vertex search
TRANSCRIPT
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 1/90
!""#$ %& !""#
HLT/NAACL 2006 2
–
– ! !
– "
# ! $% &&
– $ # % % '
– ( # )$ % * % * +%' *% *
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 2/90
HLT/NAACL 2006 3
'
$ – ,% -*%
– .-'% % /
– )$ * /
01(2
!
– % $3
– !% + /
HLT/NAACL 2006 4
& ()&
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 3/90
HLT/NAACL 2006 5
'
-*
HLT/NAACL 2006 6
* +
( 4 5 %1 #
5 #
1# 5
( 4 %
1*#
a b
c
d e
V= {a,b,c,d,e}
E= {(a,b),(a,c),(a,d),(b,e),(c,d),(c,e),
(d,e)}
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 4/90
HLT/NAACL 2006 7
,- .& & /
6% 7 % – 6 7 3
– , 6% 7 6 7
8 6% 79
– 6 3 7% 7 3 6
– , 8 6% 79 6 7
, * $ *
: – * $
– * $
HLT/NAACL 2006 8
0
1 2
3 4 5 6
G2
3 2
3 3
1 1 1 1
directed graph
in-degree
out-degree
0
1
2
G3
in:1, out: 1
in: 1, out: 2
in: 1, out: 0
0
1 2
3
33
3
0(
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 5/90
HLT/NAACL 2006 9
,-
# ; 7% <%& & & ' =7 3&
3
3 3
3
2
a b
c
d e
a b
c
d e
a b e d c b e d c
HLT/NAACL 2006 10
,- 1 & ,
#
# % * * *
a b
c
d e
b e c
a b
c
d e
a c d a
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 6/90
HLT/NAACL 2006 11
,- & 1
$# $
# * $& 1&&% $ > &
connected not connected
• connected graph: any two vertices are connected by some path
HLT/NAACL 2006 12
/ 2
( % 6% 7 4 7% 6
( % 8 6% 79 ?4 8 7% 69
tail head
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 7/90
HLT/NAACL 2006 13
* 2)
1 -
weighted unweighted
4
8 13
5
HLT/NAACL 2006 14
345- ., 6(
45%1 $ &
, 3 * $ % 3@
% 3 1% 3@ABA3B47
1% 3@ABA3B46
, 3 * C 3 * $
!# : 3 %
.#
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 8/90
HLT/NAACL 2006 15
0
1
1
1
1
0
1
1
1
1
0
1
1
1
1
0
0
1
0
1
0
0
0
1
0
0
1
1
0
0
0
0
0
1
0
0
1
0
0
0
0
1
0
0
1
0
0
0
0
0
1
1
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
1
0
1
0
0
0
0
0
0
1
0
1
0
0
0
0
0
0
1
0
G1
G2
G3
0
1 2
3
0
1
2
1
0
2
3
4
5
6
7
symmetric
undirected: n2 /2
directed: n2
Examples of Adjacency Matrices
HLT/NAACL 2006 16
3!5- .,
1 3 *
• ( 49 <
!#
–
.#
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 9/90
HLT/NAACL 2006 17
0
1
2
3
0
1
2
0
1
2
3
4
56
7
1 2 3
0 2 3
0 1 3
0 1 2
G1
1
0 2
G3
1 2
0 3
0 3
1 2
5
4 65 7
6
G4
0
1 2
3
0
1
2
1
0
2
3
4
5
6
7
0( .,
HLT/NAACL 2006 18
'
-*
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 10/90
HLT/NAACL 2006 19
,
/ :
– " $ %
0 :
– %
HLT/NAACL 2006 20
M N O P
I J K L
E F G H
A B C D
/7 1
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 11/90
HLT/NAACL 2006 21
/7 1
/713
'
/713) 5
HLT/NAACL 2006 22
87 1
M N O P
I J K L
E F G H
A B C D
0
M N O P
I J K L
E F G H
A B C D
0 1
M N O P
I J K L
E F G H
A C DB
0 1 2
M N O P
I J K L
E F G H
A B C D
0 1 2 3
d) c)
b) a)
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 12/90
HLT/NAACL 2006 23
87135-
- (
'- 9,: 9:
; " (
←←←← "
) ,
<4 , ,
(
)
( ) (
,
) <4
←←←← < 4
87 1
HLT/NAACL 2006 24
7
: * *
D
– $
D E
D $ E
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 13/90
HLT/NAACL 2006 25
7 ) / 7 1
E
F
G
B
CD
A start
destination
A DFS on A A
DFS on BBA
DFS on C
B
C
AB Return to call on B
D Call DFS on D
A
BD
Call DFS on GG found destination - done
path is implicitly stored in DFS recursionpath is: A, B, D, G
HLT/NAACL 2006 26
E
F
G
B
CD
Astart
destination
A
Initial call to BFS on A
Add A to queue
B
Dequeue A
Add B
frontrear frontrear
C
Dequeue B
Add C, D
frontrear
D D
Dequeue C
Nothing to add
frontrear
G
Dequeue D
Add G
frontrear
found destination - donepath must be stored separately
7 ) 8 7 1
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 14/90
HLT/NAACL 2006 27
'
-*
HLT/NAACL 2006 28
6 1
( % 45%1
– $
–
– &
45%1%)% $
– $
(
– &
