f rom Names and Phrases t o Entities and Relations Gerhard Weikum Max Planck Institute for Informatics Saarbrücken Germany httpwwwmpiinfmpgdeweikum ID: 605273
Download Presentation The PPT/PDF document "Big Text:" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
Big Text:from Names and Phrasesto Entities and Relations
Gerhard Weikum
Max Planck Institute
for
Informatics
Saarbrücken, Germany
http://www.mpi-inf.mpg.de/~weikum/Slide2
From Natural-Language Text to Knowledge
Web
Contents
Knowledge
knowledge
acquisition
intelligent
interpretation
m
ore
knowledge
,
analytics
,
insightSlide3
Cyc
TextRunner/
ReVerb
WikiTaxonomy/
WikiNet
SUMO
ConceptNet
5
BabelNet
ReadTheWeb
http://richard.cyganiak.de/2007/10/lod/lod-datasets_2011-09-19_colored.png
Web
of
Data & Knowledge
(
Linked
Open Data)
> 50 Bio.
s
ubject-predicate-object
triples
from
> 1000
sourcesSlide4
10M
entities
in
350K
classes
120M
facts
for
100
relations
100
languages
95%
accuracy
4M
entities
in
250
classes
500M
facts
for
6000
properties
live
updates
40M
entities
in
15000
topics
1B facts for 4000
properties core of
Google Knowledge
Graph
Web of Data & Knowledge
600M
entities
in
15000
topics
20B
facts
> 50 Bio. subject-predicate-object triples from > 1000 sourcesSlide5
Web
of
Data & Knowledge
> 50 Bio.
s
ubject
-
predicate
-
object
triples
from
> 1000
sources
taxonomic
knowledge
factual
knowledge
temporal
knowledge
evidence
& belief
knowledget
erminological knowledge
Yimou_Zhang
type
movie_director
Yimou_Zhang
type
olympic_games_participant
movie_director subclassOf
artist
Yimou_Zhang
directed Shanghai
Triad
Li Gong actedIn
Shanghai Triad
Yimou_Zhang
memberOf Beijing_film_academy
validDuring
[1978, 1982]
Yimou_Zhang
„was classmate
of“ Kaige_Chen
Yimou_Zhang
„
had love
affair with
“ Li_Gong
Li_Gong knownAs
„China‘s
most beautiful
“Slide6
Knowledge for Intelligent
Applications
Enabling
technology
for
:
disambiguation
in
written
&
spoken
natural
language
deep reasoning
(e.g. QA to
win quiz
game)machine
reading
(e.g. to
summarize book or
corpus)
semantic search
in
terms of entities&relations (not keywords&pages
)entity-level
linkage
for Big Data & Big Text analyticsSlide7
Big Text Analytics: Who Covered Whom?1000‘s of Databases100 Mio‘s of Web Tables100 Bio‘s of Web & Social Media Pages
Elvis Presley Frank Sinatra
My
Way
Robbie Williams Frank Sinatra
My
Way
Sex
Pistols
Frank Sinatra
My
Way
Frank Sinatra Claude Francois
Comme
d‘Habitude
. . . . . . . . . . . . . . .
Musician
Original Title
in different
language
,
country
,
key
, …
w
ith
more
sales
,
awards
,
media
buzz
, …
.....Slide8
Big Text Analytics: Who Covered Whom?1000‘s of Databases100 Mio‘s of Web Tables100 Bio‘s of Web & Social Media Pages
Claudia
Leitte
Bruno Mars
Billionaire
Only
Won Bruno Mars
I
wanna
be
an
engineer
Yip
Sin-Man Celine Dion
我心永恒
Norah
Jones Bob Dylan
Forever
Young
Beyonce
Alphaville
Forever
Young
Jay-Z
Alphaville
Forever
Young
. . . . . . . . . . . . . . .
Musician
Original Title
in different
language
,
country
,
key
, …
w
ith
more
sales
,
awards
,
media
buzz
, …
.....Slide9
Big Text Analytics: Who Covered Whom?1000‘s of Databases100 Mio‘s of Web Tables100 Bio‘s of Web & Social Media Pagesin different language, country, key, …with more sales, awards,
media
buzz
, …
.....
