20-12-2012 àBravo
Given a drug under development, what are other
drugs or biomedical compounds that it might
interact with?
Which are the proteins
targeted by celecoxib?
Which are the genes that
cause lung cancer?
TEXT
MINING
Which are the genes that
cause lung cancer?
Given a drug under development, what are other
drugs or biomedical compounds that it might
interact with?
Which are the proteins
targeted by celecoxib?
TEXT
MINING
Which are the genes that
cause lung cancer?
Given a drug under development, what are other
drugs or biomedical compounds that it might
interact with?
Which are the proteins
targeted by celecoxib?
TEXT
MINING
Which are the genes that
cause lung cancer?
Given a drug under development, what are other
drugs or biomedical compounds that it might
interact with?
Which are the proteins
targeted by celecoxib?
INFORMATION
EXTRACTION
TEXT
MINING
Which are the genes that
cause lung cancer?
Given a drug under development, what are other
drugs or biomedical compounds that it might
interact with?
Which are the proteins
targeted by celecoxib?
RESULTS
Workflow
ENTITY
EXTRACTION
RELATION
EXTRACTION
PRE
PROCESSING
Workflow
ENTITY
EXTRACTION
RELATION
EXTRACTION
PRE
PROCESSING
Workflow
Dictionary X-Ref
ENTITY
EXTRACTION
RELATION
EXTRACTION
PRE
PROCESSING
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SCORE
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SCORE PMI
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SCORE PMI
DiseasesGenes
g d
NER
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SCORE PMI
DiseasesGenes
g d
NER
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SCORE PMI
g d
Diseases
g
Genes
d
DiseasesGenes
X
PMI =
g d
g d
NER
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
SCORE
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
Supervised classification algorthm
Training Binary
SCORE
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
Supervised classification algorthm
Training Binary
SCORE
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
Supervised classification algorthm
Training Binary
SCORE
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
Supervised classification algorthm
Training Binary
SCORE
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
Supervised classification algorthm
Training Binary
SUPPORT
VECTOR
MACHINE
(SVM)
SCORE
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
Non-linear
cases???
SCORE
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
SCORE
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
RBF
SCORE
(RADIAL BASIS FUNCTIONS)
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
RBF
SCORE
(RADIAL BASIS FUNCTIONS)
Gene-Disease
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
RBF
SCORE
(RADIAL BASIS FUNCTIONS)
Gene-Disease
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
RBF
KERNELSCORE
Gene-Disease
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
SCORE
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
RBF
YEEES!
KERNEL
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
SCORE
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
KERNEL
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
KERNELSCORE
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
KERNELSCORE
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
KERNELSCORE
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
SCORE
CALCULATION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
KERNEL
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
SCORE
CALCULATION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
KERNEL
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
SCORE
CALCULATION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
KERNEL
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
KERNELSCORE
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
TEXT MINING???KERNELSCORE
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
TEXT MINING???KERNEL
N-dimensional feature vector
len position POS…
len position POS…
len position POS…
SCORE
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
TEXT MINING???KERNEL
N-dimensional feature vector
len position POS…
len position POS…
len position POS…
Trained
Model
SVM SYSTEM
SCORE
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
TEXT MINING???KERNEL
N-dimensional feature vector
len position POS…
len position POS…
len position POS…
Trained
Model
Or ? SVM SYSTEMlen position POS…
SCORE
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
TEXT MINING???KERNEL
N-dimensional feature vector
len position POS…
len position POS…
len position POS…
Trained
Model
Or ?
Prediction
SVM SYSTEMlen position POS…
SCORE
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
Feature vector
Annotated Corpus
Input Space Feature Space
KERNEL
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
Feature vector
Annotated Corpus
Input Space Feature Space
KERNEL
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
Feature vector
Annotated Corpus
Input Space Feature Space
KERNEL
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
Feature vector
Annotated Corpus
Input Space Feature Space
Lib
SVM
KERNEL
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
Feature vector
Annotated Corpus
Input Space Feature Space
Lib
SVM
JSRE
Java tool for Relation Extraction.
When (deep) linguistic processing is not
available.
Combination of kernel functions
to integrate two different information
sources: Global and Local.
KERNEL
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
Feature vector
Annotated Corpus
Input Space Feature Space
Lib
SVM
JSRE K1 K2
K
KERNEL
U =
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
JSRE
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
JSRE
Expression of the sigma(K) dependent cwlH gene depended on gerR.
Expression of the sigma(K) dependent cwlH gene depended on gerR.
