This page shows visualizations of some width-4 1-d convolutional filters from Google's lm_1b language model. Each column corresponds to one position in the filter, and shows the characters with the most positive weights. Use the checkbox in the bottom-right to also see the most negative weights (may be slow).
Below that are examples of words for which the filter emits the highest values. A filter's response is its maximum value over all substrings it sees in the word. So if a filter has high weights on 'c' in the first position, then 'a', then 't', it will assign equally high scores to 'cat', 'fatcat', 'concatenate', etc. The portion of the string in blue is the substring the filter is responding to.
'^' and '$' represent beginning and end of word markers, respectively. '_' is a padding character. Literal versions of those characters are escaped with a backslash.
Use the links at the top to see filters of other widths.
Check out my blog post here for a bit more context.
Show most negative weights
Filter 0 (bias = -0.47) #
Z
l
M
r
8
-
c
a
C
t
5
p
6
i
X
k
K
b
F
I
/
T
Q
u
4
q
9
\195
V
E
y
A
G
h
7
o
"
z
\194
v
E
y
R
m
-
p
j
K
r
n
Q
L
I
f
J
,
b
c
9
i
u
/
"
C
(
o
e
A
7
M
\195
h
v
D
s
l
g
x
k
%
W
j
B
-
a
g
K
d
A
G
Q
D
w
0
,
J
N
v
/
c
U
l
X
p
H
u
q
o
I
R
Z
h
r
T
P
e
Z
p
N
z
Q
t
M
l
8
v
R
m
9
i
"
a
r
k
3
x
7
T
X
s
H
f
$
d
4
B
/
A
-
U
c
,
P
Non-zero for 19.0% of words.
^FRAN CISCO$____ (3.0) ^NEW$ ____ (2.7) ^Ear lier$____(2.7) ^Ear th$____(2.7) ^Ear ly$____(2.7) ^Ear l$____(2.7) ^MIAM I$____ (2.7) ^MRI$ ____ (2.6) ^EU$ ____(2.3) ^EBI TDA$____(2.2) ^IBM $____(2.2) ^Rau l$____(2.1) ^CEO$ ____ (2.1) ^CIA$ ____ (2.1) ^DENV ER$____ (2.1) ^Ran gers$____(2.1) ^Ran dy$____(2.1) ^Ran dolph$____(2.1) ^Ran gel$____(2.1) ^RBI $____(2.0) ^RBI s$____(2.0) ^cran e$____ (2.0) ^TEHRAN $____ (2.0) ^EAD S$____(2.0) ^RAF $____(1.9) ^TIME$_ ___ (1.9) ^Eag les$____(1.9) ^Eag le$____(1.9) ^rar e$____(1.9) ^rar ely$____(1.9) ^HMRC$ ____ (1.9) ^10-K$ ____ (1.9) ^5-2$ ____ (1.8) ^RBS $____(1.8) ^FIA$ ____ (1.8) ^Ray $____(1.8) ^Ray mond$____(1.8) ^Ray s$____(1.8) ^Equ ity$____(1.8) ^Jar ed$____(1.7) Filter 1 (bias = -0.50) #
<BOS>
h
w
T
K
t
Z
D
E
d
X
H
W
C
Q
u
2
y
V
l
9
g
<EOS>
c
J
A
b
p
M
r
6
k
3
i
5
1
"
Y
f
j
F
z
x
g
h
G
f
p
S
c
\$
w
B
-
4
T
N
I
6
D
\194
O
8
r
:
P
"
\195
9
d
(
!
H
U
7
C
W
m
,
0
Q
m
R
x
C
a
r
h
j
L
I
B
'
i
"
y
$
l
7
o
F
f
-
p
S
w
K
W
v
A
q
b
H
l
k
L
U
7
T
X
v
5
B
6
u
n
s
.
f
j
c
/
t
D
y
Z
m
4
E
\194
z
8
R
1
b
d
i
Q
r
2
w
H
p
Non-zero for 14.1% of words.
^overwhel ming$____ (2.4) ^overwhel mingly$____ (2.4) ^overwhel med$____ (2.4) ^awful $____ (2.3) ^unlawful $____ (2.3) ^Excl uding$____ (2.2) ^when $____ (1.9) ^when ever$____ (1.9) ^whol e$____ (1.8) ^whol esale$____ (1.8) ^whol ly$____ (1.8) ^excl usive$____ (1.7) ^excl uding$____ (1.7) ^excl usively$____ (1.7) ^excl uded$____ (1.7) ^whil e$____ (1.7) ^Meanwhil e$____ (1.7) ^meanwhil e$____ (1.7) ^whil st$____ (1.7) ^FRA NCISCO$____(1.7) ^whal es$____ (1.6) ^whal e$____ (1.6) ^whal ing$____ (1.6) ^Ful ham$____(1.6) ^Ful l$____(1.6) ^Ful ler$____(1.6) ^Fel ipe$____(1.6) ^Fel ix$____(1.6) ^Fel low$____(1.6) ^FOX $____(1.5) ^FOX News.com$____(1.5) ^worl d$____ (1.5) ^worl dwide$____ (1.5) ^worl ds$____ (1.5) ^worl d-class$____ (1.5) ^Karl $____ (1.5) ^Expl orer$____ (1.5) ^When $____ (1.4) ^NFC$ ____ (1.4) ^careful ly$____ (1.4) Filter 2 (bias = -0.71) #
Z
l
8
-
9
d
Q
t
3
m
N
v
K
j
C
p
4
f
W
.
7
e
2
g
1
T
X
u
"
b
5
z
V
s
R
x
6
i
B
o
f
g
;
G
'
0
-
1
r
L
t
h
k
D
"
5
Q
A
M
c
R
d
N
)
:
4
m
2
O
7
/
6
o
8
W
C
K
Z
(
x
g
f
c
M
A
j
C
-
a
J
G
e
Q
o
z
F
p
u
d
S
1
s
.
t
X
B
Z
E
q
i
7
m
\163
N
Y
K
D
v
0
w
Q
h
P
j
R
g
k
e
r
D
b
4
'
M
z
E
p
5
"
H
K
F
U
L
\195
t
I
u
a
0
X
.
B
i
$
1
Y
S
V
c
Non-zero for 18.9% of words.
^fak e$____(2.8) ^far $____(2.8) ^far mers$____(2.8) ^far m$____(2.8) ^far ms$____(2.8) ^fab ric$____(2.6) ^fab ulous$____(2.6) ^McQ ueen$____(2.3) ^warfar e$____ (2.1) ^rar e$____(2.0) ^rar ely$____(2.0) ^tak e$____(2.0) ^tak en$____(2.0) ^tak ing$____(2.0) ^tak es$____(2.0) ^tar get$____(1.9) ^tar gets$____(1.9) ^tar geted$____(1.9) ^tar geting$____(1.9) ^Mak e$____(1.9) ^Mak ing$____(1.9) ^backgr ound$____ (1.9) ^backgr ounds$____ (1.9) ^Mar ch$____(1.9) ^Mar k$____(1.9) ^Mar tin$____(1.9) ^Mar yland$____(1.9) ^intak e$____ (1.8) ^Braz il$____ (1.8) ^Braz ilian$____ (1.8) ^parliamentar y$____ (1.8) ^documentar y$____ (1.8) ^voluntar y$____ (1.8) ^commentar y$____ (1.8) ^non-pr ofit$____ (1.8) ^rap idly$____(1.8) ^rap e$____(1.8) ^rap id$____(1.8) ^rap ed$____(1.8) ^Ankar a$____ (1.8) Filter 3 (bias = -0.71) #
x
D
V
T
b
y
<BOS>
t
s
c
9
d
S
r
B
u
5
H
<EOS>
M
J
O
4
\163
f
g
6
1
3
.
