1"
DataXFormer:"Leveraging"the"Web" for"Seman7c"Transforma7ons" Ziawasch"Abedjan,"John"Morcos," Michael"Gubanov,"Ihab"F."Ilyas,"Mike" Stonebraker,"Paolo"PapoL,"Mourad" Ouzzani" 2"
Integra7on"of"mul7ple"sources" 3"
Different"value"representa7ons" Departure" BOS" Boston","MA"(BOS)" Boston"–"Logan"Interna7onal" Des7na7on" SFO" San"Fransisco,"CA"(SFO)" San"Fransisco,"CA"(SFO)" Cabin" Choice" Coach" Economy"Restricted" Time" 5:50"am"" 5:50"AM"" 14:26" Price" 563"$" 561"$" 613"€" Departure" Boston"(BOS)" Boston"(BOS)" Boston"(BOS)" Des7na7on" San"Fransisco"(SFO)" San"Fransisco"(SFO)" San"Fransisco"(SFO)" Economy" Cabin" Economy" Economy" 02:26p" Time" 5:50a" 5:50a" 4" …" 563"$" 561"$" 731"$"
Data"Transforma7on"Tasks" Airport"code" "! "City" Airport'code' City' • model" ! "brand" BOS" Boston" – Iphone"6" ! "Apple"Inc." JFK" New"York" • ISBN ! "7tle" ORD" Chicago" – 0f553f57340f3" ! "“A"Game"of" BER" Berlin" Thrones”" CDG" Paris" • unit"conversion" " – 1"Mi" ! "1.6"km" • date"format"transforma7ons" • long/lat" ! "loca7on" – MM/DD/YYYY" ! "DD/MM/YY"" • language"transla7on " • currency"conversion" • …" – 1"USD" ! "0.7?"EUR" 5"
Problem"Statement:"" Automa7cally"discover"transforma7ons!" Given' Find' airport City airport City BER ? BER Berlin JFK ? JFK New York ORD ? ORD Chicago HBE ? HBE Alexandria IST ? IST Istanbul FRA ? FRA Frankfurt BOS ? BOS Boston DFW ? DFW Dallas .. … .. … 6"
Syntac7c"Transforma7ons" Liter' Gallon' US'date' EU'date' 1" 0.26" 11/01/2014" 01.11.2014" 5" 1.04" 11/02/2014" 02.11.2014" 100" 26.42" 10/30/2014" 30.10.2014" 34" 8.98" 11/05/2014" 05.11.2014" 6" 1.58" 11/04/2014" 04.11.2014" GB' MB' Name' Last'name' 1" 1,024" Michael"Stonebraker" Stonebraker" 0.49" 500" Michael"Bay" Bay" 100" 102,400" Michael"Brodie" Brodie" 2" 2,048" Michael"Jordan" Jordan" 6" 6,144" 7"
Seman7c"Transforma7ons" ISBN' Title' Name' Nickname' 0f553f57340f3"" A"Game"of"Thrones" Michael" Mike" 0f553f80202fX" Universe"in"a"Nutshell" Samuel" Sam" 0f671f62964f6" The"Hitchhiker's"Guide" to"the"Galaxy" Ziawasch" Zia" 0f374f53355f7" Thinking"Fast"and"Slow" Rebeccea" Becca" 0f875f84585f1" The"Innovator’s" Dilemma" Airport'code' City' BOS" Boston" Model' Category' JFK" New"York" Iphone"6" Mobile"Phone" ORD" Chicago" MacBook"Air" Notebook" BER" Berlin" Logitech"mouse" Accessory" Nexus"5" Mobile"Phone" CDG" Paris" 8"
Problem"Statement"" Given' Find' airport airport City City airport City airport City BER BER ? Berlin BER Berlin BER Berlin Example" JFK JFK ? New York JFK New York JFK New York transforma7ons" ORD ORD ? Chicago ORD Chicago ORD Chicago HBE HBE ? ? HBE Alexandria HBE Alexandria IST IST ? ? IST Istanbul IST Istanbul FRA FRA ? ? FRA Frankfurt FRA Frankfurt BOS BOS ? ? BOS Boston BOS Boston DFW DFW ? ? DFW Dallas DFW Dallas .. .. … … .. … .. … 9"
DataXFormer:"The"Web"as"general" Repository" Given' Web'Tables' airport City BER Berlin JFK New York ORD Chicago HBE ? IST ? FRA ? BOS ? DFW ? .. … Web'Forms' 10"
