View
220
Download
0
Tags:
Embed Size (px)
Citation preview
[email protected]://www.mpi-inf.mpg.de/~weikum/
Gerhard Weikum
DB & IR: Both Sides Now
in collaboration with Georgiana Ifrim, Gjergji Kasneci, Josiane Parreira, Maya Ramanath, Ralf Schenkel, Fabian Suchanek, Martin Theobald
Gerhard Weikum June 14, 2007 2/41
DB and IR: Two Parallel Universes
canonical application: accounting libraries
data type: numbers,short strings
text
foundation: algebraic /logic based
probabilistic /statistics based
searchparadigm:
Boolean retrieval(exact queries,result sets/bags)
ranked retrieval(vague queries,result lists)
Database Systems Information Retrieval
parallel universes forever ?
marketleaders:
Oracle, IBM DB2,MS SQL Server, etc.
Google, Yahoo!, MSN,Verity, Fast, etc.
Gerhard Weikum June 14, 2007 3/41
Why DB&IR Now? – Application Needs
• Global health-care management for monitoring epidemics• News archives for journalists, press agencies, etc. • Product catalogs for houses, cars, vacation places, etc.• Customer support & CRM in insurances, telcom, retail, software, etc.• Bulletin boards for social communities• Enterprise search for projects, skills, know-how, etc.• Personalized & collaborative search in digital libraries, Web, etc.• Comprehensive archive of blogs with time-travel search
Simplify life for application areas like:
Typical data:Disease (DId, Name, Category, Pathogen …) UMLS-Categories ( … )Patient (… Age, HId, Date, Report, TreatedDId) Hospital (HId, Address …)Typical query: symptoms of tropical virus diseases and reported anomalieswith young patients in central Europe in the last two weeks
Gerhard Weikum June 14, 2007 4/41
Why DB&IR Now? – Platform Desiderata
Platform desiderata (from app developer‘s viewpoint):• Flexible ranking on text, categorical, numerical attributes
• cope with „too many answers“ and „no answers“
• High update rate concurrently with high query load
• Ontologies (dimensions, facets) for products, locations, org‘s, etc.• for query rewriting (relaxation, strengthening)
• Complex queries combining text & structured attributes• XPath/XQuery Full-Text with ranking
Structured data (records) Unstructured data (documents)
Unstructuredsearch(keywords)
Structuredsearch(SQL,XQuery)
DB Systems
IR SystemsSearch Engines
Keyword Search onRelational Graphs(IIT Bombay, UCSD, MSR, Hebrew U,CU Hong Kong, Duke U, ...)
Querying entities &relations from IE(MSR Beijing, UW Seattle,IBM Almaden, UIUC, MPI, … )
IntegratedDB&IR Platform
Gerhard Weikum June 14, 2007 5/41
Why DB&IR Forever?Turn the Web, Web2.0, and Web3.0 into the world‘s
most comprehensive knowledge base („semantic DB“) !
• Data enrichment at very large scale• Text and speech are key sources of knowledge production (publications, patents, conferences, meetings, ...)
2000 2007
indexed Web 2 Bio. 20 Bio.Flickr photos --- 100 Mio.digital photos ? 150 Bio. Wikipedia 8 000 1.8 Mio.OECD researchers 7.4 Mio. 8.4 Mio.patents world-wide ? 60 Mio.US Library of Congres 115 Mio. 134 Mio.Google Scholar --- 500 Mio.
Gerhard Weikum June 14, 2007 6/41
Outline
• Past
• Future
• Present
: Matter, Antimatter, and Wormholes
: From Data to Knowledge
: XML and Graph IR
Gerhard Weikum June 14, 2007 7/41
Parallel Universes: A Closer Look
Matter Antimatter
• user = programmer• query = precise spec. of info request• interaction via API
• user = your kids• query = approximation of user‘s real info needs• interaction process via GUI
• strength: indexing, QP• weakness: user model
• strength: ranking model• weakness: interoperability
• eval. measure: efficiency (throughput, response time, TPC-H, XMark, …)
• eval. measure: effectiveness (precision, recall, F1, MAP, NDCG, TREC & INEX benchmarks, …
[email protected]://www.mpi-inf.mpg.de/~weikum/
Gerhard Weikum
DB & IR: Both Sides Now
DBDBDBDB
IRIRIRIR
19901990 19951995 20002000 20052005
VAGUE(Motro)
Proximal Nodes(Baeza-Yates et al.)