W
C
7
e
w
h
8
m
a
q
F
g
f
z
M
G
N
w
e
!
j
c
B
a
;
p
S
C
\$
A
r
d
J
i
"
s
K
U
Q
1
P
?
X
.
9
n
8
-
t
&
z
f
A
F
m
x
.
S
Z
h
w
j
g
R
X
r
G
e
/
N
l
9
U
o
L
E
a
u
C
3
c
-
D
J
T
8
I
"
V
4
6
m
7
r
X
M
\194
k
Q
T
8
g
d
U
5
u
x
t
2
o
9
w
W
y
4
c
1
E
q
f
I
s
a
O
$
G
"
R
3
_
Non-zero for 29.7% of words.
^exemp t$____ (3.0) ^exemp tion$____ (3.0) ^Hoffma n$____ (2.7) ^Vega s$____ (2.6) ^basema n$____ (2.5) ^defensema n$____ (2.5) ^bega n$____ (2.4) ^Luxemb ourg$____ (2.4) ^seld om$____ (2.4) ^FA$ ____(2.4) ^NASDAQ $____ (2.4) ^bend $____ (2.3) ^mixed$ ____ (2.3) ^fixed$ ____ (2.3) ^relaxed$ ____ (2.3) ^sewa ge$____ (2.3) ^rebel$ ____ (2.3) ^label$ ____ (2.3) ^Nobel$ ____ (2.3) ^libel$ ____ (2.3) ^beca use$____ (2.2) ^beca me$____ (2.2) ^Quebec$ ____ (2.2) ^Halifax $____ (2.2) ^embedd ed$____ (2.1) ^fema le$____ (2.1) ^fema les$____ (2.1) ^Gonzalez$ ____ (2.1) ^UEFA$ ____ (2.1) ^Martinez$ ____ (2.0) ^send $____ (2.0) ^send ing$____ (2.0) ^send s$____ (2.0) ^Caribbean $____ (2.0) ^bean s$____ (2.0) ^fad ed$____(2.0) ^fad ing$____(2.0) ^fad e$____(2.0) ^halfwa y$____ (2.0) ^Karza i$____ (2.0) Filter 4 (bias = -0.48) #
A
d
B
p
w
-
<BOS>
l
N
D
Q
P
U
j
Z
v
W
h
K
g
/
x
k
J
<EOS>
T
C
y
s
0
,
m
I
e
V
G
S
o
R
1
z
f
&
F
Q
M
I
h
O
y
!
m
G
n
Y
x
:
k
X
j
\194
e
l
H
(
)
a
B
"
i
R
u
\195
\$
E
4
W
,
s
H
-
I
c
A
o
q
s
X
m
W
G
2
u
i
v
a
M
7
f
1
'
4
.
,
z
Q
g
6
y
B
R
5
U
3
x
t
D
V
r
X
C
v
A
W
s
q
y
-
U
J
t
w
S
6
c
b
h
9
,
e
r
2
F
E
k
Q
m
"
l
7
R
\194
L
8
O
0
g
Z
o
Non-zero for 14.4% of words.
^BEIJ ING$____ (2.9) ^waiv er$____ (2.8) ^Giv en$____(2.4) ^Giv e$____(2.4) ^Giv ing$____(2.4) ^Gav in$____(2.4) ^Aziz $____ (2.4) ^Qae da$____(2.3) ^Alab ama$____ (2.3) ^IAE A$____(2.3) ^Cliv e$____ (2.3) ^Xav ier$____(2.2) ^Inv estors$____(2.2) ^Inv estment$____(2.2) ^Inv estigators$____(2.2) ^Inv estigation$____(2.2) ^slav ery$____ (2.2) ^slav e$____ (2.2) ^slav es$____ (2.2) ^liv e$____(2.1) ^liv es$____(2.1) ^liv ing$____(2.1) ^liv ed$____(2.1) ^liv er$____(2.1) ^lav ish$____(2.1) ^III$ ____ (2.1) ^Ponzi$ ____ (2.1) ^influenza$ ____ (2.0) ^Aviv $____ (2.0) ^naiv e$____ (2.0) ^Brav es$____ (2.0) ^Brav o$____ (2.0) ^Riv er$____(2.0) ^Riv era$____(2.0) ^Riv ers$____(2.0) ^Riv erside$____(2.0) ^puzzle $____ (2.0) ^Elizab eth$____ (2.0) ^sizab le$____ (2.0) ^Rav ens$____(2.0) Filter 5 (bias = -0.53) #
.
J
c
i
d
B
y
j
Q
E
'
3
/
4
C
k
"
0
D
5
U
e
u
V
\194
K
m
S
-
9
s
6
^
p
Z
2
a
f
P
o
A
Y
z
"
d
R
U
h
D
x
g
9
m
W
!
S
.
N
L
\194
w
;
I
3
c
8
a
:
C
O
t
(
Z
'
P
M
)
Q
n
Q
m
E
-
8
o
2
f
7
n
9
l
4
p
I
v
X
'
3
y
F
t
6
.
"
i
5
M
Z
g
N
u
U
w
1
c
A
x
B
h
Q
t
X
f
G
F
Y
h
"
n
O
u
r
s
R
m
K
d
V
v
$
x
\195
.
b
y
9
L
W
D
P
A
z
c
J
i
l
j
Non-zero for 14.1% of words.
^SEO UL$____(2.8) ^char ges$____ (2.7) ^char ged$____ (2.7) ^char ge$____ (2.7) ^Richar d$____ (2.7) ^corr uption$____ (2.5) ^corr ect$____ (2.5) ^corr espondent$____ (2.5) ^corr upt$____ (2.5) ^researcher s$____ (2.4) ^teacher s$____ (2.4) ^teacher $____ (2.4) ^researcher $____ (2.4) ^Researcher s$____ (2.4) ^N.Y.$ ____ (2.2) ^D-N.Y.$ ____ (2.2) ^soar ing$____ (2.2) ^soar ed$____ (2.2) ^soar $____ (2.2) ^Yar d$____(2.2) ^NEW $____(2.1) ^Char les$____ (2.1) ^Char lie$____ (2.1) ^Char lotte$____ (2.1) ^Char gers$____ (2.1) ^Jacob$ ____ (2.1) ^1.25$ ____ (2.1) ^98$ ____(2.1) ^Broncos$ ____ (2.1) ^CO2$ ____ (2.1) ^92$ ____(2.0) ^uproar $____ (2.0) ^97$ ____(2.0) ^VEG AS$____(2.0) ^embryos$ ____ (2.0) ^JER USALEM$____(2.0) ^ambassador$ ____ (2.0) ^Ambassador$ ____ (2.0) ^Ecuador$ ____ (2.0) ^Salvador$ ____ (2.0) Filter 6 (bias = -0.43) #
K
t
Z
g
9
l
8
-
B
r
3
p
6
d
2
j
5
.
4
T
N
'
M
h
J
k
U
m
W
I
L
A
X
S
<EOS>
i
w
q
1
\163
y
x
r
5
T
J
m
4
k
w
U
0
P
6
;
E
'
2
?
3
t
9
Q
S
C
j
u
7
"
o
!
e
D
l
H
L
c
B
d
8
H
f
Z
S
I
p
X
F
L
k
q
s
1
x
u
y
\195
t
7
c
Q
m
A
e
Y
M
D
j
a
O
/
'
.
i
R
K
9
o
J
E
W
A
"
h
Q
g
-
D
\194
H
$
C
K
n
'
l
9
c
X
r
E
L
O
y
w
p
M
d
S
t
s
T
f
F
1
_
a
Non-zero for 16.5% of words.
