anhai doan university of wisconsin-madison the cimple project on community information management
TRANSCRIPT
AnHai DoanUniversity of Wisconsin-Madison
The Cimple Project on The Cimple Project on Community Information Management Community Information Management
2
The CIM ProblemThe CIM Problem Numerous online communities
– database researchers, movie fans, legal professionals, bioinformatics, enterprise intranets, tech support groups
Each community = many data sources + many members Database community
– home pages, project pages, DBworld, DBLP, conference pages, ...
Movie fan community– review sites, movie home pages, theatre listings, ...
Legal profession community– law firm home pages
3
The CIM ProblemThe CIM Problem Members often want to discovery, query, monitor
information in the community
Database community– what is new in the past week in the database community?– any interesting connection between researchers X and Y?– find all citations of this paper in the past one week on the Web– what are current hot topics? who has moved where?
Legal profession community– which lawyers have moved where? – which law firms have taken on which cases?
4
The CIM ProblemThe CIM Problem To address such needs, build data portals Starting out topic-based, now structured data portals
– DBLP, Citeseer, IMDB, GlobalSpec, etc.
Limitations of current solutions– mostly by hand, labor intensive, error prone
– hard-to-port solutions
– few services other than browsing and keyword search
5
Cimple Project @ Wisconsin / Yahoo! ResearchCimple Project @ Wisconsin / Yahoo! Research
Researcher
Homepages
Conference
Pages
Group Pages
DBworld
mailing list
DBLP
Web pages
Text documents
* **
** * ***
SIGMOD-04
**
** give-talk
Jim Gray
Keyword search
SQL querying
Question answering
Browse
Mining
Alert/Monitor
News summary
Jim Gray
SIGMOD-04
**
Personalize system, provide feedback
Develop generic solutions to create structured data portals via extraction + integration + mass collaboration
6
The Research TeamThe Research Team
Faculty / Vice President– AnHai Doan– Raghu Ramakrishnan
Current students– Pedro DeRose– Warren Shen– Fei Chen– Yoonkyong Lee– Doug Burdick– Mayssam Sayyadian – Xiaoyong Chai – Ting Chen
7
Prototype System: DBLifePrototype System: DBLife Integrate data of the DB research community 1164 data sources
Crawled daily, 11000+ pages = 160+ MB / day
10
Resulting ER GraphResulting ER Graph
“Proactive Re-optimization
Jennifer Widom
Shivnath Babu
SIGMOD 2005
David DeWitt
Pedro Bizarrocoauthor
coauthor
coauthor
advise advise
write
write
write
PC-Chair
PC-member
11
Querying The ER GraphQuerying The ER Graph
Query: “David DeWitt Jennifer Widom”
1.
2.
3.
Jennifer Widom
David DeWittcoauthor
Jennifer Widom
SIGMOD 2005
David DeWittcoauthor
PC-Chair
PC-member
Jennifer Widom
Shivnath Babu
David DeWitt
coauthor
coauthoradvise
13
Mass Collaboration: Example 1Mass Collaboration: Example 1
Picture is removed if enough users vote “no”.