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 15/90
HLT/NAACL 2006 29
= 6 1
$ * 0 $
–
–
$
– *
5 3 #
– , # %
– : # % $ 3 * %
– D #
HLT/NAACL 2006 30
=
( !,
+
$ *
3
)
$ * *
C
3
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 16/90
HLT/NAACL 2006 31
=
v1 v2
v3 v4 v5
v6 v7
2
14
1
4
10
6
72
5 8
3
v1 v2
v3 v4 v5
v6 v7
2
14
1
4
10
6
72
5 8
3
v1 v2
v3 v4 v5
v6 v7
2
14
1
4
10
6
72
5 8
3
v1 v2
v3 v4 v5
v6 v7
2
14
1
4
10
6
72
5 8
3
HLT/NAACL 2006 32
=
v1 v2
v3 v4 v5
v6 v7
2
14
1
4
10
6
72
5 8
3
v1 v2
v3 v4 v5
v6 v7
2
14
1
4
10
6
72
5 8
3
v1 v2
v3 v4 v5
v6 v7
2
14
1
4
10
6
72
5 8
3
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 17/90
HLT/NAACL 2006 33
6 1
/# – $ % 45%1%)% , $
& ,#
,%
*
– , ,
(
# % 4 5% 1% )% ,7 ,< ,%
, $ F5F7
HLT/NAACL 2006 34
'
-*
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 18/90
HLT/NAACL 2006 35
7) )>
) * G
1*# – ! ' &&% %
1 E % E 3
– / ' 1 E ' % E 3
. # – *
– * – : E
–
HLT/NAACL 2006 36
7;
: ' E
– /% ≥ 6
– , #
– : * →→→→
: E # × →
– #∀ ∈ # ≤
– # ∀ ∈∈∈∈ # 4 E – ∀ ∈∈∈∈ E H I#
∈
∈
v V
v V
, o f(u,v)= f(u,V)= 0
f(v,u)= f(V
r
,u)=0
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 19/90
HLT/NAACL 2006 37
)
/ $ $ G
? $ &
13
11
54
15
10
14
19
3
s t 9
a b
c d
2 5
3
3
6
6
13
11
54
15
10
14
19
3
s t 9
a b
c d
3
3
3
6
6
HLT/NAACL 2006 38
6( )
) *+
#∈
v V
f = f(s,v)= f(s,V)= f(V,t)
Find a flow of maximum value.
13
11
54
15
10
14
19
3
s t 9
a b
c d
5
13
3
8
6
10
8
2
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 20/90
HLT/NAACL 2006 39
# – %
! 9 6% # = ≤≤≤≤
– ' % $
( G
13
11
54
1510
14
19
3
s t 9
a b
c d
5
13
3
8
6
108
2
HLT/NAACL 2006 40
77>
Ford-Fulkerson(G,s,t)
1 initialize flow f to 0 everywhere
2 while there is an augmenting path p do
3 augment flow f along p
4 return f
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 21/90
HLT/NAACL 2006 41
)>
J G
" #
– # 4 E
– '# 4 % 4 H ∈∈∈∈ ×××× # 9 6I
Compute residual network:
13
11
54
15
10
14
19
3
s t 9
a b
c d
5
13
3
8
6
108
2
HLT/NAACL 2006 42
,
J G
" #
4 H # ∈ I
/ #
– ∀ % $
– &
13
11
54
15
10
14
19
3
s t 9
a b
c d
5
13
3
8
6
108
2What is the residual capacity of the
path (s,a,b,t)?
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 22/90
HLT/NAACL 2006 43
77> & )
Ford-Fulkerson(G,s,t)
1 for each edge (u,v)∈G.E do
2 f(u,v) = f(v,u) = 0
3 while ∃ path p from s to t in residual network Gfdo
4 cf= min{c
f(u,v): (u,v)∈p}
5 for each edge (u,v) in p do
6 f(u,v) = f(u,v) + cf
7 f(v,u) = -f(u,v)
8 return f
, $ >&
HLT/NAACL 2006 44
77> & (
13
11
54
15
10
14
19
3
s t
a b
c d
3
3
3
13
11
54
15
10
14
19
3
s t
a b
c d
13
3
3
3
13
13
13
11
54
15
10
14
19
3
s t
a b
c d
18
3
8
8
13
13
5
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 23/90
HLT/NAACL 2006 45
( % , ∈
% ∈
,& – , % , #
# &
∈∈∈∈∈∈∈∈∈∈∈∈
====
T w S v
E w v S e
w v u e u
,),(ofout
).,(:)(
s
2
3
4
5
6
7
t
15
5
30
15
10
8
15
9
6 10
10
10154
4
Capacity = 30
HLT/NAACL 2006 46
s
2
3
4
5
6
7
t
15
5
30
15
10
8
15
9
6 10
10
10154
4
Capacity = 62
s
2
3
4
5
6
7
t
15
5
30
15
10
8
15
9
6 10
10
10154
4
Capacity = 28
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 24/90
HLT/NAACL 2006 47
7)
10
6
6
10
10
0 10
4 8 8
0
4 0
00
s
2
3
4
5
6
7
t
15
5
30
15
10
8
15
9
6 10
10
10154
4
Value = 24
7& $ % % , $ & ,%
; &
f e f e f e f
s e S e S e
========−−−− ofouttoinofout
)()()(
HLT/NAACL 2006 48
6( 7) 6
.& $ % % , $ & FF 4% ,% * % , &
* - ::'# '% * ; &
s
2
3
4
5
6
7
t
15
5
30
15
10
8
15
9
6 10
10
10154
4
10
9
9
15
15
4 10
4 8 9
10 0
Flow value = 28
00
Cut capacity = 28
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 25/90
HLT/NAACL 2006 49
'
-*
HLT/NAACL 2006 50
, 45%1 J4)%: $3 # 5 → ) % 5# – HI ∈1 ⇔ H%I ∈ :
! "#
! $ → → % !