Expression of the sigma(K) dependent cwlH gene depended on gerR.
Expression of the sigma(K) dependent cwlH gene depended on gerR.
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
JSRE
Expression of the sigma(K) dependent cwlH gene depended on gerR.
Expression of the sigma(K) dependent cwlH gene depended on gerR.
Expression of the sigma(K) dependent cwlH gene depended on gerR.
Expression of the sigma(K) dependent cwlH gene depended on gerR.
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
JSRE
Expression of the sigma(K) dependent cwlH gene depended on gerR.
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
JSRE
Expression of the sigma(K) dependent cwlH gene depended on gerR.
BETWEEN AFTERFORE [1] [2]
GLOBAL
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
JSRE
Expression of the sigma(K) dependent cwlH gene depended on gerR.
BETWEEN AFTERFORE [1] [2]
GLOBAL
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
JSRE
Expression of the sigma(K) dependent cwlH gene depended on gerR.
BETWEEN AFTERFORE [1] [2]
GLOBAL
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
JSRE
Expression of the sigma(K) dependent cwlH gene depended on gerR.
BETWEEN AFTERFORE [1] [2]
GLOBAL
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
JSRE
Expression of the sigma(K) dependent cwlH gene depended on gerR.
GLOBAL
BETWEEN AFTERFORE [1] [2]
FEATURES 3-GRAM
Expression_of_the
of_the_sigma(k)
the_sigma(k)_dependent
sigma(k)_dependent_cwlH dependent_cwlH_gene
cwlJ_gene_depended
gene_depended_on
depended_on_gerB
FORE-BETWEEN BETWEEN BETWEEN-AFTER
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
JSRE
Expression of the sigma(K) dependent cwlH gene depended on gerR.
GLOBAL
BASIC FEATURESLOCAL
Token
Stem
POS
Orthographic
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
JSRE
Expression of the sigma(K) dependent cwlH gene depended on gerR.
GLOBAL
LOCAL
KERNEL
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
JSRE
Expression of the sigma(K) dependent cwlH gene depended on gerR.
GLOBAL
LOCAL
KERNEL
DEP
Workflow
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
MACHINE
LEARNING
SUPPORT
VECTOR
MACHINE
(SVM)
JSRE
Expression of the sigma(K) dependent cwlH gene depended on gerR.
GLOBAL
LOCAL
KERNEL
DEP
?
Workflow: Gene-Disease
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
Workflow: Gene-Disease
ENTITY
EXTRACTION
PRE
PROCESSING
RELATION
EXTRACTION
19098994
Workflow: Gene-Disease
PRE
PROCESSING
RELATION
EXTRACTION
The low frequencies of the at-risk alleles at rs1048661 and
rs2165241 may be one of the factors that led to the low
prevalence of exfoliation syndrome in the general populations of
the Chinese.
ENTITY
EXTRACTION
Workflow: Gene-Disease
PRE
PROCESSING
The low frequencies of the at-risk alleles at rs1048661 and
rs2165241 may be one of the factors that led to the low
prevalence of exfoliation syndrome in the general populations of
the Chinese.
ENTITY
EXTRACTION
RELATION
EXTRACTION
The low frequencies of the at-risk alleles at rs1048661 and
rs2165241 may be one of the factors that led to the low
prevalence of exfoliation syndrome in the general populations of
the Chinese.
The low frequencies of the at-risk alleles at rs1048661 and
rs2165241 may be one of the factors that led to the low
prevalence of exfoliation syndrome in the general populations of
the Chinese.
Workflow: Gene-Disease
PRE
PROCESSING
The low frequencies of the at-risk alleles at rs1048661 and
rs2165241 may be one of the factors that led to the low
prevalence of exfoliation syndrome in the general populations of
the Chinese.
ENTITY
EXTRACTION
RELATION
EXTRACTION
The low frequencies of the at-risk alleles at rs1048661 and
rs2165241 may be one of the factors that led to the low
prevalence of exfoliation syndrome in the general populations of
the Chinese.
The low frequencies of the at-risk alleles at rs1048661 and
rs2165241 may be one of the factors that led to the low
prevalence of exfoliation syndrome in the general populations of
the Chinese.
Workflow: Gene-Disease
PRE
PROCESSING
The low frequencies of the at-risk alleles at rs1048661 and
rs2165241 may be one of the factors that led to the low
prevalence of exfoliation syndrome in the general populations of
the Chinese.