Web EntitySearch:Libra, Avatar,ExDB …
Faceted Search:Flamenco …
1st Gen.XML IR:
XXL,XIRQL,Elixir,JuruXML
Multimedia IR
WHIRL(Cohen)
Web QueryLanguages:W3QS, WebOQL,Araneus …
Semistructured Data: Lore, Xyleme …
2nd Gen. XML IR:XRank,Timber, TIJAH,XSearch, FleXPath,CoXML, TopX,MarkLogic, Fast …
Prob. Datalog(Fuhr et al.)
Uncertain &Prob. Relations:Mystiq, Trio …
Struct. Docs
Deep Web Search
INEX
XPath
XPathFull-Text
Digital Libraries
GraphIR
Prob. DB(Cavallo&Pittarelli)
Prob. Tuples(Barbara et al.)
Gerhard Weikum June 14, 2007 9/41
WHIRL: IR over Relations [W.W. Cohen: SIGMOD’98]
Add text-similarity selection and join to relational algebraExample: Select * From Movies M, Reviews R Where M.Plot ~ ”fight“ And M.Year > 1990 And R.Rating > 3 And M.Title ~ R.Title And M.Plot ~ R.Comment
Title Plot … Year
Movies
Title Comment … Rating
Reviews
Matrix
Hero
Matrix 1
MatrixReloaded
MatrixEigenvalues
Ying xiongaka. Hero
Shrek 2
… matrix spectrum … orthonormal …
… fight for peace …… sword fight … dramatic colors …
…
In ancient China … fights … sword fight …fights Broken Sword …
In the near future … computer hacker Neo …… fight training …
… cool fights …new techniques …
… fights …and more fights …… fairly boring …
1999
2002
2004In Far Far Away …our lovely herofights with cat killer …
4
1
5
5
Scoring and ranking:
s (<x,y>, q: A~B) = cosine (x.A, y.B)
s (<x,y>, q1 … qm) =
m
iiqyxs
1
),,(
xj ~ tf (word j in x) idf (word j)with dampening & normalization
• DB&IR for query-time data integration• More recent work: MinorThird, Spider, DBLife, etc.• But scoring models fairly ad hoc
Gerhard Weikum June 14, 2007 10/41
Professor
Name:GerhardWeikum
Address...
City: SBCountry: Germany
TeachingResearch
Course
Title: IR
Description: Information retrieval ...
Syllabus
...Book Article
... ...
ProjectTitle: IntelligentSearch ofHeterogeneousXML Data
Funding: EU
...
Name:RalfSchenkel
Lecturer
Address:Max-PlanckInstitute forInformatics,Germany
Activities
Seminar
Contents:Ranked retrieval …
Literature: …
Scientific
Name:INEX taskcoordinator(Initiative for the Evaluation of XML …)
Other
Sponsor: EU
…
XXL: Early XML IR [Anja Theobald, GW: Adding Relevance toXML, WebDB’00]
Which professors from Saarbruecken (SB)are teaching IR and haveresearch projects on XML?
Union of heterogeneous sources without global schema Similarity-aware XPath://~Professor [//* = ”~SB“] [//~Course [//* = ”~IR“] ] [//~Research [//* = ”~XML“] ]
Similarity-aware XPath://~Professor [//* = ”~SB“] [//~Course [//* = ”~IR“] ] [//~Research [//* = ”~XML“] ]
Gerhard Weikum June 14, 2007 11/41
Professor
Name:GerhardWeikum
Address...
City: SBCountry: Germany
TeachingResearch
Course
Title: IR
Description: Information retrieval ...
Syllabus
...Book Article
... ...
ProjectTitle: IntelligentSearch ofHeterogeneousXML Data
Funding: EU
...
Name:RalfSchenkel
Lecturer
Address:Max-PlanckInstitute forInformatics,Germany
Activities
Seminar
Contents:Ranked retrieval …
Literature: …
Scientific
Name:INEX taskcoordinator(Initiative for the Evaluation of XML …)
Other
Sponsor: EU
…
XXL: Early XML IR [Anja Theobald, GW: Adding Relevance toXML, WebDB’00]
Scoring and ranking:• tf*idf for content condition• ontological similarity for relaxed tag condition• score aggregation with probabilistic independence
Wu&Palmer: |path| through lca(x,y)
Dice coeff.: 2 #(x,y) / (#x + #y) on Web
Similarity-aware XPath://~Professor [//* = ”~Saarbruecken“] [//~Course [//* = ”~IR“] ] [//~Research [//* = ”~XML“] ]
Which professors from Saarbruecken (SB)are teaching IR and haveresearch projects on XML?