^AZUZ$ ____ (3.2) ^Kraf t$____ (3.0) ^KABUL$ ____ (3.0) ^Brus sels$____ (2.8) ^ATLANTA$ ____ (2.8) ^3rd$ ____ (2.7) ^23rd$ ____ (2.7) ^Braw n$____ (2.7) ^Mr.$ ____ (2.6) ^Jr.$ ____ (2.6) ^WTA$ ____ (2.6) ^UPI$ ____ (2.5) ^Brav es$____ (2.5) ^Brav o$____ (2.5) ^Kyle $____ (2.4) ^ETA$ ____ (2.3) ^UAW $____(2.2) ^Kenya$ ____ (2.2) ^Chechnya$ ____ (2.2) ^NHL$ ____ (2.2) ^Braz il$____ (2.2) ^Braz ilian$____ (2.2) ^FRANCIS CO$____ (2.1) ^THE $____(2.0) ^Peru$ ____ (2.0) ^EPA$ ____ (2.0) ^Kyi$ ____ (2.0) ^Zelaya$ ____ (2.0) ^Lt.$ ____ (2.0) ^MRI$ ____ (1.9) ^3.1$ ____ (1.9) ^LCD$ ____ (1.9) ^frus tration$____ (1.9) ^frus trated$____ (1.9) ^frus trating$____ (1.9) ^Libya$ ____ (1.9) ^3.7$ ____ (1.9) ^DETRO IT$____ (1.9) ^TD$ ____(1.9) ^camera$ ____ (1.9) Filter 7 (bias = -0.46) #
-
x
R
B
u
c
<BOS>
A
r
o
Q
p
j
K
P
a
"
w
E
v
I
h
'
n
H
W
d
l
F
z
M
5
U
i
J
m
s
0
^
,
V
-
k
u
C
.
)
y
B
o
0
E
X
r
Y
?
v
s
b
d
9
N
7
w
6
t
G
O
\194
(
p
e
5
%
x
a
8
U
c
\$
w
p
W
r
Z
T
X
h
Q
g
V
o
.
D
B
c
K
j
6
C
2
d
b
R
/
t
a
0
U
l
4
y
$
O
m
P
G
k
f
z
W
g
M
l
F
R
y
T
6
r
4
u
X
A
,
b
5
G
n
\195
8
Y
K
U
3
k
2
-
Z
v
N
E
/
J
e
D
x
s
Non-zero for 20.2% of words.
^low-key $____ (2.6) ^VW$ ____(2.6) ^Arkan sas$____ (2.6) ^Turkey $____ (2.4) ^turkey $____ (2.4) ^Van $____(2.4) ^Van couver$____(2.4) ^Van essa$____(2.4) ^servan ts$____ (2.2) ^servan t$____ (2.2) ^Cuban $____ (2.1) ^Cuban s$____ (2.1) ^urban $____ (2.1) ^suburban $____ (2.1) ^Urban $____ (2.1) ^V.$ ____(2.0) ^occupan ts$____ (2.0) ^occupan cy$____ (2.0) ^survey $____ (1.9) ^survey ed$____ (1.9) ^survey s$____ (1.9) ^Survey $____ (1.9) ^Harvey $____ (1.9) ^Va$ ____(1.9) ^Milwaukee $____ (1.9) ^Ven ezuela$____(1.9) ^Ven ezuelan$____(1.9) ^Ven us$____(1.9) ^Ven ice$____(1.9) ^dubbe d$____ (1.9) ^rubbe r$____ (1.9) ^SOURCE$ ____ (1.8) ^Mercan tile$____ (1.8) ^0.6 $____(1.8) ^Crawf ord$____ (1.8) ^Secretary-Gen eral$____ (1.8) ^Duke$ ____ (1.8) ^Luke$ ____ (1.8) ^works$ ____ (1.8) ^networks$ ____ (1.8) Filter 8 (bias = -0.38) #
<BOS>
f
X
y
.
U
g
F
Q
P
q
K
-
u
V
s
Y
c
b
D
W
,
\194
C
w
o
l
M
^
B
I
R
k
L
N
r
W
f
T
x
H
r
i
b
U
p
1
F
Y
l
M
n
E
.
4
P
2
'
y
q
O
j
G
-
u
k
)
R
3
N
w
\195
&
\$
D
a
L
-
U
f
A
v
z
c
l
j
Y
M
Q
e
a
k
/
n
X
x
G
'
K
t
Z
p
7
g
6
r
I
h
o
F
q
w
x
k
8
g
"
w
S
T
y
i
\194
-
6
r
5
I
9
t
Q
H
,
j
L
J
F
m
7
b
f
\195
c
A
o
u
N
q
K
V
/
P
Non-zero for 13.5% of words.
^Wax man$____(2.6) ^WAS HINGTON$____(2.5) ^Tax $____(2.5) ^MLS $____(2.4) ^Max $____(2.2) ^pre-tax $____ (2.1) ^WTA$ ____ (2.0) ^Way ne$____(1.9) ^Way $____(1.9) ^Kyrgyzs tan$____ (1.9) ^engulf ed$____ (1.8) ^tax $____(1.8) ^tax es$____(1.8) ^tax payers$____(1.8) ^tax payer$____(1.8) ^Tay lor$____(1.8) ^Wild $____ (1.7) ^Wild life$____ (1.7) ^Wild cats$____ (1.7) ^UAW $____(1.7) ^YOU$ ____ (1.7) ^wild $____ (1.7) ^wild life$____ (1.7) ^wild ly$____ (1.7) ^wild fires$____ (1.7) ^Hay es$____(1.7) ^Hay den$____(1.7) ^Hay ward$____(1.7) ^Wils on$____ (1.7) ^Oly mpic$____(1.7) ^Oly mpics$____(1.7) ^Wac hovia$____(1.7) ^max imum$____(1.6) ^max imize$____(1.6) ^DAL LAS$____(1.6) ^Wad e$____(1.6) ^1,8 00$____(1.6) ^EU$ ____(1.6) ^highly $____ (1.6) ^roughly $____ (1.6) Filter 9 (bias = -0.54) #
o
F
O
k
K
V
<BOS>
b
N
h
/
g
T
H
D
x
z
4
"
s
W
j
-
A
M
i
\194
a
Y
n
,
Z
Q
C
R
u
G
e
r
7
Q
m
O
g
N
v
(
k
I
b
X
u
K
-
&
x
W
)
/
h
,
n
2
c
"
i
E
p
:
V
S
s
3
'
\$
d
5
j
8
f
V
-
F
y
j
o
Z
u
7
d
X
m
A
f
5
s
B
p
0
O
4
i
9
U
J
t
Q
'
8
a
b
v
6
w
L
W
C
z
H
,
h
p
u
K
H
z
R
c
Y
P
k
G
j
d
S
y
F
O
V
w
4
D
B
m
A
o
t
2
\194
a
M
\163
r
f
N
X
s
0
b
L
Non-zero for 13.6% of words.