14
Mass Collaboration Meets Jeff NaughtonMass Collaboration Meets Jeff Naughton
Jeffrey F. Naughton swears that this is David J. DeWitt
15
Mass Collaboration: Example 2Mass Collaboration: Example 2
Community Wikipedia
backed up by a structured underlying database
16
What We Have DoneWhat We Have Done
Define the CIM problem / understand it a little bit– start to talk about it in the DB community
[SIGMOD-06 tutorial, IEEE DEB-06, CIDR-07]
Build DBLife / helps clarify research issues– live at dblife.cs.wisc.edu– latest stuff at dblife-labs.cs.wisc.edu
Start some preliminary research– ICDE-07a, ICDE-07b, ICDE-07b
17
What We Would Like to Do NextWhat We Would Like to Do Next Release DBLife
– as a research / education toolpossible service to the DB community demo of CIM systems benchmark / challenge for data integration / extraction
Develop and release a generic Cimple platform– anyone can use it to build structured data portals
Build CimBase: a hosting service– anyone can specify a structured portal on CimBase– we will build and host it
Continue research / expand team / build alliance
18
Research Challenges (1)Research Challenges (1)
Information extraction Data integration Mass collaboration
Researcher
Homepages
Conference
Pages
Group Pages
DBworld
mailing list
DBLP
Web pages
Text documents
* **
** * ***
SIGMOD-04
**
** give-talk
Jim Gray
Keyword search
SQL querying
Question answering
Browse
Mining
Alert/Monitor
News summary
Jim Gray
SIGMOD-04
**
Personalize system, provide feedback
19
Research Challenges (2)Research Challenges (2)
Researcher
Homepages
Conference
Pages
Group Pages
DBworld
mailing list
DBLP
Web pages
Text documents
* **
** * ***
SIGMOD-04
**
** give-talk
Jim Gray
Keyword search
SQL querying
Question answering
Browse
Mining
Alert/Monitor
News summary
Jim Gray
SIGMOD-04
**
Personalize system, provide feedback
Exploiting extracted data Handling uncertainty / provenance / explanation Dealing with evolving data, versioning, temporal data
20
Research Challenges (3)Research Challenges (3)
Researcher
Homepages
Conference
Pages
Group Pages
DBworld
mailing list
DBLP
Web pages
Text documents
* **
** * ***
SIGMOD-04
**
** give-talk
Jim Gray
Keyword search
SQL querying
Question answering
Browse
Mining
Alert/Monitor
News summary
Jim Gray
SIGMOD-04
**
Personalize system, provide feedback
What is the right architecture? What is the right data model / storage? How to build continuously running systems How to build massively scalable hosting services? How to build a generic CIM platform?
21
Rest of the TalkRest of the Talk
The CIM problem The Cimple solution approach What we have done / plan to do Research challenges
– information extraction– data integration (focus on entity matching)– mass collaboration
Broader perspectives
22
Declarative IEDeclarative IE Current IE research
– develops learning- & rule-based solutions [SIGMOD-06 tutorial]– focuses largely on improving accuracy
Real-world IE applications– glue multiple such solutions together, using Perl
Serious problems– hard to develop, understand, debug, and optimize
DECLARATIVE IE
Dr. R. Ramakrishnan
This is a fun topic ...
23
Example in DBLifeExample in DBLife
Find conference name in raw text#############################################################################
# Regular expressions to construct the pattern to extract conference names#############################################################################
# These are subordinate patternsmy $wordOrdinals="(?:first|second|third|fourth|fifth|sixth|seventh|eighth|ninth|tenth|eleventh|twelfth|thirteenth|fourteenth|fifteenth)";
my $numberOrdinals="(?:\\d?(?:1st|2nd|3rd|1th|2th|3th|4th|5th|6th|7th|8th|9th|0th))";my $ordinals="(?:$wordOrdinals|$numberOrdinals)";
my $confTypes="(?:Conference|Workshop|Symposium)";my $words="(?:[A-Z]\\w+\\s*)"; # A word starting with a capital letter and ending with 0 or more spaces
my $confDescriptors="(?:international\\s+|[A-Z]+\\s+)"; # .e.g "International Conference ...' or the conference name for workshops (e.g. "VLDB Workshop ...")
my $connectors="(?:on|of)";my $abbreviations="(?:\\([A-Z]\\w\\w+[\\W\\s]*?(?:\\d\\d+)?\\))"; # Conference abbreviations like "(SIGMOD'06)"
# The actual pattern we search for. A typical conference name this pattern will find is# "3rd International Conference on Blah Blah Blah (ICBBB-05)"
my $fullNamePattern="((?:$ordinals\\s+$words*|$confDescriptors)?$confTypes(?:\\s+$connectors\\s+.*?|\\s+)?$abbreviations?)(?:\\n|\\r|\\.|<)";
############################## ################################# Given a <dbworldMessage>, look for the conference pattern
##############################################################lookForPattern($dbworldMessage, $fullNamePattern);
########################################################## In a given <file>, look for occurrences of <pattern>
# <pattern> is a regular expression#########################################################
sub lookForPattern { my ($file,$pattern) = @_;
24
Example in DBLife (cont.)Example in DBLife (cont.)