Motivation: Union of heterogeneous sources has no schema
query expansion model:disjunction of tags
magician
wizard
intellectual
artist
alchemist
directorprimadonna
professor
teacher
scholar
academic,academician,faculty member
scientist
researcher
HYPONYM (0.749)HYPONYM (0.749)
investigator
mentor
RELATED (0.48)RELATED (0.48)
lecturer
Gerhard Weikum June 14, 2007 12/41
The Past: Lessons Learned
• DB&IR: added flexible ranking to (semi) structured querying to cope with schema and instance diversity
• but ranking seems „ad hoc“ and not consistently good in benchmarks
• to win benchmark: tuning needed, but tuning is easier if ranking is principled !
• ontologies are mixed blessing: quality diverse, concept similarity subtle, danger of topic drift
• ontology-based query expansion (into large disjunctions) poses efficiency challenge
prec
isio
n
recall
// ~Professor [...]
// { Professor, Researcher, Lecturer, Scientist, Scholar, Academic, ... }[...]
element
gold
produce
Golden Delicious
entity
food
substancesolid
edible fruit
applepome
Gerhard Weikum June 14, 2007 13/41
Outline
Past
• Future
• Present
: Matter, Antimatter, and Wormholes
: From Data to Knowledge
: XML and Graph IR
Gerhard Weikum June 14, 2007 14/41
TopX: 2nd Generation XML IR
”Semantic“ XPath Full-Text query: /Article [ftcontains(//Person, ”Max Planck“)] [ftcontains(//Work, ”quantum physics“)]//Children[@Gender = ”female“]//Birthdates
supported by TopX engine: http://infao5501.ag5.mpi-sb.mpg.de:8080/topx/ http://topx.sourceforge.net
• Exploit tags & structure for better precision• Can relax tag names & structure for better recall• Principled ranking by probabilistic IR (Okapi BM25 for XML)• Efficient top-k query processing (using improved TA)• Robust ontology integration (self-throttling to avoid topic drift)• Efficient query expansion (on demand, by extended TA)• Relevance feedback for automatic query rewriting
[Martin Theobald, Ralf Schenkel, GW: VLDB’05, VLDB Journal]
Gerhard Weikum June 14, 2007 15/41
Commercial Break[Martin Theobald, Ralf Schenkel, GW: VLDB’95]
TopX demo today 3:30 – 5:30
Gerhard Weikum June 14, 2007 16/41
Principled Ranking by Probabilistic IR
]|[
]|[
]|)([
]|)([),(
dRP
dRP
dofcontentsqRdP
dofcontentsqRdPqds
odds for item d withterms di being relevant for query q = {q1, …, qm}
binary features, conditional independence of features [Robertson & Sparck-Jones 1976]
mi RdP
RdP
i
i1 ]|[
]|[~
dqii
i
i
i
q
q
p
p 1log
1log~ ]|[ RdPp ii
]|[ RdPq ii
Now estimate pi and qi values from •relevance feedback,•pseudo-relevance feedback, •corpus statistics
by MLE (with statistical smoothing)and store precomputed pi, qi in index
docsdocsrelpi /#).(#ˆ
]|[ corpusdPq ii
ki kdf
idfq
)(
)(ˆ
ki dktf
ditfp
),(
),(ˆ
i
k
kidf
kdf
dk
ditf
)(
)(
),(
),(log
Relationship to tf*idf
„God does not play dice.“ (Einstein)IR does.
with
related to but different fromstatistical language models
• led to Okapi BM25 (wins TREC tasks)• adapted and extended to XML in TopX, ...