^shotgu n$____ (2.5) ^NAS A$____(2.3) ^NAS CAR$____(2.3) ^NAS DAQ$____(2.3) ^OF$ ____(2.3) ^TOKY O$____ (2.2) ^foreh and$____ (2.1) ^OFT $____(2.1) ^Och oa$____(2.1) ^DNA$ ____ (2.1) ^Nku nda$____(2.1) ^botch ed$____ (2.1) ^NBA $____(2.0) ^WASHINGTON$ ____ (1.9) ^BOSTON$ ____ (1.9) ^HOUSTON$ ____ (1.9) ^NHS $____(1.9) ^corru ption$____ (1.9) ^corru pt$____ (1.9) ^NYS E$____(1.9) ^NFC $____(1.9) ^LONDON$ ____ (1.9) ^ON$_ ___ (1.8) ^IV$ ____(1.8) ^QC$ ____(1.8) ^FOXNe ws.com$____ (1.8) ^Obs erver$____(1.8) ^downh ill$____ (1.7) ^torch $____ (1.7) ^orch estra$____ (1.7) ^orch estrated$____ (1.7) ^TEHR AN$____ (1.7) ^NCA A$____(1.7) ^NAT O$____(1.7) ^DETROIT$ ____ (1.6) ^MIAM I$____ (1.6) ^DENVE R$____ (1.6) ^Oth er$____(1.6) ^Oth ers$____(1.6) ^Oth erwise$____(1.6) Filter 10 (bias = -0.44) #
<BOS>
p
Z
t
M
T
V
y
-
d
J
i
Q
a
b
h
R
,
X
c
j
A
9
O
<EOS>
1
'
o
.
D
N
q
/
\163
^
l
"
x
8
I
.
P
W
G
\194
J
q
p
(
K
:
f
t
%
Q
r
"
R
!
o
?
\195
v
C
H
j
X
k
d
n
a
)
&
3
h
s
T
i
\$
z
P
g
9
h
8
m
2
k
6
A
X
.
F
t
J
Y
N
c
K
s
e
z
3
i
7
'
Q
l
5
v
I
o
D
x
Z
u
E
T
f
b
W
.
i
r
4
c
2
l
3
x
H
d
w
b
M
L
V
D
X
A
6
a
1
f
5
z
Z
F
Y
o
$
R
J
N
E
u
7
v
t
Non-zero for 15.2% of words.
^Wei ss$____(2.9) ^Wei r$____(2.9) ^Wei nstein$____(2.9) ^Madi son$____ (2.3) ^Hei ghts$____(2.3) ^Hei neken$____(2.3) ^Hei di$____(2.3) ^Bernstei n$____ (2.2) ^7.2$ ____ (2.2) ^Zuri ch$____ (2.2) ^7.6$ ____ (2.2) ^6.8$ ____ (2.2) ^Mae$ ____ (2.1) ^6.2$ ____ (2.1) ^4.9$ ____ (2.1) ^4.8$ ____ (2.1) ^9.5$ ____ (2.1) ^6.6$ ____ (2.1) ^4.2$ ____ (2.1) ^3.9$ ____ (2.1) ^3.8$ ____ (2.1) ^3.2$ ____ (2.1) ^4.6$ ____ (2.1) ^unvei led$____ (2.1) ^unvei l$____ (2.1) ^unvei ling$____ (2.1) ^hei ght$____(2.1) ^hei ghtened$____(2.1) ^hei ghts$____(2.1) ^hei r$____(2.1) ^3.6$ ____ (2.0) ^5.9$ ____ (2.0) ^5.8$ ____ (2.0) ^5.2$ ____ (2.0) ^Mari ne$____ (2.0) ^Mari a$____ (2.0) ^Mari nes$____ (2.0) ^Mari o$____ (2.0) ^5.6$ ____ (2.0) ^outfi t$____ (2.0) Filter 11 (bias = -0.56) #
d
M
I
h
-
c
2
Y
P
B
<BOS>
o
a
y
E
k
6
m
s
K
z
N
l
r
p
b
i
T
7
S
1
G
e
g
J
x
j
A
Q
Z
A
f
U
o
Z
-
E
v
X
x
L
n
Q
t
H
'
!
c
G
j
2
p
a
M
V
h
b
\194
Y
u
z
d
)
,
4
S
I
N
&
k
Q
p
N
l
W
h
X
d
M
m
w
L
"
y
9
A
$
g
Z
s
\194
a
K
i
2
U
3
x
P
F
D
u
C
k
x
T
b
D
.
t
n
y
l
u
L
P
5
O
a
U
V
M
A
d
7
r
B
k
6
c
q
R
X
E
w
\163
9
i
Z
p
4
H
N
C
Non-zero for 16.1% of words.
^dawn $____ (3.3) ^IAEA $____ (3.2) ^long-awa ited$____ (2.7) ^MIAMI $____ (2.7) ^Awa rds$____(2.6) ^Awa rd$____(2.6) ^Awa kening$____(2.6) ^DENV ER$____ (2.6) ^CIA$_ ___ (2.6) ^FIA$_ ___ (2.6) ^PHILADELPHIA$_ ___ (2.6) ^Lex us$____(2.4) ^ISLAMA BAD$____ (2.4) ^lawn $____ (2.3) ^spawn ed$____ (2.2) ^lawl ess$____ (2.2) ^cardboa rd$____ (2.2) ^Delawa re$____ (2.1) ^diox ide$____ (2.1) ^Secretary-Gen eral$____ (2.0) ^NASDAQ$ ____ (2.0) ^Xbox $____ (2.0) ^sidewa lk$____ (2.0) ^Dawn $____ (2.0) ^Anb ar$____(2.0) ^box $____(1.9) ^box es$____(1.9) ^box ing$____(1.9) ^box er$____(1.9) ^GMA C$____(1.9) ^AOL $____(1.9) ^FEMA $____ (1.9) ^AIG$_ ___ (1.9) ^mid-199 0s$____ (1.9) ^anx iety$____(1.9) ^anx ious$____(1.9) ^EPA$_ ___ (1.9) ^PA$_ ___ (1.9) ^UAW$ ____ (1.9) ^DAX$ ____ (1.8) Filter 12 (bias = -0.45) #
w
R
.
P
Z
l
W
J
m
j
g
p
A
h
X
f
c
o
M
d
<BOS>
7
a
9
Q
D
E
S
z
F
y
0
/
i
2
r
e
x
q
,
V
t
G
-
K
d
k
e
Y
u
Z
j
c
l
9
.
C
F
z
H
B
D
b
f
)
\$
X
E
Q
L
8
h
W
%
"
I
R
s
0
y
5
k
8
T
6
h
2
A
3
u
4
a
Z
b
7
t
9
m
X
r
Q
l
M
B
F
U
K
q
j
v
S
H
e
Y
$
z
G
\195
/
i
I
j
a
-
W
M
A
f
t
g
,
F
U
x
T
b
i
m
O
.
z
J
2
e
1
Z
Q
'
H
c
B
r
q
h
p
n
/
V
v
Non-zero for 20.8% of words.
^awkwa rd$____ (3.1) ^Hawks$ ____ (2.1) ^Seahawks$ ____ (2.1) ^downwa rd$____ (2.0) ^GMA C$____(1.9) ^Kea ne$____(1.7) ^ticket $____ (1.7) ^ticket s$____ (1.7) ^rocket $____ (1.7) ^cricket $____ (1.7) ^jacket $____ (1.7) ^rocket s$____ (1.7) ^pocket $____ (1.7) ^pocket s$____ (1.7) ^Yea r$____(1.7) ^Yea h$____(1.7) ^Yea rs$____(1.7) ^1.25$ ____ (1.7) ^Vet erans$____(1.7) ^impea chment$____ (1.7) ^Blackwa ter$____ (1.7) ^backwa rd$____ (1.7) ^backwa rds$____ (1.7) ^two-t hirds$____ (1.7) ^two-t ime$____ (1.7) ^wast e$____ (1.7) ^wast ed$____ (1.7) ^wast ing$____ (1.7) ^nickna me$____ (1.7) ^nickna med$____ (1.7) ^breakfa st$____ (1.7) ^want $____ (1.7) ^want ed$____ (1.7) ^want s$____ (1.7) ^want ing$____ (1.7) ^McNa mee$____ (1.7) ^1.9$_ ___ (1.6) ^2.9$_ ___ (1.6) ^0.9$_ ___ (1.6) ^3.9$_ ___ (1.6) Filter 13 (bias = -0.57) #
O
B
d
f
T
n
G
F
D
b
"
x
z
N
W
k
p
h
Q
J
t
L
\194
A
^
9
g
r
c
R
-
a
\163
Z
y
\195
X
u
Y
V
M
x
t
p
U
d
A
v
Z
6
?