# Only look for conference names in the top 20 lines of the file my $maxLines=20;
my $topOfFile=getTopOfFile($file,$maxLines);
# Look for the match in the top 20 lines - case insenstive, allow matches spanning multiple lines if($topOfFile=~/(.*?)$pattern/is) { my ($prefix,$name)=($1,$2);
# If it matches, do a sanity check and clean up the match # Get the first letter
# Verify that the first letter is a capital letter or number if(!($name=~/^\W*?[A-Z0-9]/)) { return (); }
# If there is an abbreviation, cut off whatever comes after that if($name=~/^(.*?$abbreviations)/s) { $name=$1; }
# If the name is too long, it probably isn't a conference if(scalar($name=~/[^\s]/g) > 100) { return (); }
# Get the first letter of the last word (need to this after chopping off parts of it due to abbreviation my ($letter,$nonLetter)=("[A-Za-z]","[^A-Za-z]");
" $name"=~/$nonLetter($letter) $letter*$nonLetter*$/; # Need a space before $name to handle the first $nonLetter in the pattern if there is only one word in name
my $lastLetter=$1; if(!($lastLetter=~/[A-Z]/)) { return (); } # Verify that the first letter of the last word is a capital letter
# Passed test, return a new crutch return newCrutch(length($prefix),length($prefix)+length($name),$name,"Matched pattern in top $maxLines lines","conference
name",getYear($name)); }
return ();}
25
Solution: Declarative, Compositional IESolution: Declarative, Compositional IE
Treat each solution as a “black box” Glue black boxes using a Datalog-like language
– author(y,d) :- docs(d), name(y,d), title(x,d), distance-line(x,y)<3– name(y,d) :- docs(d), seeds(s), namepatterns(s,p), match(p,d,y)– title(x,d) :- docs(d), lines(x,n,d), allcaps(x), (n<5)
DECLARATIVE IE
Dr. R. Ramakrishnan
This is a fun topic ...
Raghu, Ramakrishnan
Divesh, Srivastava
...
seeds(s)
p = Raghu Ramakrishnan R. Ramakrishnan Dr. Ramakrishnan, etc.
26
IE Execution PlanIE Execution Plan
docs(d)
lines(x,n,d)
SELECT_[allcaps(x) and (n<5)]
seeds(s)
namepatterns(p,s) docs(d)
match(y,p,d)
distance-line(x,y)<3
PROJECT_[y,d]
DECLARATIVE IE
Dr. R. Ramakrishnan
This is a fun topic ...
27
Sample Optimization: Push Down SelectionsSample Optimization: Push Down Selections
docs(d)
lines(x,n,d)
SELECT_[allcaps(x) and (n<5)]
seeds(s)
namepatterns(p,s) docs(d)
match(y,p,d)
distance-line(x,y)<3
PROJECT_[y,d]
DECLARATIVE IE
Dr. R. Ramakrishnan
This is a fun topic ...
28
Sample Optimization: Order OperationsSample Optimization: Order Operations
docs(d)
lines(x,n,d)
SELECT_[allcaps(x) and (n<5)]
seeds(s)
namepatterns(p,s) docs(d)
match(y,p,d)
distance-line(x,y)<3
PROJECT_[y,d]
DECLARATIVE IE
Dr. R. Ramakrishnan
This is a fun topic ...
29
Sample Optimization: Sample Optimization: Efficient Large-Scale Pattern MatchingEfficient Large-Scale Pattern Matching
docs(d)
lines(x,n,d)
SELECT_[allcaps(x) and (n<5)]
seeds(s)
namepatterns(p,s) docs(d)
match(y,p,d)
distance-line(x,y)<3
PROJECT_[y,d]
DECLARATIVE IE
Dr. R. Ramakrishnan
This is a fun topic ...
30
Related Project: Avatar @ IBM AlmadenRelated Project: Avatar @ IBM Almaden
Person followed by ContactPattern followed by PhoneNumber
ContactPattern RegularExpression(Email.body,”can be reached at”)
PersonPhone Precedes ( Precedes (Person, ContactPattern, D), Phone, D)
Person can be reached at PhoneNumber
Declarative Query Language
31
DECLARATIVE IE
Dr. R. Ramakrishnan
This is a fun topic ...
Information Extraction: Another ExampleInformation Extraction: Another Example
DECLARATIVE IE
Dr. R. Ramakrishnan
This is a great topic ...
DECLARATIVE IE
Dr. R. Ramakrishnan
More will follow soon ...
time 0
time 1
time 2
How to efficiently extract information over text streams?