Gerhard Weikum June 14, 2007 17/41
Probabilistic Ranking for SQL
SQL queries that return many answers need rankingExamples: • Houses (Id, City, Price, #Rooms, View, Pool, SchoolDistrict, …) Select * From Houses Where View = ”Lake“ And City In (”Redmond“, ”Bellevue“)• Movies (Id, Title, Genre, Country, Era, Format, Director, Actor1, Actor2, …) Select * From Movies Where Genre = ”Romance“ And Era = ”90s“
]|[
]|[
]|[
]|[~),(
RdP
RdP
dRP
dRPqds
odds for tuple d withattributes XY relevant for query q: X1=x1 … Xm=xm
]|[
]|[
RXYP
RXYP
][]|[
]|[1
YPRYP
YXP
Estimate prob‘s, exploiting workload W:
[S. Chaudhuri, G. Das, V. Hristidis, GW: TODS‘06]
]|[]|[ XWYPRYP Example: frequent queries
• … Where Genre = ”Romance“ And Actor1 = ”Hugh Grant“• … Where Actor1 = ”Hugh Grant“ And Actor2 = ”Julia Roberts“boosts HG and JR movies in ranking for Genre = ”Romance“ And Era = ”90s“
Gerhard Weikum June 14, 2007 18/41
From Tables and Trees to Graphs
Example: Conferences (CId, Title, Location, Year) Journals (JId, Title)CPublications (PId, Title, CId) JPublications (PId, Title, Vol, No, Year) Authors (PId, Person) Editors (CId, Person)Select * From * Where * Contains ”Gray, DeWitt, XML, Performance“ And Year > 95
Schema-agnostic keyword search over multiple tables:graph of tuples with foreign-key relationships as edges
[BANKS, Discover, DBExplorer, KUPS, SphereSearch, BLINKS]
Result is connected tree with nodes that contain as many query keywords as possible
Ranking: 1)(1)1(),(),(
eedgesnnodes
eedgeScoreqnnodeScoreqtrees
with nodeScore based on tf*idf or prob. IRand edgeScore reflecting importance of relationships (or confidence, authority, etc.)
Related use cases:• XML beyond trees• RDF graphs• ER graphs (e.g. from IE)• social networks
Top-k querying: compute best trees, e.g. Steiner trees (NP-hard)
Gerhard Weikum June 14, 2007 19/41
The Present: Observations & Opportunities• Probabilistic IR and statistical language models
yield principled ranking and high effectiveness (related to prob. relational models (Suciu, Getoor, …) but different)
• Structural similarity and ranking based on tree edit distance (FleXPath, Timber, …)
• Aim for comprehensive XML ranking model capturing content, structure, ontologies
• Aim to generate structure skeleton in XPath query from user feedback
• Good progress on performance but still many open efficiency issues
actor
movie movie
plot directormovie
actor actor director
plot
”life physicist Max Planck“
//article[//person ”Max Planck“] [//category ”physicist“] //biography
Gerhard Weikum June 14, 2007 20/41
Outline
Past
• Future
Present
: Matter, Antimatter, and Wormholes
: From Data to Knowledge
: XML and Graph IR
Gerhard Weikum June 14, 2007 21/41
Knowledge Queries
Nobel laureate who survived both world wars and his children
drama with three women making a prophecy to a British nobleman that he will become king
proteins that inhibit both protease and some other enzyme
connection between Thomas Mann and Goethe
differences in Rembetiko music from Greece and from Turkey
neutron stars with Xray bursts > 1040 erg s-1 & black holes in 10‘‘
market impact of Web2.0 technology in December 2006
sympathy or antipathy for Germany from May to August 2006
Turn the Web, Web2.0, and Web3.0 into the world‘s most comprehensive knowledge base („semantic DB“) !
Answer „knowledge queries“ such as:
Gerhard Weikum June 14, 2007 22/41
Three Roads to Knowledge
• Handcrafted High-Quality Knowledge Bases (Semantic-Web-style ontologies, encyclopedias, etc.)
• Large-scale Information Extraction & Harvesting: (using pattern matching, NLP, statistical learning, etc. for product search, Web entity/object search, ...)
• Social Wisdom from Web 2.0 Communities (social tagging, folksonomies, human computing, e.g.: del.icio.us, flickr, answers.yahoo, iknow.baidu, ...)
• Social Wisdom from Web 2.0 Communities (social tagging, folksonomies, human computing, e.g.: del.icio.us, flickr, answers.yahoo, iknow.baidu, ...)
Gerhard Weikum June 14, 2007 23/41
High-Quality Knowledge Sources• universal „common-sense“ ontologies:
• SUMO (Suggested Upper Merged Ontology): 60 000 OWL axioms• Cyc: 5 Mio. facts (OpenCyc: 2 Mio. facts)
• domain-specific ontologies:• UMLS (Unified Medical Language System): 1 Mio. biomedical concepts 135 categories, 54 relations (e.g. virus causes disease | symptom)• GeneOntology, etc.