0
C
J
m
o
r
-
k
9
H
7
/
l
T
h
.
5
&
8
Q
a
(
i
R
1
'
2
u
3
c
-
y
j
K
l
G
J
O
u
C
I
U
b
M
H
W
e
D
i
z
h
"
E
/
q
Z
g
T
F
m
R
o
s
Q
r
,
7
8
x
X
u
Q
m
\194
r
7
U
6
k
5
f
4
o
W
b
S
-
$
y
2
\195
8
c
"
z
I
w
R
v
n
s
p
B
Non-zero for 15.7% of words.
^McQ ueen$____(3.5) ^YORK$ ____ (2.9) ^dry$ ____ (2.9) ^laundry$ ____ (2.9) ^Try$ ____ (2.8) ^MyS pace$____(2.8) ^My$ ____(2.8) ^empty$ ____ (2.8) ^pretty$ ____ (2.7) ^Betty$ ____ (2.7) ^petty$ ____ (2.7) ^UK$ ____(2.5) ^Guy$ ____ (2.5) ^Orch estra$____ (2.5) ^McD onald$____(2.3) ^McD onnell$____(2.3) ^country$ ____ (2.3) ^industry$ ____ (2.3) ^try$ ____ (2.3) ^Ministry$ ____ (2.3) ^entry$ ____ (2.3) ^ministry$ ____ (2.3) ^Country$ ____ (2.3) ^Industry$ ____ (2.3) ^GAO$ ____ (2.3) ^my$ ____(2.3) ^angry$ ____ (2.3) ^hungry$ ____ (2.3) ^cry$ ____ (2.2) ^outcry$ ____ (2.2) ^Kentucky$ ____ (2.2) ^lucky$ ____ (2.2) ^Ricky$ ____ (2.2) ^tricky$ ____ (2.2) ^Once $____ (2.2) ^McC ain$____(2.2) ^McC hrystal$____(2.2) ^McC arthy$____(2.2) ^McC onnell$____(2.2) ^OTC$ ____ (2.1) Filter 14 (bias = -0.49) #
X
t
m
v
y
c
Y
z
K
C
Q
s
"
I
b
D
W
u
L
d
M
T
<BOS>
0
f
g
V
k
/
w
Z
n
H
U
^
j
8
A
P
E
N
j
W
p
w
g
K
C
O
k
(
d
o
P
&
F
E
)
Q
l
?
V
"
h
:
v
/
i
a
D
U
n
.
m
B
c
2
0
3
T
X
t
p
F
G
n
Y
u
x
s
b
A
6
j
8
I
v
H
0
w
K
N
"
r
z
f
W
.
V
C
9
M
P
e
\194
k
7
S
J
E
V
y
Q
H
b
t
\194
u
7
o
S
h
z
w
9
q
G
T
R
i
X
p
Y
r
5
-
C
a
j
W
J
O
l
e
s
d
Z
1
6
N
Non-zero for 15.5% of words.
^NYS E$____(3.0) ^NYC $____(2.8) ^NY$ ____(2.5) ^maps $____ (2.5) ^mob$ ____ (2.5) ^robb ery$____ (2.4) ^robb ed$____ (2.4) ^robb ers$____ (2.4) ^robb eries$____ (2.4) ^presumabl y$____ (2.3) ^payabl e$____ (2.3) ^box$ ____ (2.3) ^Xbox$ ____ (2.3) ^crops $____ (2.2) ^drops $____ (2.2) ^Shrops hire$____ (2.2) ^Obs erver$____(2.2) ^map$ ____ (2.2) ^obj ect$____(2.2) ^obj ects$____(2.2) ^obj ective$____(2.2) ^obj ections$____(2.2) ^obj ectives$____(2.2) ^obj ected$____(2.2) ^obl igations$____(2.2) ^obl igation$____(2.2) ^obl iged$____(2.2) ^obs ervers$____(2.1) ^obs erved$____(2.1) ^obs tacles$____(2.1) ^obs ervation$____(2.1) ^obs tacle$____(2.1) ^obs cure$____(2.1) ^obs ession$____(2.1) ^probabl y$____ (2.1) ^probabl e$____ (2.1) ^probl ems$____ (2.1) ^probl em$____ (2.1) ^probl ematic$____ (2.1) ^lobb y$____ (2.1) Filter 15 (bias = -0.45) #
w
F
<BOS>
f
W
P
X
D
Q
x
.
c
-
s
q
C
Y
d
I
j
^
U
H
S
p
y
k
h
L
v
0
B
h
p
Y
d
B
-
N
P
H
z
(
f
W
s
A
c
4
I
L
m
X
'
\194
C
:
g
9
r
8
D
q
t
5
G
6
i
7
\163
Q
k
Y
k
G
F
K
f
y
v
O
j
L
t
M
x
/
I
Z
q
m
n
X
e
W
r
U
-
"
b
$
p
o
d
3
a
s
.
B
H
f
h
K
1
z
4
o
u
B
y
p
g
l
d
O
D
r
7
P
Z
J
6
R
8
s
W
\195
Y
b
X
w
q
k
2
x
"
N
M
S
Non-zero for 7.8% of words.
^BAGH DAD$____ (2.2) ^hyg iene$____(2.1) ^why$ ____ (2.1) ^hyd rogen$____(2.1) ^Why$ ____ (2.0) ^WASH INGTON$____ (2.0) ^24-hou r$____ (1.8) ^two-hou r$____ (1.8) ^half-hou r$____ (1.8) ^in-hou se$____ (1.8) ^Loh an$____(1.8) ^Hyu ndai$____(1.7) ^NY$ ____(1.7) ^Hyd e$____(1.7) ^hou se$____(1.7) ^hou rs$____(1.7) ^hou sing$____(1.7) ^hou r$____(1.7) ^who$ ____ (1.7) ^BEIJING$ ____ (1.6) ^You $____(1.6) ^You ng$____(1.6) ^You r$____(1.6) ^You Tube$____(1.6) ^WHO$ ____ (1.6) ^Who$ ____ (1.6) ^throughou t$____ (1.5) ^Throughou t$____ (1.5) ^Bou levard$____(1.5) ^Bou rnemouth$____(1.5) ^Boy le$____(1.5) ^Boy $____(1.5) ^Boy s$____(1.5) ^Boy d$____(1.5) ^NYC $____(1.5) ^Amy $____(1.5) ^Nou ri$____(1.5) ^Bod y$____(1.5) ^WAM$ ____ (1.4) ^BMW $____(1.4) Filter 16 (bias = -0.51) #
V
w
7
f
j
o
Y
a
C
t
Z
p
4
-
F
y
8
W
G
v
R
O
6
m
X
K
H
q
5
z
L
.
g
x
0
i
Q
,
9
N
w
L
W
D
I
h
q
y
-
%
.
G
v
Y
!
l
Q
P
?
U
t
R
a
u
2
J
:
C
X
o
(
m
e
F
'
r
E
T
\194
H
L
k
K
-
X
i
Z
h
/
R
D
'
m
p
A
v
Q
u
E
C
U
n
M
x
.
g
6
r
e
s
z
o
G
t
N
j
l
c
B
d
X
L
7
m
v
u
6
o
9
l
Q
A
V
U
\194
.