32
Data Integration Research: Setting the ContextData Integration Research: Setting the Context Past and current work
– build the foundation: TSIMMIS, Information Manifold, UPenn, P2P, etc.
– develop solutions for specific integration tasks: wrapping, schema matching, entity matching, adaptive QP, etc.
– branching into many app. domains: bioinformatics, PIM (e.g., semex, iMemex), etc.
– top-k, topX query processing
Our work in Cimple– compositional solutions for schema matching, entity matching, etc.
[VLDB-05a, VLDBJ-06, ICDE-07a, Tech Report-07a] – best-effort data integration:
e.g. keyword search + automatic schema matching + automatic entity matching over relational databases [ICDE-07b]
– data integration for masses [Tech Report-07b]
33
Sample Data Integration Challenge in Cimple:Sample Data Integration Challenge in Cimple:Matching Mentions of EntitiesMatching Mentions of Entities
Researcher
Homepages
Conference
Pages
Group Pages
DBworld
mailing list
DBLP
Web pages
Text documents
* **
** * ***
SIGMOD-04
**
** give-talk
Jim Gray
Keyword search
SQL querying
Question answering
Browse
Mining
Alert/Monitor
News summary
Jim Gray
SIGMOD-04
**
Personalize system, provide feedback
34
Extremely Important Problem!Extremely Important Problem!
Appears in numerous real-world contexts Plagues many applications that we have seen
– Citeseer, Rexa, DBLP, InfoZoom, etc.
Why so important? Many services rely on correct mention matching Incorrect matching propagates errors
35
An ExampleAn Example
DBLife incorrectly matches this mention “J. Han” with “Jiawei Han”, but it actually refers to “Jianchao Han”.
Discover related organizations using occurrence analysis:
“J. Han ... Centrum voor Wiskunde en Informatica”
36
Classical Mention MatchingClassical Mention Matching
Applies just a single “matcher” Focuses mainly on improving matcher accuracy
Our key observation: A single matcher often has limited utility
37
Illustrating ExampleIllustrating Example
L. Gravano, K. Ross.Text Databases. SIGMOD 03
L. Gravano, J. Sanz.Packet Routing. SPAA 91
MembersL. Gravano K. Ross J. Zhou
L. Gravano, J. Zhou.Text Retrieval. VLDB 04
C. Li.Machine Learning. AAAI 04
C. Li, A. Tung.Entity Matching. KDD 03
Luis Gravano, Kenneth Ross. Digital Libraries. SIGMOD 04
Luis Gravano, Jingren Zhou.Fuzzy Matching. VLDB 01
Luis Gravano, Jorge Sanz.Packet Routing. SPAA 91
Chen Li, Anthony Tung.Entity Matching. KDD 03
Chen Li, Chris Brown. Interfaces. HCI 99
d4: Chen Li’s Homepage
d1: Luis Gravano’s Homepage d2: Columbia DB Group Page d3: DBLP
Only one Luis Gravano
Two Chen Li-s What is the best way to match mentions here?
38
A liberal matcher: A liberal matcher: good for matching Luis Gravano, good for matching Luis Gravano,
bad for matching Chen Libad for matching Chen Li
L. Gravano, K. Ross.Text Databases. SIGMOD 03
L. Gravano, J. Sanz.Packet Routing. SPAA 91
MembersL. Gravano K. Ross J. Zhou
L. Gravano, J. Zhou.Text Retrieval. VLDB 04
C. Li.Machine Learning. AAAI 04
C. Li, A. Tung.Entity Matching. KDD 03
Luis Gravano, Kenneth Ross. Digital Libraries. SIGMOD 04
Luis Gravano, Jingren Zhou.Fuzzy Matching. VLDB 01
Luis Gravano, Jorge Sanz.Packet Routing. SPAA 91
Chen Li, Anthony Tung.Entity Matching. KDD 03
Chen Li, Chris Brown. Interfaces. HCI 99
d4: Chen Li’s Homepage
d1: Luis Gravano’s Homepage d2: Columbia DB Group Page d3: DBLP
s0 matcher: two mentions match if they share the same name.
