completeness Statements. ISWC (Poster) 2014. For more info …