!&" $ → – " $ ' ( ∈ ⇔' ( ∈
– " $ " %
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 26/90
HLT/NAACL 2006 51
1
( 45%1 $ J4)%: $ J7 J J7
HLT/NAACL 2006 52
?(
F5F ≠ F)F% F1F ≠ F:F% # & /&
"% J %
– 1 J
– J% $ % &
: # $ % &
– % ;
* &
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 27/90
HLT/NAACL 2006 53
1
; 4 ( 7
$ ( 7
45% ; J 4 07 $ J 074)&
– ( $ ( 7%K%( 0 J $ 0 7%K%0&
( 7%K%( (
0 7%K%0 0
: # ( # 0#
– . * ( 0 $ ( # 0 # % # &
– ( 4 H ( F $ ( # I
– 0 4 H 0 F $ 0 # I
– F( F ≠ F0 F% #
HLT/NAACL 2006 54
,>& &
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 28/90
HLT/NAACL 2006 55
'
. J
1
'
HLT/NAACL 2006 56
,
. * & $ $
, .5#
– / # $ 5L $
– / # 5
– 0 # 5
– 1 # ' ' 5
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 29/90
HLT/NAACL 2006 57
>
, ' !' ! &
$ &
( L $&
( M N# % #
– 1 ' ' – " 3
HLT/NAACL 2006 58
)>
/'L '# – 6
) $& O $ 6G
J #
– !6 4 6 – ! 4 7
– !* 4 PQ*7= PQ*=7% 68*8
– % P $ '
0 1 2 3 4 5
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 30/90
HLT/NAACL 2006 59
/
/$ & " 46%
# 47&
L ;#
, ; #
% ; &
02=+=∇ yy xx uuu
t uuk =∇2
(**)
HLT/NAACL 2006 60
, *
– / *&
– ( * $ &
– ( $ &
– (3 $ $&
– &
.
– ! ' &
– . $$ ' * &
1
– ' $ '
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 31/90
HLT/NAACL 2006 61
0
( MN *
+% $ $ * $
. #
0)det( =− I A λ
vv A
λ =
HLT/NAACL 2006 62
0
1*#
/ (λ 4 7λQλ>Q<46
,# λ=λ<R46C λ74<C λ<4>
: λ1=2:
# *74*<
−=
02
31 A
−
−−=−
λ
λ λ
2
31 I A
022
33
2
1
=
−
−
x
x
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 32/90
HLT/NAACL 2006 63
1
# 7 6& 1*#
, * # $ 7 4 7&
: $ 7% % 4
% ,
4 &
=
43
41
85
83
A
HLT/NAACL 2006 64
0 )> )>
= y
xy xC C
1 ' * %
11
1 0.5
0.5a b
c
d
xy
xy R
C 1
=
x
xy
xyC
C P =
05
2
5
2
5
12
1
04
1
4
13
2
3
100
2
1
2
100
d cba
a
b
c
d
w=Pw T
14
514
414
314
2
From Doyle and Snell 2000
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 33/90
HLT/NAACL 2006 65
0 )> )>
xy y x
xy
y x
xy C vv R
vvi )( −=
−
=
y
y
xy y
y x
xy
x vPvc
cv ==
11
1 0.5
0.5a
c
d
1 V
b
= y xyi 0
0
1
=
=
b
a
v
v
$$ '
$ &&
, ' . &
8
3
5
2
5
116
7
2
1
4
1
=+=
=+=
cd
d c
vv
vv
HLT/NAACL 2006 66
0,
,
" %
, * #
.
yx xy f f −=
xy xy Ri2
= y x
xy xy Ri E ,
2
2
1
ba,yfor,0 ≠= y
xy f
−=− y x
xy y xaba jww jww,
)(2
1)(
(**)
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 34/90
HLT/NAACL 2006 67
=
" #
, $ $46 47 & , $ $,L !
= y x
xy xyeff xy Ri Ri,
22
2
1
(**)
HLT/NAACL 2006 68
6>
1
2
34
5
7
6 8
S
T
77R7777O
7U
77>
7<
777
STROU><7
Transition matrixGraph G (V,E )
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 35/90
HLT/NAACL 2006 69
6>
( ' $ $ ' * &
! 4 ; '% (% K% & ) 4 *(Q
, $$ $ $$ &
' 4 ) # '% % # &
HLT/NAACL 2006 70
1,
, 1 , ' A +' 7VSVB ' $ #
–
– $
–
, ' #
– ( 7
– '
– ' $
1*# !' A0 ! 7VVSB# MN
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 36/90
HLT/NAACL 2006 71
1
2
34
5
7
6 8
This graph G has a second graph G’
(not drawn) superimposed on it:
G’ is a uniform transition graph.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1 2 3 4 5 6 7 8
P a g e R a n k
t=0
00.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1 2 3 4 5 6 7 8
P a g e R a n k
t=1
(
HLT/NAACL 2006 72
,
0)( =−
=
pG I
pG p
T
T
function PowerStatDist (G):
begin
p(0) = u; (or p(0) = [1,0,…0])
i=1;
repeat
p(i) = E T p(i-1)
L = || p(i)-p(i-1)||1;
i = i + 1;
until L < ε
return p(i)
end
Solution for the
stationary distribution
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 37/90
HLT/NAACL 2006 73
1
2
34
5
7
6 8 0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1 2 3 4 5 6 7 8
P a g e R a n k t=0
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1 2 3 4 5 6 7 8
P a g e R a n k
t=1
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1 2 3 4 5 6 7 8
P a g e R a n k
t=10
0(
HLT/NAACL 2006 74
*1 (
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 38/90
HLT/NAACL 2006 75
'
!' J,
HLT/NAACL 2006 76
;% ' $ ;&
" )$% ; 4 < &
W E + ' )$G
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 39/90
HLT/NAACL 2006 77
>
/ $ &
;% ; *&
% * * &
( L ;% !'&
HLT/NAACL 2006 78
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 40/90
HLT/NAACL 2006 79
HLT/NAACL 2006 80
6 >
!$
– !' &
– ( ' &
– ( M$N & *% M$N&
– , ' $&
– ' $ * ' $ &
– !+ !' $ '
– ,$ J <66<# /"X !' &