Sex
Pistols
My
Way
Frank Sinatra
My
Way
Only
Won I
wanna
be an
engineer Beyonce
Forever Young
Musician
PerformedTitle
Francis Sinatra
My
Way
Paul Anka
My
Way
Bruno Mars
Billionaire
Bob Dylan
Forever
Young
Musician
CreatedTitle
Name Group
Sid
Vicious
Sex
Pistols
Bono
U2
Name Event
Beyonce
Grammy
Norah
J. Jobs
Mem
.
Claudia L. FIFA 2014Slide10
Big Text Analytics: Who Covered Whom?1000‘s of Databases100 Mio‘s of Web Tables100 Bio‘s of Web & Social Media Pagesin different language, country, key, …with more sales, awards,
media
buzz
, …
.....
Sex
Pistols
My
Way
Frank Sinatra
My
Way
Claudia
Leitte
Famo$a
Only
Won I wanna
be an engineer
Beyonce
Forever Young
Musician
PerformedTitle
Francis Sinatra
My
Way
Paul Anka
My
Way
Bruno Mars
Billionaire
Travie
McCoy
Billionaire
Bob Dylan
Forever Young
Musician CreatedTitle
Big Data
Volume
Velocity
Variety
Veracity
Big Data & Big Text
Volume
Velocity Variety
VeracitySlide11
Big Data & Big Text Analytics Who covered which other singer?Who influenced which other musicians
?
Entertainment:
Drugs (
combinations
)
and
their
side effects
Health:
Politicians‘ positions on
controversial topics and their
involvement with industry
Politics:
Customer opinions on
small-company products,gathered from social media
Business:
Identify relevant contents
sources Identify
entities of interest & their
relationships Position in time &
space Group and aggregate
Find insightful patterns & predict
trendsGeneral Design Pattern:
Trends in society,
cultural factors, etc.
Culturomics:Slide12
OutlineLovely NERD The New Chocolate
Conclusion
Introduction
The Dark SideSlide13
Lovely NERDSlide14
Named
Entity
Recognition &
Disambiguation
(NERD)
p
rior
popularity
o
f
name-
entity
pairs
Zhang
played
in
Lee‘s
e
pic
masterpiece
,
w
ith
Ma‘s
score.
Ma also
covered
t
he
Ecstasy.
contextual
similarity
:
mention
vs.
entity
(
bag-of-words
,
l
anguage
model
)Slide15
Named
Entity
Recognition &
Disambiguation
(NERD)
Zhang
played
in
Lee‘s
e
pic
masterpiece
,
w
ith
Ma‘s
score.
Ma also
covered
t
he
Ecstasy.
Coherence
of
entity
pairs
:
s
emantic
relationships
s
hared
types
(
categories
)
overlap
of
Wikipedia linksSlide16
Named
Entity
Recognition &
Disambiguation
(NERD)
Zhang
played
in
Lee‘s
e
pic
masterpiece
,
w
ith
Ma‘s
score.
Ma also
covered
t
he
Ecstasy.
Coherence
: (partial)
overlap
o
f
(
statistically
weighted
)
entity-specific
keyphrases
Crouching
Tiger
Memoirs
of
a Geisha
Chinese Film
Actress
Grammy Award
Best
Classical
Album
Tan Dun
Composition
Morricone Tribute
Good
, Bad,
Ugly
Western Movie Score
Ennio Morricone
Metallica
cover
song
Crouching
Tiger
Academy Award
Chinese American
Score
by
Tan DunSlide17
Named
Entity
Recognition &
Disambiguation
Zhang
played
in
Lee‘s
e
pic
masterpiece
,
w
ith
Ma‘s
score.
Ma also
covered
t
he
Ecstasy.