ENTITY
EXTRACTION
RELATION
EXTRACTION
GLOBAL LOCAL
Workflow: Gene-Disease
PRE
PROCESSING
The low frequencies of the at-risk alleles at rs1048661 and
rs2165241 may be one of the factors that led to the low
prevalence of exfoliation syndrome in the general populations of
the Chinese.
ENTITY
EXTRACTION
RELATION
EXTRACTION
Biolemmatizer Stanford Parser
GLOBAL LOCAL
DEP
Workflow: Gene-Disease
PRE
PROCESSING
The low frequencies of the at-risk alleles at rs1048661 and
rs2165241 may be one of the factors that led to the low
prevalence of exfoliation syndrome in the general populations of
the Chinese.
ENTITY
EXTRACTION
RELATION
EXTRACTION
Biolemmatizer Stanford Parser
GLOBAL LOCAL
DEP
Workflow: Gene-Disease
PRE
PROCESSING
The low frequencies of the at-risk alleles at rs1048661 and
rs2165241 may be one of the factors that led to the low
prevalence of exfoliation syndrome in the general populations of
the Chinese.
ENTITY
EXTRACTION
RELATION
EXTRACTION
Tokens + POS
STANFORD PARSER
Workflow: Gene-Disease
PRE
PROCESSING
The low frequencies of the at-risk alleles at rs1048661 and
rs2165241 may be one of the factors that led to the low
prevalence of exfoliation syndrome in the general populations of
the Chinese.
ENTITY
EXTRACTION
RELATION
EXTRACTION
Tokens + POS
STANFORD PARSER
The low
frequencies
the
risk
alleles
rs1048661
rs2165241
may be
one
the
factors
led
the low
prevalence
exfoliation_syndrome
the general
populations
the
Chinese
at
Workflow: Gene-Disease
PRE
PROCESSING
The low frequencies of the at-risk alleles at rs1048661 and
rs2165241 may be one of the factors that led to the low
prevalence of exfoliation syndrome in the general populations of
the Chinese.
ENTITY
EXTRACTION
RELATION
EXTRACTION
Tokens + POS
STANFORD PARSER
DEP
The low
frequencies
the
risk
alleles
rs1048661
rs2165241
may be
one
the
factors
led
the low
prevalence
exfoliation_syndrome
the general
populations
the
Chinese
at
Workflow: Gene-Disease
PRE
PROCESSING
ENTITY
EXTRACTION
RELATION
EXTRACTION
DEP
The low
frequencies
the
risk
alleles
rs1048661
rs2165241
may be
one
the
factors
led
the low
prevalence
exfoliation_syndrome
the general
populations
the
Chinese
at
Workflow: Gene-Disease
PRE
PROCESSING
ENTITY
EXTRACTION
RELATION
EXTRACTION
DEP
The low
frequencies
the
risk
alleles
rs1048661
rs2165241
may be
one
the
factors
led
the low
prevalence
exfoliation_syndrome
the general
populations
the
Chinese
at
Least Common Subsumer
Workflow: Gene-Disease
PRE
PROCESSING
ENTITY
EXTRACTION
RELATION
EXTRACTION
DEP
The low
frequencies
the
risk
alleles
rs1048661
rs2165241
may be
one
the
factors
led
the low
prevalence
exfoliation_syndrome
the general
populations
the
Chinese
at
Least Common Subsumer
Workflow: Gene-Disease
PRE
PROCESSING
ENTITY
EXTRACTION
RELATION
EXTRACTION
DEP
frequencies
alleles
rs1048661
rs2165241
one
factors
led
prevalence
exfoliation_syndrome
at
Least Common Subsumer
Workflow: Gene-Disease
PRE
PROCESSING
ENTITY
EXTRACTION
RELATION
EXTRACTION
DEP
frequenciesallelesrs1048661rs2165241 one factors led prevalence exfoliation_syndromeat
Least Common Subsumer
Workflow: Gene-Disease
PRE
PROCESSING
ENTITY
EXTRACTION
RELATION
EXTRACTION
DEP
frequenciesallelesrs1048661rs2165241 one factors led prevalence exfoliation_syndromeat
Least Common Subsumer
We know:
• Token
• Stem
• Lemma
• Role  NEW!!!
• Dependency Type
• Dependency Node
frequencies oneat prep_of nsubj
Workflow: Gene-Disease
PRE
PROCESSING
ENTITY
EXTRACTION
RELATION
EXTRACTION
DEP
frequenciesallelesrs1048661rs2165241 one factors led prevalence exfoliation_syndromeat
Least Common Subsumer
We know:
• Token
• Stem
• Lemma
• Role  NEW!!!