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 41/90
HLT/NAACL 2006 81
1
J* * & / $ Y 2$
0 ( .151 &
J, ;&
J$ % && $'' $ & $ &
Honda
Ford
VW
Car and Driver
HLT/NAACL 2006 82
1
1 ' !& ! &
$&
1* ; M3N#
– &&
– 3&&
– &--K , Y
– &&&-Z-3-3$'& – &&-3;-3;&
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 42/90
HLT/NAACL 2006 83
1
J, # – $
;
– * $ + 7666O666
– $
– ' $
1 #
hGa T =
'
Gah ='
)( GGaT
ω = )( T GGh ω = )(G p =
HLT/NAACL 2006 84
1
J, $ ('& ,& &
$ &&% $ &
1* M3N – ! $
– , $ $
– , $ &
1* M$N
– , M N M N
– , MN&
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 43/90
HLT/NAACL 2006 85
1
, * *&
, $ $ &
% $ *&
HLT/NAACL 2006 86
(
, * *&
4 / E % / * &
1*#
A
B
C
G
F
E
D 666677
666767!667766"
667#776
6777#66$
76676#7%
776667# &
!"$% &
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 44/90
HLT/NAACL 2006 87
7
, : &
A
B
C
G
F
E
D 6&O>UU
6&6666
6&O>UU
6&<S6S
6&>RS<
6&<S6S
6&>RS<
.7
G!
.<"
.<
.<$
.7%
.7 &
HLT/NAACL 2006 88
1
. λ<
. <
:
– < 8 6
( .7
– 1 < 9 6
( .<
– 1 < 4 6 ( .7 .<
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 45/90
HLT/NAACL 2006 89
?
1
HLT/NAACL 2006 90
'
$
$
-
*
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 46/90
HLT/NAACL 2006 91
* 1 /
, $ $&
– &
– 2 % % $
)
/
–Ex: “chair” – furniture or person
–Ex: “child” – young person or human offspring
“Sit on a chair” “Take a seat on this chair”“The chair of the Math Department” “The chair of the meeting”
HLT/NAACL 2006 92
1 *1/
D - '
0 MN
– $ % $ *
/$ +, (---. +, (---.
/$ +" (-/-.carnivore
wild dogwolf
bearfeline, felidcanine, canidfissiped mamal, fissiped
dachshund
hunting doghyena dogdingo
hyenadog
terrier
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 47/90
HLT/NAACL 2006 93
1 1, 6
# "# 1&& +, " !" (--/.
– 1&& % 4 6&R6 % $ 46&U<
" # – + (--0. +,
(--/.
– $ +1! " 1" (---.
– . $ * +2 " (--0.
−=
D
C C PathC C Similarity
2
),(log),( 21
21
where D is the taxonomy depth
HLT/NAACL 2006 94
( *1/
( $ *
* +3 " *4 (-//. +3" " 3 (-56.
M 2 ! 7 "" !! " & ! " ! ! "N
( * #
7& *& , % &
<& : % % % $ $ % &
>& % C % &
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 48/90
HLT/NAACL 2006 95
(
A very long train traveling along the rails with a constant velocity v in a
certain direction …
train #1: public transport
#2: order set of things
#3: piece of cloth
travel
#1 change location
#2: undergo transportation
rail #1: a barrier
# 2: a bar of steel for trains
#3: a small bird
HLT/NAACL 2006 96
( *1/
* *
– D
$ $ *
1#
– + " 18 9'':. * TU . * R<&6V[
– +1! " 1" 9'''. . * V6[ R6[
* MN
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 49/90
HLT/NAACL 2006 97
> 1 )>
# $ *
# ( *
"# – MN
–
* – +3" " 3 (-5-.
– : % M'N
$ '
HLT/NAACL 2006 98
Two U.S. soldiers and an unknown number of civilian
contractors are unaccounted for after a fuel convoy was
attacked near the Baghdad International Airport today,
a senior Pentagon official said. One U.S. soldier and an
Iraqi driver were killed in the incident.…
…
…
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 50/90
HLT/NAACL 2006 99
Two U.S. soldiers and an unknown number of civilian
contractors are unaccounted for after a fuel convoy was
attacked near the Baghdad International Airport today,
a senior Pentagon official said. One U.S. soldier and an
Iraqi driver were killed in the incident.…
…
…
HLT/NAACL 2006 100
Two U.S. soldiers and an unknown number of civilian
contractors are unaccounted for after a fuel convoy was
attacked near the Baghdad International Airport today,
a senior Pentagon official said. One U.S. soldier and an
Iraqi driver were killed in the incident.…
…
…
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 51/90
HLT/NAACL 2006 101
Two U.S. soldiers and an unknown number of civilian
contractors are unaccounted for after a fuel convoy was
attacked near the Baghdad International Airport today,
a senior Pentagon official said. One U.S. soldier and an
Iraqi driver were killed in the incident.…
…
…
HLT/NAACL 2006 102
6 1
(# !
– % * '+% %+
<# ( $ *
–
:# / $
;# ( $ '
– !'% J,% !
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 52/90
HLT/NAACL 2006 103
1
$ ) – ( -
– !(,": -
–
– $
– +
– '
1 – -
– $
"# * – MN
–
HLT/NAACL 2006 104
* 1 /
' -
D $ * +1! < 9'';. +1! 9''0.