Web"Tables" • Dataset" – Dresden"Web"table"Corpus" – 120"Million"tables" • Efficiently"discovering" transforma7on"Examples :" Filter"and"Refine"approach" – Filter"irrelevant"tables" Mul7ple"itera7ons"and" – Overcome"fragmenta7on" example"augmenta7on" • Average"rowcount"="12" Rate"transforma7ons"based"on" – Dirty"and"Heterogeneous"" example"hits"and"majority"vote" 11"
TransformaCon'task' Table"1" Table"3" airport City code location 1' 2' apc city … BER Berlin FRA Frankfurt DFW Dallas … JFK New York JFK New York HBE Alexandria … ORD Chicago LookF ORD Chicago Web"tables" IST Istanbul … Filter' HBE ? up' BOS Boston FRA Frankfurt … IST ? BER Berlin Table"4" …….." FRA ? Table"2" apc location BOS ? … … airport city DFW ? JFK New York … … FRA Frankfurt BER Berlin .. … … … DFW Dallas ORD Illinoise … … JFK New York FRA Frankfurt airport City FRA Hessen Augment' … … BER Berlin BOS Boston … … BER Berlin query' … .. … … DFW Dallas JFK New York airport City ORD Chicago Refine' 3' FRA Frankfurt BER Berlin BOS Boston JFK New York Augment'' X Y Score Lineage database' DFW Dallas ORD Chicago FRA Frankfurt 0.83 T1,T2 FRA Frankfurt BOS Boston 1 T1 BOS Boston DFW Dallas 0.67 T2 Result' DFW Dallas FRA Hessen 0.67 T4 HBE ? … … … … IST ? 4' 12" … …
Web"Forms" • How"to"find"Web"forms?" – Use"search"engine" • How"to"use"a"Web"form?" – Generate"a"wrapper" • How"to"avoid"high" response"7me?" – Cache"results"as"new"tables" 13"
Wrapping"Web"forms" • Parse"the"HTML"and" find"request" parameters" " • Locate"output"path" by"probing"with" examples" 14"
Expert"System"for"Corner"Cases" • Evaluate"transforma7ons" • Solve"conflicts" • Create"Transforma7ons" 15"
Experiments" • Collected"50"queries"from"computer"scien7sts" and"Tamr"customers" 1. Fahrenheit"to"Celsius"" 17. sentence"to"language"" 34. patent"ID"to"name"" 2. miles"to"km"" 18. text"to"encoding"" 35. city"to"long/lat"" 3. pound"to"kg" 19. Gregorian"to"Hijri"" 36. En7ty"to"wikipedia"link"" 4. USD"to"EUR" 20. CUSIP"to"company"" 37. En7ty"to"google"graph"id" 5. zip"to"state" 21. CUSIP"to"7cker"" 38. person"to"twiwer"id"" 6. zip"to"city" 22. symbol"to"company"" 39. ip"to"domain"" 7. UPS"tracking" 23. iban"to"bank"name"" 40. company"to"CEO"" to"address" 24. Loca7on"to"temperature" 41. company"to"industry"" 8. english"to"german"" 25. loca7on"to"humidity"" 42. US"standard"to"metric" 9. swiv"code"to"bank"" 26. car"plate"to"details"" 43. frac7ons"to"decimals"" 10. hex"to"RGB"" 27. country"code"to"country" 44. country"to"code"" 11. ISBN"to"publisher"" 28. ascii"to"char"" 45. State"to"state"abbrv"" 12. ISBN"to"7tle"" 29. car"model"to"brand"" 46. 7me"zone"to"abbrv"" 13. ISBN"to"author"" 30. country"to"demonym"" 47. city"to"country"" 14. ISSN"to"7tle"" 31. country"to"language"" 48. airport"code"to"city"" 15. ip"adress"to"country"" 32. country"to"currency"" 49. RGB"to"color" 16. Domain"to"primary"ip"" 33. company"to"BBGID"" 50. ASCII"to"unicode"" 16"
Recommend
More recommend