• thesauri and concept networks:• WordNet: 200 000 concepts (word senses) and hypernym/hyponym relations• can be cast into OWL-lite (or typed graph with statistical weights)
• lexical sources:• Wikipedia (1.8 Mio. articles, 40 Mio. links, 100 languages) etc.
• hand-tagged natural-language corpora:• TEI (Text Encoding Initiative) markup of historic encyclopedia• FrameNet: sentences classified into frames with semantic roles
growing with strong momentum
Gerhard Weikum June 14, 2007 24/41
High-Quality Knowledge SourcesGeneral-purpose thesauri and concept networks: WordNet family
enzyme -- (any of several complex proteins that are produced by cells and act as catalysts in specific biochemical reactions) => protein -- (any of a large group of nitrogenous organic compounds that are essential constituents of living cells; ...) => macromolecule, supermolecule ... => organic compound -- (any compound of carbon and another element or a radical)... => catalyst, accelerator -- ((chemistry) a substance that initiates or accelerates a chemical reaction without itself being affected) => activator -- ((biology) any agency bringing about activation; ...)
can be cast into • OWL-lite or into • graph, with weights for relation strengths (derived from co-occurrence statistics)
Gerhard Weikum June 14, 2007 25/41
High-Quality Knowledge SourcesWikipedia and other lexical sources
Gerhard Weikum June 14, 2007 26/41
{{Infobox_Scientist| name = Max Planck| birth_date = [[April 23]], [[1858]] | birth_place = [[Kiel]], [[Germany]]| death_date = [[October 4]], [[1947]]| death_place = [[Göttingen]], [[Germany]]| residence = [[Germany]] | nationality = [[Germany|German]] | field = [[Physicist]]| work_institution = [[University of Kiel]]</br> [[Humboldt-Universität zu Berlin]]</br> [[Georg-August-Universität Göttingen]]| alma_mater = [[Ludwig-Maximilians-Universität München]]| doctoral_advisor = [[Philipp von Jolly]]| doctoral_students = [[Gustav Ludwig Hertz]]</br>… | known_for = [[Planck's constant]], [[Quantum mechanics|quantum theory]]| prizes = [[Nobel Prize in Physics]] (1918)…
Exploit Hand-Crafted KnowledgeWikipedia, WordNet, and other lexical sources
Gerhard Weikum June 14, 2007 27/41
YAGO: Yet Another Great Ontology[F. Suchanek, G. Kasneci, GW: WWW 2007]
• Turn Wikipedia into explicit knowledge base (semantic DB)
• Exploit hand-crafted categories and templates
• Represent facts as explicit knowledge triples:
relation (entity1, entity2)
(in 1st-order logic, compatible with RDF, OWL-lite, XML, etc.)
• Map (and disambiguate) relations into WordNet concept DAG
entity1 entity2relation
Max_Planck KielbornIn
Kiel CityisInstanceOf
Examples:
Gerhard Weikum June 14, 2007 28/41
YAGO Knowledge RepresentationEntity
Max_Planck April 23, 1858
Person
City Country
subclass Location
subclass
instanceOf
subclass subclass
bornOn
“Max Planck”
means
“Dr. Planck”
means
subclass
October 4, 1947 diedOn
KielbornInNobel Prize Erwin_Planck
FatherOfhasWon
Scientist
means
“Max Karl Ernst Ludwig Planck”
Physicist
instanceOf
subclassBiologist
subclass
concepts
individuals
words
Knowledge Base # Facts
KnowItAll 30 000SUMO 60 000WordNet 200 000OpenCyc 300 000Cyc 5 000 000YAGO 6 000 000
Online access and download at http://www.mpi-inf.mpg.de/~suchanek/yago/
Accuracy: 97%
Gerhard Weikum June 14, 2007 29/41
NAGA: Graph IR on YAGO [G. Kasneci et al.: WWW‘07]
queries with regular expressions
Ling $x scientistisa hasFirstName | hasLastName
$y ZhejianglocatedIn*
worksFor
conjunctive queries
Beng Chin Ooi
(coAuthor| advisor)*
Kiel $x scientistisa bornIn
Graph-based search on YAGO-style knowledge bases with built-in ranking based on confidence and informativeness
statistical language model for result graphs
Gerhard Weikum June 14, 2007 30/41
Ranking FactorsConfidence:Prefer results that are likely to be correct
Certainty of IE Authenticity and Authority of Sources
Informativeness:Prefer results that are likely importantMay prefer results that are likely new to user
Frequency in answer Frequency in corpus (e.g. Web) Frequency in query log
Compactness:Prefer results that are tightly connected
Size of answer graph
bornIn (Max Planck, Kiel) from„Max Planck was born in Kiel“(Wikipedia)
livesIn (Elvis Presley, Mars) from„They believe Elvis hides on Mars“(Martian Bloggeria)
q: isa (Einstein, $y)
isa (Einstein, scientist)isa (Einstein, vegetarian)
q: isa ($x, vegetarian)
isa (Einstein, vegetarian)isa (Al Nobody, vegetarian)
Einstein
vegetarian
BohrNobel Prize
Tom Cruise
1962
isa isa bornIn
diedInwon
won
Gerhard Weikum June 14, 2007 31/41
Information Extraction (IE): Text to Records
Max Planck 4/23, 1858 KielAlbert Einstein 3/14, 1879 Ulm Mahatma Gandhi 10/2, 1869 Porbandar
Person BirthDate BirthPlace ...
