Gong show Stefano Ceri Politecnico di Milano Higher

Gong show Stefano Ceri Politecnico di Milano

Higher order data mining • Given a database, give me all the results to the interesting queries – Needs to find “(almost) all the queries” (from the schema) and to sort out those which have an “interesting result”. – Much more general – and useful – than association rules et. al. (narrow focus so far? ? ? ) • Example: Lowell’s participants list • (only one from Asia) • (only two women) • (everyone’s age is above 40 and below 60(, except …. )) • Example: serves(bar, beer), likes(person, beer), visits(person, bar) [Bart Goethals] • • (Bob likes all the beers) (Alfredo’s is the most popular bar serving all beers) (people who like Guinness only go to bars serving Guinness) (Carol is the only person going to a bar that does not serve a beer she likes)

Higher order search engines • Current search engines looks for “terms” (“attributes”) in a page (“entity”) – Fail to find “ethnical restaurants in a nice place outside Milano” • Higher-order search: more expressive model, and simple ways to get instructions from users – “domains” (e. g. : city, time, artist, politician, meal, wine. . ) – “relationships” between pages (e. g. Page linked-to/linked-by Page) • Example: Find ethnical restaurants close to Milano: (city: lombardia) (ethnical restaurant) pavia 30 entries como 20 entries (outside milano) linked by (ethnical restaurant) certosa di pavia. . . . Ethiopian restaurant monza park. . . Egyptian restaurant