C
r
2
s
8
y
W
f
0
h
4
-
5
O
I
_
1
\195
p
E
"
a
q
t
Non-zero for 14.1% of words.
^Gwen $____ (3.1) ^10-K$ ____ (3.0) ^CIA$ ____ (3.0) ^7-6$ ____ (3.0) ^wav e$____(2.9) ^wav es$____(2.9) ^wav ing$____(2.9) ^wav ed$____(2.9) ^7.6$ ____ (2.8) ^FIA$ ____ (2.8) ^we$ ____(2.8) ^7-5$ ____ (2.8) ^jam$ ____ (2.7) ^Swed en$____ (2.7) ^Swed ish$____ (2.7) ^7.2$ ____ (2.7) ^Rwan da$____ (2.6) ^7.5$ ____ (2.6) ^PHILADELP HIA$____ (2.6) ^wen t$____(2.5) ^wed ding$____(2.5) ^wed dings$____(2.5) ^Camp bell$____ (2.5) ^Camp $____ (2.5) ^Camp aign$____ (2.5) ^wak e$____(2.4) ^Swee t$____ (2.3) ^Swee ney$____ (2.3) ^4-6$ ____ (2.3) ^Swan sea$____ (2.3) ^Swan n$____ (2.3) ^Imp erial$____(2.2) ^wei ght$____(2.2) ^wei ghed$____(2.2) ^wei gh$____(2.2) ^wei ghing$____(2.2) ^BERLIN$ ____ (2.2) ^NL$ ____(2.2) ^7.4$ ____ (2.2) ^Zimbabwe$ ____ (2.2) Filter 17 (bias = -0.36) #
Q
b
8
m
7
B
d
.
3
k
6
A
2
w
1
v
"
q
5
T
4
r
O
g
s
h
,
t
9
l
S
M
C
n
^
\195
I
f
\194
u
&
j
W
u
K
-
Q
F
X
f
z
e
\194
k
/
r
Y
H
!
P
8
t
5
g
(
m
"
i
G
h
o
d
O
E
9
M
N
J
c
)
C
v
U
x
G
q
Z
o
V
f
g
-
A
W
R
e
H
w
Y
a
P
N
k
.
r
B
s
d
Q
b
/
E
m
h
M
9
S
\194
$
\163
w
T
n
c
-
G
W
z
2
k
6
U
X
Y
Z
R
N
r
e
C
q
S
4
g
3
D
5
A
I
s
H
h
M
l
f
p
J
O
/
b
Non-zero for 12.6% of words.
^QC$ ____(2.4) ^WAM $____(2.1) ^Win dows$____(2.0) ^Win ter$____(2.0) ^Win frey$____(2.0) ^Win gs$____(2.0) ^Kin g$____(1.9) ^Kin gdom$____(1.9) ^Kin gs$____(1.9) ^Kin dle$____(1.9) ^Kin gston$____(1.9) ^XVI $____(1.9) ^\194\174$ ____(1.9) ^IOC$ ____ (1.9) ^Wi- Fi$____(1.8) ^Corn wall$____ (1.8) ^Corn ell$____ (1.8) ^Wagn er$____ (1.8) ^Ki- moon$____(1.8) ^Kre mlin$____(1.8) ^Xin hua$____(1.7) ^Xin jiang$____(1.7) ^doin g$____ (1.7) ^wrongdoin g$____ (1.7) ^Kuw ait$____(1.7) ^S.C. $____ (1.6) ^ACORN $____ (1.6) ^Carn egie$____ (1.6) ^Carn ival$____ (1.6) ^HSBC$ ____ (1.6) ^AC$ ____(1.6) ^message $____ (1.6) ^message s$____ (1.6) ^passage $____ (1.6) ^usage $____ (1.6) ^ICC$ ____ (1.6) ^Warw ickshire$____ (1.5) ^dose $____ (1.5) ^dose s$____ (1.5) ^overdose $____ (1.5) Filter 18 (bias = -0.64) #
Q
f
U
x
X
p
W
j
Y
g
"
F
H
c
Z
v
<BOS>
h
^
n
/
-
2
m
I
l
1
e
3
t
.
k
b
s
S
v
f
W
l
c
F
?
P
w
j
q
%
T
m
"
L
0
p
9
S
2
r
a
s
1
;
z
,
)
e
E
J
N
'
(
i
Y
/
8
d
z
h
O
F
C
b
p
e
d
H
G
M
c
u
Q
m
o
j
\194
f
D
k
0
B
1
E
/
w
,
g
K
r
$
.
I
A
8
q
P
V
m
v
X
u
Z
U
.
R
M
s
y
z
L
k
H
E
/
B
g
T
Q
9
V
a
l
0
A
d
n
I
$
x
r
o
Y
J
t
i
Non-zero for 15.9% of words.
^vom iting$____(2.6) ^Wom en$____(2.6) ^Wom an$____(2.6) ^com pany$____(2.4) ^com e$____(2.4) ^com panies$____(2.4) ^com es$____(2.4) ^wom en$____(2.3) ^wom an$____(2.3) ^voy age$____(2.2) ^Viacom $____ (1.8) ^Tom $____(1.8) ^Tom my$____(1.8) ^Tom linson$____(1.8) ^Tom as$____(1.8) ^adm inistration$____(1.8) ^adm itted$____(1.8) ^adm it$____(1.8) ^adm inistrative$____(1.8) ^cog nitive$____(1.8) ^vol ume$____(1.8) ^vol unteers$____(1.8) ^vol untary$____(1.8) ^vol atile$____(1.8) ^Wol f$____(1.7) ^Wol ves$____(1.7) ^von $____(1.7) ^newcom ers$____ (1.7) ^newcom er$____ (1.7) ^Edm onton$____(1.7) ^Won der$____(1.7) ^cam e$____(1.7) ^cam paign$____(1.7) ^cam p$____(1.7) ^cam era$____(1.7) ^200m $____ (1.6) ^Wim bledon$____(1.6) ^Wor ld$____(1.6) ^Wor kers$____(1.6) ^Wor k$____(1.6) Filter 19 (bias = -0.39) #
9
m
7
f
Y
k
"
A
\194
U
8
t
-
p
R
a
<BOS>
B
Q
b
3
F
6
w
o
z
1
s
^
y
X
P
0
i
4
l
N
K
J
L
D
k
d
b
"
A
8
f
O
i
0
m
&
B
c
V
\194
n
o
F
G
h
Q
H
6
r
(
l
2
a
X
s
1
g
9
t
E
x
7
p
U
f
z
-
C
o
G
j
Q
x
A
M
a
e
s
n
I
h
c
l
R
J
Z
N
E
m
k
v
P
q
D
i
T
w
"
t
1
F
\163
.
Q
j
"
D
W
0
U
d
Y
l
$
v
r
c
H
n
b
p
y
g
X
5
Z
J
k
x
'
o
t
C
e
F
1
-
Non-zero for 14.5% of words.