39
A conservative matcher: A conservative matcher: good for matching Chen Li, good for matching Chen Li,
bad for matching Luis Gravanobad for matching Luis Gravano
L. Gravano, K. Ross.Text Databases. SIGMOD 03
L. Gravano, J. Sanz.Packet Routing. SPAA 91
MembersL. Gravano K. Ross J. Zhou
L. Gravano, J. Zhou.Text Retrieval. VLDB 04
C. Li.Machine Learning. AAAI 04
C. Li, A. Tung.Entity Matching. KDD 03
Luis Gravano, Kenneth Ross. Digital Libraries. SIGMOD 04
Luis Gravano, Jingren Zhou.Fuzzy Matching. VLDB 01
Luis Gravano, Jorge Sanz.Packet Routing. SPAA 91
Chen Li, Anthony Tung.Entity Matching. KDD 03
Chen Li, Chris Brown. Interfaces. HCI 99
d4: Chen Li’s Homepage
d1: Luis Gravano’s Homepage d2: Columbia DB Group Page d3: DBLP
s1 matcher: two mentions match if theyshare the same name and at least one co-author name.
40
Better solution: Better solution: apply both matchers in a workflowapply both matchers in a workflow
L. Gravano, K. Ross.Text Databases. SIGMOD 03
L. Gravano, J. Sanz.Packet Routing. SPAA 91
MembersL. Gravano K. Ross J. Zhou
L. Gravano, J. Zhou.Text Retrieval. VLDB 04
C. Li.Machine Learning. AAAI 04
C. Li, A. Tung.Entity Matching. KDD 03
Luis Gravano, Kenneth Ross. Digital Libraries. SIGMOD 04
Luis Gravano, Jingren Zhou.Fuzzy Matching. VLDB 01
Luis Gravano, Jorge Sanz.Packet Routing. SPAA 91
Chen Li, Anthony Tung.Entity Matching. KDD 03
Chen Li, Chris Brown. Interfaces. HCI 99
d4: Chen Li’s Homepage
d1: Luis Gravano’s Homepage d2: Columbia DB Group Page d3: DBLP
union
d1 d2
s0
s1
union
d3
d4
s0 s0 matcher: two mentions match if they share the same name.
s1 matcher: two mentions match if theyshare the same name and at least one co-author name.
41
Key ChallengesKey Challenges
How to compose matchers, to form a space of workflows?
How to estimate the accuracy of each workflow?
How to efficiently find one with high accuracy?
union
d1 d2
s0
s1
union
d3
d4
s0
[See ICDE-07a]
42
Mass Collaboration: The General IdeaMass Collaboration: The General Idea
Many applications have multiple developers / users– how to exploit feedback from all of them?
Variants of this is known as – collective development of system, mass collaboration,
collective curation, Web 2.0 applications, social software, etc.
Has been applied to many applications– open-source software, bug detection, tech support group, Yahoo!
Answers, Google Co-op, and many more
Studied in some academic contexts, e.g., ESP Game
Little has been done in extraction / integration contexts– except in industry, e.g., epinions.com
45
Key ChallengesKey Challenges
What types of extraction / integration tasks are most amenable to mass collaboration?– e.g., see MOBS project at Illinois [WebDB-03, ICDE-05]
How to entice people to contribute? What can they contribute? What is the underlying data model? How to handle the Naughton effect? How to propagate user contributions? How to undo? How to reconcile multiple conflicting editions?
– e.g., see ORCHESTRA project at Penn [Taylor & Ives, SIGMOD-06]
46
Sample Research: SummarySample Research: Summary
Information extraction– how to do it in a declarative / compositional fashion? – how to apply database-like optimization techniques?
Data integration– how to do it incrementally (best effort, pay-as-you-go)?
an example of a Data Space? – how to do it in a compositional fashion?
Human computation / mass collaboration– new! (Though industry has been doing it for years.)– how to do it for data management tasks?
47
ConclusionsConclusions Community Information Management
– increasingly crucial problem
The Cimple project– sample challenges: information extraction
data integration human computation
– extends the footprints of DB technologies to Web data– develops new DB technologies
DBLife prototype– research/education tool, community service, benchmark
Search “cimple wisc” for project homepage
48
Broader PerspectivesBroader Perspectives[speculation mode][speculation mode]
Current Web: keyword search over text Future Web
– should have increasingly more structure– should have more ways to exploit structure– should be more “social”
This future Web should be great for our community– we are the “Structure King” – if the Web remains text-centric not as good for us
How to accelerate the coming of this future Web?– Cimple and many current projects can contribute– but as a community we need more efforts in this direction!