# – '
– 0 ; -
J$# – $ ' = '
– $ ' = ;
1 – MN
– MDN /
– < *% + R66
– . $ *# % % $% %
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 53/90
HLT/NAACL 2006 105
0
Size(words) Random Lesk TextRank TextRank+Lesk
SemCor
law 825 37.12% 39.62% 46.42% 49.36%
sports 808 29.95% 33.00% 40.59% 46.18%
education 898 37.63% 41.33% 46.88% 52.00%
debates 799 40.17% 42.38% 47.80% 50.52%
en tertainmen t 802 39.27% 43.05% 43.89% 49.31%
Avg. 826 36.82% 39.87% 45.11% 49.47%
Senseval-2
d00 471 28.97% 43.94% 43.94% 47.77%
d01 784 45.47% 52.65% 54.46% 57.39%
d02 514 39.24% 49.61% 54.28% 56.42%
Avg. 590 37.89% 48.73% 50.89% 53.86%
Average (al l) 740 37.22% 43.19% 47.27% 51.16%
“uninformed” (no sense order)
HLT/NAACL 2006 106
0
Size(words) MFS Lesk TextRank TextRank+Lesk
SemCor
law 825 69.09% 72.65% 73.21% 73.97%
sports 808 57.30% 64.21% 68.31% 68.31%
education 898 64.03% 69.33% 71.65% 71.53%
debates 799 66.33% 70.07% 71.14% 71.67%
en tertain men t 802 59.72% 64.98% 6 6.02% 66.16%
Avg. 826 63.24% 68.24% 70.06% 70.32%
Senseval-2
d00 471 51.70% 53.07% 58.17% 57.74%
d01 784 60.80% 64.28% 67.85% 68.11%
d02 514 55.97% 62.84% 63.81% 64.39%
Avg. 590 56.15% 60.06% 63.27% 63.41%Averag e (al l) 740 60.58% 6 5.17% 67.51% 67.72%
“informed” (sense order integrated)
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 54/90
HLT/NAACL 2006 107
'
$ $
-
*
HLT/NAACL 2006 108
0
$ #
– / & Y% / Y% /& Y
– / Y % / Y (, \ ,
$ #
– ) % ) % )
$ $ *#
– ;# && 766
– # && 766
– # && 766 – &#
. $ $
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 55/90
HLT/NAACL 2006 109
/
1* $ E && #
– .
– !
– 5
1 $ $
( #
– 1&& + ,:% ,:-/:
.
+3 < 9''0.
HLT/NAACL 2006 110
1
( '
, #
– /0!# U66%666 E 7U$
– )$$ # 77 MY& N 7O MY& (N% ORT
.
– /$ 4 (A%B $ *
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 56/90
HLT/NAACL 2006 111
/ @
/0! #
)$$ – 77 MY&N# SU&T[ ' TO&U[
– 7O MY&(N# T7&<[ ' RT&<[
HLT/NAACL 2006 112
'
$
$
-
*
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 57/90
HLT/NAACL 2006 113
# D % ' – += " >"!" (---.
: #
– ? # 7V7< )$L /
– 4 # ' $
, $ ) $ +@ < (--:.
HLT/NAACL 2006 114
7& 1*
&& $
<& "$
>& ' ('
$ '
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 58/90
HLT/NAACL 2006 115
3A5
/ " &
! $ # $ ' % % % % % $$ % % % *% &
#
' #
Transport. To carry or bear from one
place to another;…
head word
Transport
to
carry
or
bear
…
7& 1* /
HLT/NAACL 2006 116
3A5
<& "$
Transport
carry
|as| outgoing arcs
e edge
s source node
t target node
re =
ps / |as|
pt
)#
'
'
' FF $
rs,t =
ps / |as|
pt
e=1
mFor more than 1edge (m) betweens and t
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 59/90
HLT/NAACL 2006 117
3A5
&'
# % % %
7&
<& % %3 ' % 3
>& % % '
%
U& ('# %% %
O& ' ( #
H6&V% 6&TO% 6&TO% 6&TO% 6&R% 6&O% K& % 6&7I H7% <% <% <% KI ; ' '
H7% <% <% <% >% KI $ '
3. ' (' • ' $
HLT/NAACL 2006 118
()
VR%S66
77<%SVT
*
87[
86&6O[
86&6O[
6[
( $ )$
(
VV%RU<
7T>%VU7
*
Z6&7[
Z776[ \
7OV%666
T7>%666 $
+ & &.
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 60/90
HLT/NAACL 2006 119
,
, )$L
+,-
)
$ *
MN# * +
$,-
,
, % %
$ %
&& )
HLT/NAACL 2006 120
- 1
( * ;
– 1&&
D $ %
– $ !"
5$
"
(3
K
– $
– $
– +>"" A @ 9''9.
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 61/90
HLT/NAACL 2006 121
( B
( ; . – 1&& A:D,B % $% % K
( – ! * ' $
-
1&& M $N
– % $
1&& M $N% M N% M N% M $N% M $N% M$ $N
– (
–
apple
strawberrybanana
pear
orange
HLT/NAACL 2006 122
0( 0
1 <6 ) – 1&& %
% % K
! S<[ – (
$
*
$
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 62/90
HLT/NAACL 2006 123
'
$ $
-
*
HLT/NAACL 2006 124
( 0
M,* ' % * % * $ *&N +@ < 9''0.