^YOU$ ____ (3.5) ^NASDAQ $____ (3.0) ^DC$ ____(2.9) ^DUP $____(2.8) ^EU$ ____(2.6) ^DAX $____(2.5) ^MDC$ ____ (2.5) ^Da$ ____(2.5) ^Dar ling$____(2.4) ^Dar fur$____(2.4) ^Dar ren$____(2.4) ^Dar win$____(2.4) ^Dar k$____(2.4) ^Day $____(2.3) ^Day s$____(2.3) ^Dak ota$____(2.3) ^USDA$ ____ (2.2) ^SOUR CE$____ (2.1) ^CDC$ ____ (2.1) ^Dam e$____(2.1) ^Dam ascus$____(2.1) ^Dam on$____(2.1) ^Dam ien$____(2.1) ^one-day $____ (2.1) ^two-day $____ (2.1) ^three-day $____ (2.1) ^day-to-day $____ (2.1) ^soda$ ____ (2.0) ^Woods$ ____ (2.0) ^goods$ ____ (2.0) ^methods$ ____ (2.0) ^periods$ ____ (2.0) ^foods$ ____ (2.0) ^neighborhoods$ ____ (2.0) ^1990s$ ____ (2.0) ^90s$ ____ (2.0) ^mid-1990s$ ____ (2.0) ^1970s$ ____ (2.0) ^70s$ ____ (2.0) ^da$ ____(2.0) Filter 20 (bias = -0.52) #
M
g
F
p
<BOS>
d
f
z
V
c
B
G
m
r
/
-
X
a
Z
o
\194
T
6
\163
S
E
K
0
4
D
H
O
Q
v
,
1
L
\195
5
I
D
p
H
x
L
b
(
f
u
k
/
w
U
v
&
-
I
g
A
o
T
m
F
'
1
a
M
G
t
c
7
r
Q
z
C
i
6
W
Z
V
Y
F
m
f
i
e
X
E
g
N
G
B
w
c
-
s
W
D
V
x
l
j
z
v
p
u
Z
S
$
t
/
8
H
9
h
r
R
F
-
0
i
5
p
V
r
\194
a
c
O
9
y
B
u
M
U
6
I
8
w
v
E
Z
s
C
d
X
P
7
H
j
\195
D
t
x
o
4
m
Non-zero for 14.3% of words.
^FDIC $____ (2.4) ^Dic k$____(2.2) ^Div ision$____(2.1) ^D-C alif$____(2.0) ^Hic ks$____(1.9) ^Dix on$____(1.9) ^DVD $____(1.8) ^DVD s$____(1.8) ^MIAM I$____ (1.7) ^FDA$ ____ (1.7) ^Doc tors$____(1.7) ^Doc tor$____(1.7) ^func tion$____ (1.6) ^func tions$____ (1.6) ^func tioning$____ (1.6) ^func tional$____ (1.6) ^MDC$ ____ (1.6) ^Dov er$____(1.6) ^Mumb ai$____ (1.6) ^DAX $____(1.6) ^D-N .Y.$____(1.5) ^Dav id$____(1.5) ^Dav is$____(1.5) ^Dav e$____(1.5) ^Dav ies$____(1.5) ^Liv erpool$____(1.5) ^Liv e$____(1.5) ^Liv ing$____(1.5) ^Liv ni$____(1.5) ^BAGHDAD $____ (1.5) ^HIV $____(1.5) ^Di$ ____(1.5) ^conflic t$____ (1.4) ^conflic ts$____ (1.4) ^conflic ting$____ (1.4) ^inflic ted$____ (1.4) ^D.C .$____(1.4) ^Hin du$____(1.4) ^Hoc key$____(1.4) ^liftin g$____ (1.4) Filter 21 (bias = -0.47) #
8
w
6
z
7
g
L
t
9
s
F
O
X
-
H
E
h
G
x
i
q
k
N
S
Z
o
4
p
B
T
1
c
n
m
Q
U
\194
I
D
'
h
c
Y
z
H
d
:
v
(
p
W
P
S
D
i
)
A
w
u
C
4
e
o
I
"
\163
L
0
N
s
3
G
l
f
y
-
;
F
/
E
N
k
5
m
8
-
X
p
K
u
9
g
Q
i
L
b
6
f
o
s
7
P
\194
'
/
U
0
r
D
v
3
d
O
y
W
t
2
V
4
w
I
m
S
-
Q
u
E
f
O
v
t
n
A
x
W
b
X
U
2
'
7
y
5
M
N
Z
4
c
T
s
H
d
,
J
e
o
.
k
Non-zero for 14.7% of words.
^NYSE $____ (3.3) ^hot el$____(3.1) ^hot $____(3.1) ^hot els$____(3.1) ^hot test$____(3.1) ^Mourinho$ ____ (2.8) ^NHL$ ____ (2.8) ^Hot el$____(2.8) ^Hot $____(2.8) ^Hot els$____(2.8) ^ATLANT A$____ (2.7) ^PHILADE LPHIA$____ (2.6) ^Phoe nix$____ (2.5) ^Foot ball$____ (2.4) ^NHS$ ____ (2.4) ^Idaho$ ____ (2.2) ^HBO S$____(2.2) ^HBO $____(2.2) ^Manhat tan$____ (2.2) ^bondhol ders$____ (2.2) ^shoot ing$____ (2.2) ^shoot $____ (2.2) ^shoot ings$____ (2.2) ^shoot out$____ (2.2) ^quot ed$____ (2.2) ^quot e$____ (2.2) ^quot es$____ (2.2) ^quot ing$____ (2.2) ^hor se$____(2.1) ^hor ses$____(2.1) ^hor ror$____(2.1) ^hor mone$____(2.1) ^hor rible$____(2.1) ^hor izon$____(2.1) ^1945$ ____ (2.1) ^Who$ ____ (2.1) ^Ho$ ____(2.1) ^hol d$____(2.0) ^hol ding$____(2.0) ^hol iday$____(2.0) Filter 22 (bias = -0.47) #
L
t
K
k
y
j
U
-
G
v
/
i
Z
g
D
I
8
w
O
q
Q
n
X
h
P
e
o
u
N
H
m
'
z
V
"
s
<EOS>
4
Y
<BOS>
c
P
h
f
0
U
Y
m
o
!
v
I
g
%
S
-
x
;
G
F
T
r
\194
:
5
a
9
d
4
\195
W
L
8
s
C
Z
1
u
7
H
Q
p
/
g
\194
i
$
x
"
P
U
h
.
v
s
0
'
J
N
b
R
e
G
d
c
1
q
r
y
k
f
X
h
B
y
K
-
W
d
Q
g
Z
t
z
u
b
r
V
f
9
j
6
p
J
s
2
n
Y
c
w
o
\194
F
8
i
U
C
"
'
5
H
Non-zero for 15.2% of words.
^Los$ ____ (3.2) ^LSU$ ____ (3.1) ^Lou$ ____ (3.0) ^MLS$_ ___ (2.8) ^LG$_ ___ (2.7) ^embryos$ ____ (2.7) ^Tokyo$_ ___ (2.6) ^Mayo$_ ___ (2.6) ^you$ ____ (2.4) ^Lisb on$____ (2.4) ^TOKYO$ ____ (2.4) ^Oh$_ ___ (2.3) ^Go$_ ___ (2.3) ^MSNB C$____ (2.2) ^LLC$_ ___ (2.2) ^PLC$_ ___ (2.2) ^1980s$ ____ (2.2) ^80s$ ____ (2.2) ^80$_ ___ (2.1) ^1980$_ ___ (2.1) ^180$_ ___ (2.1) ^Toyota $____ (2.1) ^0.9 $____(2.1) ^Pc$_ ___ (2.0) ^bloc$_ ___ (2.0) ^havoc$_ ___ (2.0) ^oh$_ ___ (2.0) ^cub ic$____(2.0) ^US$_ ___ (2.0) ^hub $____(2.0) ^Koso vo$____ (2.0) ^Do$_ ___ (2.0) ^Lt.$ ____ (2.0) ^0.6 $____(2.0) ^Low$ ____ (1.9) ^Loui s$____ (1.9) ^Loui siana$____ (1.9) ^Loui sville$____ (1.9) ^Loui se$____ (1.9) ^No.$ ____ (1.9) Filter 23 (bias = -0.36) #
.
k
-
B
<BOS>
p
Q
P
\194
f
/
K
^
i
"
U
X
x
v
J
T
C
0
b
z
c
F
R
\195
Z
k
Q
j
!
h
&
i
/
p
X
t
?
f
K
F
"
r
W
u
(
e
\194
l
M
g
.