NED
algorithms
compute
m
ention-to-entity
mapping
over
weighted
graph
of
candidates
b
y
popularity
&
similarity
&
coherence
KB
provides
building
blocks
:
n
ame-
entity
dictionary
,
relationships
,
types
,
t
ext
descriptions
,
keyphrases
,
s
tatistics
for
weightsSlide18
Joint Mapping Build mention-entity graph or probabilistic factor graph from knowledge and statistics in KB Compute
high-
likelihood
mapping
(ML
or
MAP)
or dense subgraph
(with high total edge weight) such that
: each m is connected
to exactly one
e (or at most
one e)
90
30
5
100
100
50
20
50
90
80
90
30
10
10
20
30
30
m1
m2
m3
m4
e
1
e2
e3
e4
e5
e6Slide19
Coherence Graph Algorithm
90
30
5
100
100
50
50
90
80
90
30
10
20
10
20
30
30
140
180
50
470
145
230
D5 Overview May 14, 2013
19
Compute
dense
subgraph
to
maximize
min
weighted
degree
among
entity
nodes
such
that
:
each
m
is
connected
to
exactly
one
e (
or
at
most
one
e)
Approx
.
algorithms
(
greedy
,
randomized
, …),
hash
sketches
, …
82%
precision
on CoNLL‘03
benchmark
Open-
source
software
& online
service
AIDA
http://www.mpi-inf.mpg.de/yago-naga/aida/
m1
m2
m3
m4
e
1
e2
e3
e4
e5
e6
[J. Hoffart et al.:
EMNLP‘11, CIKM‘12,
WWW’14]Slide20
NERD Online ToolsJ. Hoffart et al.: EMNLP 2011, VLDB 2011https://d5gate.ag5.mpi-sb.mpg.de/webaida/P. Ferragina, U. Scaella: CIKM 2010http://tagme.di.unipi.it/P.N. Mendes, C. Bizer et al.: I-Semantics 2011, WWW 2012http://spotlight.dbpedia.org/demo/index.htmlD. Milne, I. Witten: CIKM 2008http://wikipedia-miner.cms.waikato.ac.nz/demos/annotate/L. Ratinov, D. Roth, D. Downey
, M. Anderson: ACL 2011
http://
cogcomp.cs.illinois.edu/page/demo_view/Wikifier
D.
Odijk
, E.
Meij
, M. de
Rijke:OAIR 2013
http://semanticize.uva.nlReuters Open Calais: http://viewer.opencalais.com/
Alchemy API: http://www.alchemyapi.com/api/demo.html
Slide21
NERD at Workhttps://gate.d5.mpi-inf.mpg.de/webaida/Slide22
Research Challenges & OpportunitiesHandling long-tail and emerging entitiest
o
complement
and
continuously
update KB
key for KB life-cycle
management
Entity
name disambiguation in
difficult situations
Short and noisy
texts about long-tail
entities in social media
Efficient
interactive & high-throughput
batch NERD
a day‘s news, a month‘s
publications, a decade‘s archive
Web-
scale
entity
linkage with high quality
across text sources
, linked data, KB‘s, Web
tables, …Slide23
OutlineLovely NERD The New Chocolate
Conclusion
Introduction
The Dark Side
Slide24
Big Text: the New ChocolateSlide25
Semantic Search over News
s
ee
also J. Hoffart et al.:
demo
@ CIKM‘14
stics.mpi-inf.mpg.deSlide26
Semantic Search over News
stics.mpi-inf.mpg.deSlide27
Semantic Search over News
stics.mpi-inf.mpg.deSlide28
Semantic Search over Newsstics.mpi-inf.mpg.deSlide29
Analytics over Newsstics.mpi-inf.mpg.de/statsSlide30
Machine Reading of Scholarly Papers
https://gate.d5.mpi-inf.mpg.de/knowlife/
[P. Ernst et al.: ICDE‘14]Slide31
Machine Reading of Health Forums
https://gate.d5.mpi-inf.mpg.de/knowlife/
[P. Ernst et al.: ICDE‘14]Slide32
Big Data & Text Analytics:Side Effects of Drug Combinations
http://dailymed.nlm.nih.gov
http://www.patient.co.uk
Deeper
insight
from
both
expert
data
&
social
media
:
a
ctual
side
effects
of drugs…
and drug combinations
risk factors
and complications of (
wide-spread) diseases
alternative therapiesaggregation & comparison
by age, gender, life style, etc.