Eyeing the huge market
potential, currently led by
Google, Yahoo took over
search company Overture
Services Inc last year
Yahoo acquired Overture
Applications to:
Information Retrieval (IR)
Comparable Documents (CD)Reading Comprehension (RC)
Question Answering (QA)
Information Extraction (IE)
Machine Translation (MT)Paraphrase Acquisition (PP)
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 63/90
HLT/NAACL 2006 125
( 0
2 ; – #
+% $ % -
BC! &! 4D E 3 B4 &! & C!D
– #
% %
BC! &D E 3 BC! D
– ) '
BF! % D E 3 BF! ((''D
,1 # – ORT % S66
– $ O6[
HLT/NAACL 2006 126
,* $
* %
–
–
' %
– ⇔
Yahoo
bought
Overture Overture
was sold to
Yahoo
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 64/90
HLT/NAACL 2006 127
( 0
( – # O<&U[
– '# TR&O[ $ % >V&O[;
– +%GG < 9''0.
– % '%
– '
–
– OR&S[ – +3!! < 9''0.
HLT/NAACL 2006 128
'
$
$
-
*
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 65/90
HLT/NAACL 2006 129
1., , 1
, $3 $3 * *
D &
– 1&& # &&
$ $% % $
– ; $
HLT/NAACL 2006 130
1., 0(
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 66/90
HLT/NAACL 2006 131
1., /0(
/ $3 * $ $3 $3 *
$3 * * +
$3 *#
– # ; 5% ] 0%
– .*# * % && $3
+% " , 9'';.
HLT/NAACL 2006 132
1.,
; &
.$ ' $ *
( # ( 9 &
# " # + . $ # $
( # + .
$
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 67/90
HLT/NAACL 2006 133
*+ L % % +
: + $# xi
C 1 C 2 + #
( ) ( ) ( )∈
∈∈∈
++
2
121
,12
C xC x
k i
C xC x
k
i
x xassoc xind xind
HLT/NAACL 2006 134
, 2n $ $ n% + $
0 G {v1 ,…vn ,s,t} # – (s,vi) ind 1(xi)
– (vi ,t) ind 2(xi)
– (vi ,vk ) assoc(xi ,xk )
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 68/90
HLT/NAACL 2006 135
3A5
.# #
, S T
(
( $ *
% D - *
',' where
'}{and'}{
T t S s
T t T S sS
∉∉
∪=∪=
HLT/NAACL 2006 136
3A5
$ $ 1
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 69/90
HLT/NAACL 2006 137
1., 0(
( $3 – && $$ $ $3% $
] 0 %
( * – && * 4 $
$
D $3-$3
HLT/NAACL 2006 138
1., 0( ) 6
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 70/90
HLT/NAACL 2006 139
<666 7666 - 7666
, $3 * +
HLT/NAACL 2006 140
• Dolan, William B., L. Vanderwende, and S. Richardson. 1993. AutomaticallyDeriving Structured Knowledge Base from On-line Dictionaries. In Proceedingsof the Pacific Association for Computational Linguistics, April 21-24, 1993,Vancouver, British Columbia.
• Erkan, G. and Radev, D.. "LexRank: Graph-based Lexical Centrality as Saliencein Text Summarization", Journal of Artificial Intelligence Research , 2004
• Galley, M. and McKeown, K.: Improving Word Sense Disambiguation inLexical Chaining. IJCAI 2003.
• Hirst, G. and St-Onge, D. Lexical chains as representations of context in thedetection and correction of malaproprisms. WordNet: An electronic lexicaldatabase, MIT Press, 1998
• Haghighi, A. and A. Ng and C. Manning, Robust Textual Inference via GraphMatching, EMNLP 2005.
• Halliday, M. and Hasan, R. Cohesion in English. Longman, 1976.• Hobbs, J. A model for natural language semantics, Tech. Report, YaleUniversity, 1974.
• Jannink, J. and G. Wiederhold. Thesaurus entry extraction from an on-linedictionary. Fusion 1999.
• Lee, L. Measures of distributional similarity, ACL 1999.
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 71/90
HLT/NAACL 2006 141
• Lesk, M. Automatic sense disambiguation using machine readabledictionaries: How to tell a pine cone from an ice cream cone. SIGDOC 1986.• Lin, D. An Information-Theoretic Definition of Similarity, ICML 1998.• Mihalcea, R. and Moldovan, D. An iterative approach to word sense
disambiguation. FLAIRS 2000.• Mihalcea, R. and Tarau, P and Figa, E. PageRank on Semantic Networks
with Application to Word Sense Disambiguation, COLING 2004• Mihalcea, R. and Tarau, P. TextRank: Bringing Order Into Texts, EMNLP
2004• Mihalcea, R. Graph-based ranking algorithms for large vocabulary word
sense disambiguation, HTL-EMNLP 2005.• Han, H and Zha, H. and Giles, C.L. Name disambiguation in author citations
using a K-way spectral clustering method, ACM/IEEE – IJCDL 2005• Jannink, J. and Wiederhold, G., Thesaurus Entry Extraction from an On-line
Dictionary, Fusion 1999, Sunnyvale CA, 1999.
• Pang, B. and Lee, L., A Sentimental Education: Sentiment Analysis UsingSubjectivity Summarization Based on Minimum Cuts, ACL 2004.• Widdows, D. and Dorow, B., A Graph Model for Unsupervised Lexical
Acquisition, COLING 2002
HLT/NAACL 2006 142
?
1; 1,(
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 72/90
HLT/NAACL 2006 143
'
,* + !
!
/
2 *
HLT/NAACL 2006 144
, ;
1* + ' ' &
# &
, $ ' &&% $ &
, A & <666B&
( G
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 73/90
HLT/NAACL 2006 145
1
1 (d1s1) Iraqi Vice President Taha Yassin Ramadan announced today, Sunday, that Iraq refuses to back down from its decision to stop
cooperating with disarmament inspectors before its demands are met.