Person ScientificResult
Max Planck Quantum Theory
Person CollaboratorMax Planck Albert EinsteinMax Planck Niels Bohr
Planck‘s constant 6.2261023 Js
Constant Value Dimension
combine NLP, pattern matching, lexicons, statistical learning
Gerhard Weikum June 14, 2007 32/41
Knowledge Acquisition from the WebLearn Semantic Relations from Entire Corpora at Large Scale(as exhaustively as possible but with high accuracy)
Examples: • all cities, all basketball players, all composers• headquarters of companies, CEOs of companies, synonyms of proteins• birthdates of people, capitals of countries, rivers in cities• which musician plays which instruments• who discovered or invented what• which enzyme catalyzes which biochemical reaction
Existing approaches and tools (Snowball [Gravano et al. 2000], KnowItAll [Etzioni et al. 2004], …):
almost-unsupervised pattern matching and learning:seeds (known facts) patterns (in text) (extraction) rule (new) facts
Gerhard Weikum June 14, 2007 33/41
city(Beijing) plays(Coltrane, sax) city(Beijing) old center of Beijingplays(Coltrane, sax) sax player Coltranecity(Beijing) old center of Beijing old center of Xplays(Coltrane, sax) sax player Coltrane Y player X
Methods for Web-Scale Fact Extration
Example:city (Seattle) in downtown Seattle city (Seattle) Seattle and other towns city (Las Vegas) Las Vegas and other townsplays (Zappa, guitar) playing guitar: … Zappaplays (Davis, trumpet) Davis … blows trumpet
seeds text rules new facts
Example:city (Seattle) in downtown Seattle in downtown Xcity (Seattle) Seattle and other towns X and other townscity (Las Vegas) Las Vegas and other towns X and other townsplays (Zappa, guitar) playing guitar: … Zappa playing Y: … Xplays (Davis, trumpet) Davis … blows trumpet X … blows Y
Example:city (Seattle) in downtown Seattle in downtown Xcity (Seattle) Seattle and other towns X and other townscity (Las Vegas) Las Vegas and other towns X and other townsplays (Zappa, guitar) playing guitar: … Zappa playing Y: … Xplays (Davis, trumpet) Davis … blows trumpet X … blows Y
Example:city (Seattle) in downtown Seattle in downtown Xcity (Seattle) Seattle and other towns X and other townscity (Las Vegas) Las Vegas and other towns X and other townsplays (Zappa, guitar) playing guitar: … Zappa playing Y: … Xplays (Davis, trumpet) Davis … blows trumpet X … blows Y
in downtown Beijing city(Beijing) Coltrane blows sax plays(C., sax)
Assessment of facts & generation of rules based on statisticsRules can be more sophisticated: playing NN: (ADJ|ADV)* NP & class(NN)=instrument & class(head(NP))=person plays(head(NP), NN)
Gerhard Weikum June 14, 2007 34/41
Performance of Web-IEState-of-the-art precision/recall results:
Anecdotic evidence:invented (A.G. Bell, telephone)married (Hillary Clinton, Bill Clinton)isa (yoga, relaxation technique)isa (zearalenone, mycotoxin)contains (chocolate, theobromine)contains (Singapore sling, gin)
invented (Johannes Kepler, logarithm tables)married (Segolene Royal, Francois Hollande)isa (yoga, excellent way)isa (your day, good one)contains (chocolate, raisins)plays (the liver, central role)makes (everybody, mistakes)
relation precision recall corpus systemscountries 80% 90% Web KnowItAllcities 80% ??? Web KnowItAllscientists 60% ??? Web KnowItAllheadquarters 90% 50% News Snowball, LEILAbirthdates 80% 70% Wikipedia LEILAinstanceOf 40% 20% Web Text2Onto, LEILA
Open IE 80% ??? Web TextRunner
precision value-chain: entities 80%, attributes 70%, facts 60%, events 50%