Structured
Expert Data
SocialMediaSlide33
Machine Reading (Semantic Parsing): from Names & Phrases to Entities, Classes & Relations
Rome
(
Italy
)
Lazio
Roma
AS
Roma
Maestro
Card
Ennio
Morricone
MDMA
l
‘Estasi
d
ell‘Oro
Leonard
Bernstein
Jack
Ma
Yo-Yo
Ma
c
over
of
s
tory
about
b
orn
in
p
lays
for
f
ilm
music
goal
in
football
p
lays
sport
p
lays
musicwestern
movie
WesternDigital
Massachusetts
Ma
played
his
version
of
the
Ecstasy.
The Maestro
from
Rome
wrote
scores
for
westerns
.Slide34
Paraphrases of Relations
Dylan
wrote a
sad song
Knockin
‘ on Heaven‘s
Door, a
cover song
by the
Dead
Morricone
‘s masterpiece
is
the Ecstasy of
Gold, covered by Yo-Yo MaAmy‘s
souly
interpretation of Cupid, a classic piece
of Sam CookeNina
Simone‘s singing of
Don‘t Explain revived
Holiday‘s old song
Cat Power‘s
voice is haunting in her
version of Don‘t Explain
Cale performed Hallelujah
written by
L. Cohen
c
omposed
:
musician
song
covered
:
musician
song
Sequence
Mining
w
ith
Type Lifting(N. Nakashole et al.: EMNLP’12, ACL’13,
VLDB‘12)
composed
:
wrote
,
classic
piece of, ‘s
old song, written
by, composed, …
350 000 SOL
patterns
from
Wikipedia: http://www.mpi-inf.mpg.de/yago-naga/patty/
<singer> covered
<song> <book> covered <event>
Relational phrases are
typed:
SOL patterns
over words, wildcards, POS tags, semantic types:
<
musician
>
wrote
ADJ
piece
<song>
Relational
synsets
(and
subsumptions
):
covered
:
cover
song
, interpretation
of,
singing
of,
voice in
version, … Slide35
wrote scoresr: composed
Disambiguation
for
Entities
,
Classes
& Relations
s
cores
for
westerns
from
Rome
Maestro
Combinatorial
Optimization
by
ILP (
with
type
constraints
etc.)
e:
Rome
(
Italy
)
e: Lazio Roma
e:
MaestroCard
e: Ennio Morricone
c:
conductor
c:soundtrack
r:
soundtrackFor
r:
shootsGoalFor
r:
bornIn
r:
actedIn
c: western
movie
e: Western Digital
w
eighted
edges
(
coherence
,
similarity
, etc.)
(M. Yahya et
al
.: EMNLP’12, CIKM‘13)
c
:
musician
r:
giveExam
ILP
optimizers
like
Gurobi
solve
this
in
secondsSlide36
Research Challenges & OpportunitiesManaging text & data on par: DB&IR integration, finallytext with
SPO/ER
markup
linked
to
DB/KB
records
DB/KB records linked
to provenance, context, fact
spottingsQuery &
analytics language – API ?
Sparql Full-Text, extended
/ relaxed Sparql, data&text cube, …
User-friendly
search & exploration –
UI ? natural language, visual, multimodal, …
Context-aware
“strings, things,
cats“: suggested entities
& categories should consider
input prefix
Comprehensive repository of
relational paraphrases: binary-predicate
counterpart of WordNet – still elusive
despite Patty, WiseNet, Probase,
ReVerb, ReNoun,
Biperpedia, …Slide37
Research Challenges & Opportunities
Managing
text
&
data
on par:
DB&IR
integration
,
finally
t
ext
with
SPO/ER markup linked to DB/KB
recordsDB/KB records linked
to provenance
, context, fact spottings
Query & analytics language – API ?