2 (d2s1) Iraqi Vice president Taha Yassin Ramadan announced today, Thursday, that Iraq rejects cooperating with the United Nations
except on the issue of lifting the blockade imposed upon it since the year 1990.
3 (d2s2) Ramadan told reporters in Baghdad that "Iraq cannot deal positively with whoever represents the Security Council unless there was
a clear stance on the issue of lifting the blockade off of it.
4 (d2s3) Baghdad had decided late last October to completely cease cooperating with the inspectors of the United Nations Special
Commission (UNSCOM), in charge of disarming Iraq's weapons, and whose work became very limited since the fifth of August, and
announced it will not resume its cooperation with the Commission even if it were subjected to a military operation.
5 (d3s1) The Russian Foreign Minister, Igor Ivanov, warned today, Wednesday against using force against Iraq, which will destroy,
according to him, seven years of difficult diplomatic work and will complicate the regional situation in the area.
6 (d3s2) Ivanov contended that carrying out air strikes against Iraq, who refuses to cooperate with the United Nations inspectors, ``will end
the tremendous work achieved by the international group during the past seven years and will complicate the situation in the region.''
7 (d3s3) Nevertheless, Ivanov stressed that Baghdad must resume working with the Special Commission in charge of disarming the Iraqi
weapons of mass destruction (UNSCOM).
8 (d4s1) The Special Representative of the United Nations Secretary-General in Baghdad, Prakash Shah, announced today, Wednesday,
after meeting with the Iraqi Deputy Prime Minister Tariq Aziz, that Iraq refuses to back down from its decision to cut off cooperation with
the disarmament inspectors.
9 (d5s1) British Prime Minister Tony Blair said today, Sunday, that the crisis between the international community and Iraq ``did not end''
and that Britain is still ``ready, prepared, and able to strike Iraq.''
10 (d5s2) In a gathering with the press held at the Prime Minister's office, Blair contended that the crisis with Iraq ``will not end until Iraq
has absolutely and unconditionally respected its commitments'' towards the United Nations.
11 (d5s3) A spokesman for Tony Blair had indicated that the British Prime Minister gave permission to British Air Force Tornado planes
stationed in Kuwait to join the aerial bombardment against Iraq.
(DUC cluster d1003t)
HLT/NAACL 2006 146
)
7 < $ &
$
, $ $
&
y x
y x
y xni
ii=
=,1
),cos(
, 6 7&
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 74/90
HLT/NAACL 2006 147
( , C0> !""%D
7&666&7<6&>S6&7T6&676&6>6&7S6&676&666&666&6677
6&7<7&666&<R6&<66&666&<66&6U6&6V6&676&7O6&6R76
6&>S6&<R7&666&<O6&666&6U6&<66&666&666&6>6&6RV
6&7T6&<66&<O7&666&6R6&<V6&7O6&7T6&6U6&<76&<SS
6&676&666&666&6R7&666&6O6&6O6&<S6&6>6&6>6&6>T
6&6>6&<66&6U6&<V6&6O7&666&<V6&7R6&676&7V6&<<R
6&7S6&6U6&<66&7O6&6O6&<V7&666&676&666&6>6&6>O
6&676&6V6&666&7T6&<S6&7R6&677&666&6>6&<T6&7TU
6&666&676&666&6U6&6>6&676&666&6>7&666&7R6&6<>
6&666&7O6&6>6&<76&6>6&7V6&6>6&<T6&7R7&666&UO<
6&666&6R6&6R6&<S6&6>6&<<6&6>6&7T6&6<6&UO7&667
7776VSTROU><7
HLT/NAACL 2006 148
d4s1
d1s1
d3s2
d3s1
d2s3
d2s1
d2s2
d5s2
d5s3
d5s1
d3s3
( , 3E"AF5
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 75/90
HLT/NAACL 2006 149
d4s1
d1s1
d3s2
d3s1
d2s3
d2s1
d2s2
d5s2
d5s3
d5s1
d3s3
( , 3E"A!5
HLT/NAACL 2006 150
d4s1
d1s1
d3s2
d3s1
d2s3d3s3
d2s1
d2s2
d5s2
d5s3
d5s1
( , 3E"A45
Sentences vote for the most central sentence!
Need to worry about diversity reranking.
d4s1
d3s2
d2s1
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 76/90
HLT/NAACL 2006 151
N
d TiT E T p
T c
d TiT E T p
T c
d Ti p
nn
n
−+++=
1)()(
)(...)()(
)()( ,,11
1
F (KF ' 2 +F . F " H $ &
" M N% $$ M3N '& &
) " 4 6% $& ) " 4 7%
; & , " $ A6&7%6&<B 0 !%
7VVS&
(>
HLT/NAACL 2006 152
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 77/90
HLT/NAACL 2006 153
HLT/NAACL 2006 154
0( (>
: ' 2 <66O# $$ / 3F/& ( 2 <66R# J, '&
/ ' 1'<66R# ' 7> $ &
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 78/90
HLT/NAACL 2006 155
1*#
Example fromZhu et al. 2003
HLT/NAACL 2006 156
*
.
&
Example fromZhu et al. 2003
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 79/90
HLT/NAACL 2006 157
J &
V x02_join x01_board x0_as x11_director
N x02_is x01_chairman x0_of x11_entitynam
N x02_name x01_director x0_of x11_conglomer
N x02_caus x01_percentag x0_of x11_death
V x02_us x01_crocidolit x0_in x11_filter
V x02_bring x01_attent x0_to x11_problem
Pierre Vinken , 61 years old , will join the board as a nonexecutive director Nov. 29. Mr. Vinken is chairman of
Elsevier N.V. , the Dutch publishing group. Rudolph Agnew , 55 years old and former chairman ofConsolidated Gold Fields PLC , was named a nonexecutive director of this British industrial conglomerate.