Gerhard Weikum June 14, 2007 35/41
Beyond Surface Learning with LEILA
Almost-unsupervised Statistical Learning with Dependency Parsing
(Cologne, Rhine), (Cairo, Nile), … (Cairo, Rhine), (Rome, 0911), (, [0..9]*), …
Paris was founded on an island in the Seine
(Paris, Seine)
Ss Pv MVp Ds
Js
DG
Js
MVp
NP VPVP PP NP NPPP NPNP
Cologne lies on the banks of the Rhine
Ss MVp DMc Mp Dg
JsJp
NP PPVP NP PP NP NPNP
People in Cairo like wine from the Rhine valley
Mp Js Os
Sp Mvp DsJs
AN
NP NPPP VP PP NPNP NPNP
Limitation of surface patterns:who discovered or invented what “Tesla’s work formed the basis of AC electric power”
Learning to Extract Information by Linguistic Analysis [F.Suchanek, G.Ifrim, GW: KDD‘06]
LEILA outperforms other Web-IE methodsin terms of precision, recall, F1, but:• dependency parser is slow• one relation at a time
“Al Gore funded more work for a better basis of the Internet”
Gerhard Weikum June 14, 2007 36/41
IE Efficiency and Accuracy Tradeoffs
• precision vs. recall: two-stage processing (filter pipeline)1) recall-oriented harvesting2) precision-oriented scrutinizing
• preprocessing• indexing: NLP trees & graphs, N-grams, PoS-tag patterns ?
• exploit ontologies? exploit usage logs ?• turn crawl&extract into set-oriented query processing
• candidate finding• efficient phrase, pattern, and proximity queries• optimizing entire text-mining workflows [Ipeirotis et al.: SIGMOD‘06]
IE is cool, but what‘s in it for DB folks?
[see also tutorials by Cohen, Doan/Ramakrishnan/Vaithyanathan, Agichtein/Sarawagi]
Gerhard Weikum June 14, 2007 37/41
The Future: Challenges• Generalize YAGO approach (Wikipedia + WordNet)• Methods for comprehensive, highly accurate
mappings across many knowledge sources• cross-lingual, cross-temporal• scalable in size, diversity, number of sources
• Pursue DB support towards efficient IE (and NLP)• Achieve Web-scale IE throughput that can
• sustain rate of new content production (e.g. blogs) • with > 90% accuracy and Wikipedia-like coverage
• Integrate handcrafted knowledge with NLP/ML-based IE• Incorporate social tagging and human computing
Gerhard Weikum June 14, 2007 38/41
Outline
Past
Future
Present
: Matter, Antimatter, and Wormholes
: From Data to Knowledge
: XML and Graph IR
Gerhard Weikum June 14, 2007 39/41
Major Trends in DB and IR
malleable schema (later) deep NLP, adding structure
record linkage info extraction
graph mining entity-relationship graph IR
ontologies
ranking
Database Systems Information Retrieval
statistical language models
data uncertainty
programmability search as Web Service
dataspaces Web objects
Web 2.0 Web 2.0
Gerhard Weikum June 14, 2007 40/41
Conclusion• DB&IR integration agenda:
• models − ranking, ontologies, prob. SQL ?, graph IR ?• languages and APIs − XQuery Full-Text++ ?• systems − drop SQL, go light-weight ? − combine with P2P, Deep Web, ... ?
• Rethink progress measures and experimental methodology
• Address killer app(s) and grand challenge(s):• from data to knowledge (Web, products, enterprises)• integrate knowledge bases, info extraction, social wisdom• cope with uncertainty; ranking as first-class principle
• Bridge cultural differences between DB and IR:• co-locate SIGIR and SIGMOD
Gerhard Weikum June 14, 2007 41/41
DB&IR: Both Sides NowJoni Mitchell (1969): Both Sides Now
…I've looked at life from both sides now,From up and down, and still somehowIt's life's illusions i recall.I really don't know life at all.
Thank You !