• Dependency Type
• Dependency Node
frequencies oneat prep_of nsubj
V-WalkE-Walk
Workflow: Gene-Disease
PRE
PROCESSING
ENTITY
EXTRACTION
RELATION
EXTRACTION
DEP
frequenciesallelesrs1048661rs2165241 one factors led prevalence exfoliation_syndromeat
Least Common Subsumer
We know:
• Token
• Stem
• Lemma
• Role  NEW!!!
• Dependency Type
• Dependency Node
frequencies oneat prep_of nsubj
V-WalkE-Walk
Workflow: Gene-Disease
PRE
PROCESSING
ENTITY
EXTRACTION
RELATION
EXTRACTION
DEP
frequenciesallelesrs1048661rs2165241 one factors led prevalence exfoliation_syndromeat
Least Common Subsumer
We know:
• Token
• Stem
• Lemma
• Role  NEW!!!
• Dependency Type
• Dependency Node
frequencies oneat prep_of nsubj
V-WalkE-Walk
Workflow: Gene-Disease
PRE
PROCESSING
ENTITY
EXTRACTION
RELATION
EXTRACTION
DEP
frequenciesallelesrs1048661rs2165241 one factors led prevalence exfoliation_syndromeat
Least Common Subsumer
We know:
• Token
• Stem
• Lemma
• Role  NEW!!!
• Dependency Type
• Dependency Node
frequencies oneat prep_of nsubj
V-WalkE-Walk
POSi_dep_POSi+1
Lemmai_dep_Lemmai+1
Stemi_dep_Stemi+1
Rolei_dep_Rolei+1
Tokeni_dep_Tokeni+1
dep_POSi_dep
dep_Lemmai_dep
dep_Stemi_dep
dep_Rolei_dep
dep_Tokeni_dep
Workflow: Gene-Disease
PRE
PROCESSING
ENTITY
EXTRACTION
RELATION
EXTRACTION
DEP
frequenciesallelesrs1048661rs2165241 one factors led prevalence exfoliation_syndromeat
Least Common Subsumer
We know:
• Token
• Stem
• Lemma
• Role  NEW!!!
• Dependency Type
• Dependency Node
frequencies oneat prep_of nsubj
V-WalkE-Walk
POSi_dep_POSi+1
Lemmai_dep_Lemmai+1
Stemi_dep_Stemi+1
Rolei_dep_Rolei+1
Tokeni_dep_Tokeni+1
dep_POSi_dep
dep_Lemmai_dep
dep_Stemi_dep
dep_Rolei_dep
dep_Tokeni_dep
Workflow
PRE
PROCESSING
ENTITY
EXTRACTION
RELATION
EXTRACTION
DEP
Annotated Corpus
Drug-DiseaseDrug-Gene Gene-Disease
• Positive Association
• Negative Association
• Speculative Association
Global + Local
Global + Local + Dependencies
Workflow
PRE
PROCESSING
ENTITY
EXTRACTION
RELATION
EXTRACTION
DEP
Annotated Corpus
Drug-DiseaseDrug-Gene Gene-Disease
• Positive Association
• Negative Association
• Speculative Association
Global + Local
Global + Local + Dependencies
Corpus Kernel V-walk E-walk Precision Recall F1
Drug-Gene
GL + LC - - 0.68 0.72 0.69
GL + LC + DEP Token - 0.70 0.73 (0.81) 0.70 (0.72)
Drug-Disease
GL + LC - - 0.83 0.76 0.79
GL + LC + DEP Token - 0.81 (0.83) 0.83 (0.85) 0.82 (0.72)
Target-Disease
GL + LC - - 0.73 0.69 0.70
GL + LC + DEP Token - 0.73 (0.76) 0.77 (0.78) 0.74 (0.75)
AIMed
GL + LC - - 0.53 0.66 0.58
GL + LC + DEP - Token 0.57 (0.57) 0.62 (0.69) 0.59 (0.60)
Workflow
PRE
PROCESSING
ENTITY
EXTRACTION
RELATION
EXTRACTION
DEP
10-fold Cross Validation
Example
PRE PROCESSING
JSRE
NER
Diseases
Genes
GAD
Example
PRE PROCESSING
JSRE
NER
Diseases
Genes
GAD
Test Results
Example
PRE PROCESSING
JSRE
NER
Diseases
Genes
GAD
Test Results
Prediction Results
Relation Extraction

Relation Extraction