A form of asbestos once used to make Kent cigarette filters has caused a high percentage of cancer deathsamong a group of workers exposed to it more than 30 years ago , researchers reported . The asbestos fiber ,crocidolite , is unusually resilient once it enters the lungs , with even brief exposures to it causing symptomsthat show up decades later , researchers said . Lorillard Inc. , the unit of New York-based Loews Corp. thatmakes Kent cigarettes , stopped using crocidolite in its Micronite cigarette filters in 1956 . Althoughpreliminary findings were reported more than a year ago , the latest results appear in today 's New EnglandJournal of Medicine , a forum likely to bring new attention to the problem .
HLT/NAACL 2006 158
, ' !! $ , & <66U&
1*# # M N E* M N&
&
' # 5 % 5 5 $ % % )%
5 % 1* 0!&
1* # ST&OU[ SR&O[ $ X <66U&
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 80/90
HLT/NAACL 2006 159
0(
reported earnings for quarter
reported loss for quarter
posted loss for quarter
posted loss of quarter
posted loss of million
5
G
G
G
*
HLT/NAACL 2006 160
V
N
n1
p
v
reported earnings for quarter
posted loss of million
posted earnings for quarter
n2
, (**)
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 81/90
HLT/NAACL 2006 161
268 (**)
HLT/NAACL 2006 162
This example is slightly modified from the original.
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 82/90
HLT/NAACL 2006 163
HLT/NAACL 2006 164
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 83/90
HLT/NAACL 2006 165
HLT/NAACL 2006 166
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 84/90
HLT/NAACL 2006 167
1
"$ & <66O& $ &
, $ $ &
( $ $ $ % $&
HLT/NAACL 2006 168
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 85/90
HLT/NAACL 2006 169
HLT/NAACL 2006 170
/,
/ & <66O&
1* #
1 3 $
&" &
# ; *
& . 7VRO 1 7VRT
, &
hit
John ball with
the bat
the
root
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 86/90
HLT/NAACL 2006 171
/,
. MY N & , .1 ,
& , 3&
9
10
30
20
300
11
3
9
root
John
saw
Mary
10
30
30
11
root
John
saw
Mary
HLT/NAACL 2006 172
G,) 0(
*
A \ , <66UB
2
– ( *
– , *
– ) # %,* +% ) /$
! ' –
– A, 7VVVB% ] 0A:' 7VVVB% AJ <66>B
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 87/90
HLT/NAACL 2006 173
(> G,) 0(
D
' *
!' '
5#
– '
– '
– ' = 3
HLT/NAACL 2006 174
0(
Compatibility of systems of linear constraints over the set of natural numbers
Criteria of compatibility of a system of linear Diophantine equations, strictinequations, and nonstrict inequations are considered. Upper bounds for
components of a minimal set of solutions and algorithms of construction of
minimal generating sets of solutions for all types of systems are given.
These criteria and the corresponding algorithms for constructing a minimal
supporting set of solutions can be used in solving all the considered types of
systems and systems of mixed types.
Keywords by TextRank: linear constraints, linear diophantine equations,natural numbers, non-strict inequations, strict inequations, upper bounds
Keywords by human annotators: linear constraints, linear diophantine
equations, non-strict inequations, set of natural numbers, strict inequations,
upper bounds
types
equationssolutions
sets
bounds
numbersnatural
criteria
compatibility
minimal
algorithms
construction components
upper
constraints diophantine
systemlinear
strict
inequations
non-strict
systems
TextRank
numbers (1.46)
inequations (1.45)
linear (1.29)
diophantine (1.28)
upper (0.99)
bounds (0.99)
strict (0.77)
Frequency
systems (4)
types (4)solutions (3)
minimal (3)
linear (2)
inequations (2)
algorit hms (2)
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 88/90
HLT/NAACL 2006 175
0
• Evaluation:
– 500 INSPEC abstracts
– collection previously used in keyphrase extraction[Hulth 2003]
• Various settings. Here:
– nouns and adjectives
– select top N/3
• Evaluation in previous work
– mostly supervised learning – training/development/test : 1000/500/500 abstracts
HLT/NAACL 2006 176
Assigned CorrectMethod Total Mean Total Mean Precision Recall F-measure
Ngram with tag 7,815 15.6 1,973 3.9 25.2 51.7 33.9NP-chunks with tag 4,788 9.6 1,421 2.8 29.7 37.2 33Pattern with tag 7,012 14.0 1,523 3.1 21.7 39.9 28.1TextRank 6,784 13.7 2,116 4.2 31.2 43.1 36.2
(Hulth, 2003)
8/13/2019 Vertex Search
http://slidepdf.com/reader/full/vertex-search 89/90
HLT/NAACL 2006 177
0 . <667 E $ D$ /
% ./ <667 E . 0
!% 2//
/ E ' '
1' <66U E *!'# ! / ,*+^% 1!
1' <66R E 0 / . D )'% J,((.
Y <66> E , !%.
2% 2% <66> E % Y.(
2 <66O E !' J'# '
' $ % 2 <66R E ? J, '%
+ $ %
, <66U E ,*'# 0 " ,*% 1!
HLT/NAACL 2006 178
0 <66> E ( / )
+ Y'' <667 E ! $ . ' )'% !
X & <667 E 0 ! / .%.2
X <66< E + 2 1* D ! .%
X% % <66> E D : J :% .
) J$ <66U& : # & , 1 ! Y 0% >S#>>7_>>S
$$##--&&&-Z-$-$$&