Sparql Full-Text, extended
/ relaxed Sparql, data&text cube, …
User-friendly
search & exploration – UI ?
natural language speech
, visual, multimodal, …
Context-aware “strings,
things, cats“:
suggested entities &
categories should consider input prefix
Comprehensive repository
of relational paraphrases:
binary-predicate counterpart of WordNet - still elusive
despite Patty, WiseNet,
Probase, ReVerb, ReNoun, Biperpedia
, …Slide38
OutlineLovely NERD The New ChocolateConclusion
Introduction
The Dark Side
Slide39
The Dark Side of Big Data
Nobody
interested
in
your
research
?
We
read
your
papers
!Slide40
search
Internet
p
ublish & recommend
Entity
Linking: Privacy at Stake
Levothroid
shaking
Addison
’
s disease
………
Nive
concert
Greenland singers
Somalia elections
Steve
Biko
s
earch
engine
Zoe
female 29y
Jamame
s
ocial
network
Nive
Nielsen
Cry
Freedom
discuss
& seek
help
o
nline
forum
female
25-30 Somalia
Synthroid
tremble
……….
Addison
disorder
……….Slide41
s
ocial
network
Synthroid
tremble
……….
Addison
disorder
……….
search
Internet
Privacy
Adversaries
Linkability
Threats
:
Weak cues:
profiles
,
friends
, etc.
Semantic cues:
health
, taste, queries
Statistical
cues
:
correlations
discuss
& seek
help
p
ublish & recommend
L
evothroid
shaking
Addison
’
s disease
………
Nive
concert
Greenland
singers
Somalia
elections
Steve
Biko
Nive
Nielsen
female
25-30 Somalia
female 29y
Jamame
o
nline
forum
s
earch
engine
Cry
FreedomSlide42
s
ocial
network
search
s
ocial
network
s
earch
engine
o
nline
forum
Synthroid
tremble
……….
Addison
disorder
……….
Internet
L
evothroid
shaking
Addison’s
disease
………
Nive
concert
Greenland
singers
Somalia
elections
female
25-30 Somalia
Goal:
Automated
Privacy
Advisor
discuss
& seek
help
p
ublish & recommend
Nive
Nielsen
Cry
Freedom
female 29y
Jamame
Privacy
Adviser
(PA):
S
oftware
tool
that
analyses risk
alerts user
advises
user
explains consequences
recommends
policy
changes
Your
queries
may
lead
to
linking
your
identies
i
n
Facebook
and
patient.co.uk !
………….
Would
you
like
to
use
an
anonymization
tool
for
your
search
requests
?
………..
ERC Project
imPACT
(M. Backes, P.
Druschel
, R.
Majumdar
, G. Weikum)
s
ee
also: J.
Biega
et al.: PSBD Workshop @ CIKM‘14 Slide43
Research Challenges & OpportunitiesLong-term privacy management:policies, risks, privacy-utility trade-offs
Explain
risks
,
advise
on
consequences
,
r
ecommend
counter-
measures
and
mitigation
stepsWhich data is
(or can be
linked to become) privacy-critical
?highly user-specific, but needs a global
perspectiveHow
are privacy risks building
up over time?
where is my
data, who has seen
it, who can copy
and accumulate
it
Who
are the
adversaries? How powerful? At which
cost? role of
background knowledge &
statistical learningSlide44
OutlineLovely NERD The New ChocolateConclusion
Introduction
The Dark Side
Slide45
Big Text & Big DataBig Text & NERD: valuable content about entities lifted towards knowledge &
analytic
insight
Machine
Reading:
discover
and
interpret
names
& phrases
as entities, classes, relations
, spatio-temporal
modifiers, sentiments, beliefs, ….Big Data:
interlink natural-language text,
social media, structured data
& knowledge bases, images,
videosa
nd help users
coping with privacy
risksSlide46
intelligentinterpretation
Take-Home Message
Web
Contents
Knowledge
more
knowledge
,
analytics
,
insight
k
nowledge
acquisition
Knowledge
„Who
Covered
Whom
?“
and
More!
(
Entities
,
Classes
, Relations)