P
8
E
G
x
:
v
d
T
J
G
f
g
u
j
i
L
y
z
W
Z
h
D
x
A
-
0
,
X
o
E
q
V
a
l
k
c
H
J
p
5
n
.
v
Q
'
F
t
b
d
.
p
Q
P
b
i
w
T
E
d
Z
D
s
y
W
t
x
H
"
1
z
j
A
C
N
r
a
n
V
-
S
k
X
o
L
f
B
h
\194
u
Non-zero for 5.6% of words.
^.... $____ (4.0) ^.... .$____ (4.0) ^U.K.$ ____ (3.2) ^G.M.$ ____ (2.8) ^...$ ____ (2.7) ^non-GAA P$____ (2.6) ^U.N.$ ____ (2.4) ^N.Y.$ ____ (2.4) ^D-N.Y.$ ____ (2.4) ^middle-cla ss$____ (2.3) ^working-cla ss$____ (2.3) ^world-cla ss$____ (2.3) ^first-cla ss$____ (2.3) ^N.F.L.$ ____ (2.2) ^McQ ueen$____(2.1) ^p.m.$ ____ (2.1) ^a.m.$ ____ (2.1) ^KAB UL$____(2.0) ^MLS $____(2.0) ^1.25$ ____ (2.0) ^D.C.$ ____ (2.0) ^N.C.$ ____ (2.0) ^S.C.$ ____ (2.0) ^LG$ ____(2.0) ^Zea land$____(2.0) ^Q.$ ____(1.9) ^guardian.co. uk$____ (1.9) ^Web $____(1.9) ^Web b$____(1.9) ^Web ber$____(1.9) ^Web ster$____(1.9) ^Web er$____(1.9) ^Ph.D.$ ____ (1.8) ^WAS HINGTON$____(1.8) ^N.J. $____ (1.8) ^north-wes t$____ (1.8) ^MGM $____(1.8) ^L.A. $____ (1.8) ^Mr. $____(1.8) ^80s $____(1.7) Filter 24 (bias = -0.43) #
<BOS>
1
f
H
'
g
F
y
l
G
x
i
.
h
\194
U
j
2
S
u
v
3
t
A
b
a
m
T
-
4
Q
E
B
Y
M
D
e
Z
N
0
G
u
X
t
K
m
2
h
8
-
5
k
Z
f
Q
l
0
s
9
.
3
i
7
H
6
v
O
'
1
U
4
a
c
b
W
A
E
T
D
%
m
I
Y
E
v
2
T
r
k
d
h
N
\194
e
b
n
B
3
V
s
M
w
W
-
"
5
'
1
x
P
$
O
X
F
u
a
t
7
l
j
d
f
D
w
\194
B
7
k
Y
E
C
r
6
b
1
K
$
e
l
N
X
x
/
F
Q
a
L
o
0
t
8
S
A
p
s
_
Non-zero for 20.2% of words.
^feud $____ (2.8) ^watchd og$____ (2.7) ^2m$ ____(2.6) ^MGM$ ____ (2.5) ^fold $____ (2.4) ^3m$ ____(2.4) ^problem$ ____ (2.3) ^Jerusalem$ ____ (2.3) ^Harlem$ ____ (2.3) ^EBITD A$____ (2.2) ^50m$ ____ (2.2) ^GB$ ____(2.2) ^GM$ ____(2.2) ^Kid s$____(2.1) ^Kid d$____(2.1) ^1m$ ____(2.1) ^system$ ____ (2.1) ^stem$ ____ (2.1) ^System$ ____ (2.1) ^item$ ____ (2.1) ^post-mortem$ ____ (2.1) ^food $____ (2.0) ^food s$____ (2.0) ^seafood $____ (2.0) ^NOT$ ____ (2.0) ^God $____(2.0) ^slalom$ ____ (1.9) ^fema le$____ (1.9) ^fema les$____ (1.9) ^fell $____ (1.9) ^fell ow$____ (1.9) ^Rockefell er$____ (1.9) ^match$ ____ (1.9) ^watch$ ____ (1.9) ^pitch$ ____ (1.9) ^catch$ ____ (1.9) ^OTC $____(1.9) ^NYC $____(1.8) ^check$ ____ (1.8) ^neck$ ____ (1.8) Filter 25 (bias = -0.41) #
Z
p
9
t
F
i
8
w
V
o
7
O
X
-
L
y
Q
a
b
s
J
f
6
m
R
W
<BOS>
,
M
z
j
d
B
I
0
g
4
'
Y
T
(
v
Q
x
&
p
O
f
/
k
:
b
?
)
N
j
"
V
!
i
r
m
H
J
A
B
I
g
.
F
U
0
R
n
Y
e
y
c
w
H
b
1
x
i
v
4
.
2
-
3
f
D
'
5
r
,
z
I
k
M
a
A
q
y
c
C
p
t
o
6
R
/
\195
Z
B
T
"
O
\163
-
F
Y
t
b
f
g
y
l
,
G
D
.
c
V
B
x
C
z
K
\194
N
'
n
X
P
J
T
Q
U
$
M
"
e
m
I
\195
k
2
Non-zero for 15.6% of words.
^brib es$____ (2.6) ^brib ery$____ (2.6) ^brig ht$____ (2.4) ^brig ade$____ (2.4) ^bril liant$____ (2.3) ^bril liantly$____ (2.3) ^Brig hton$____ (2.2) ^Brig ade$____ (2.2) ^Nig eria$____(2.0) ^Nig ht$____(2.0) ^Nig erian$____(2.0) ^Nig el$____(2.0) ^Nig er$____(2.0) ^Oil $____(2.0) ^rib s$____(1.9) ^buil ding$____ (1.9) ^buil d$____ (1.9) ^buil t$____ (1.9) ^buil dings$____ (1.9) ^rig ht$____(1.8) ^rig hts$____(1.8) ^rig ht-wing$____(1.8) ^rig orous$____(1.8) ^rig htly$____(1.8) ^Hig h$____(1.8) ^Hig hway$____(1.8) ^Hig her$____(1.8) ^Hig hland$____(1.8) ^terrib le$____ (1.8) ^horrib le$____ (1.8) ^terrib ly$____ (1.8) ^Flig ht$____ (1.7) ^19th- century$____ (1.7) ^Wi- Fi$____(1.7) ^Buil ding$____ (1.7) ^Uig hurs$____(1.7) ^Uig hur$____(1.7) ^bail out$____ (1.7) ^bail $____ (1.7) ^bail ed$____ (1.7) Filter 26 (bias = -0.52) #
C
b
I
J
U
x
d
v
c
E
n
e
D
h
Q
W
/
w
,
B
'
q
s
M
t
Y
P
o
R
-
A
4
z
i
F
m
r
j
y
X
J
y
j
h
X
U
0
s
7
t
Z
a
-
u
I
.
V
f
2
x
P
m
5
A
9
S
6
"
n
c
g
W
G
,
1
o
3
k
)
B
W
u
p
D
K
j
i
.
B
g
x
c
X
F
a
d
,
-
O
C
w
r
Y
R
q
L
f
s
k
M
"
Z
6
n
V
U
Q
h
2
e
Q
f
"
J
X
P
Y
o
\194
i
Z
-
$
I
.
t
V
w
W
p
8
E
h
l
c
s
7
r
\195
j
B
e
u
z
Non-zero for 22.9% of words.
^MOSCOW$ ____ (3.0) ^Clic k$____ (2.9) ^CITY $____ (2.9) ^adjac ent$____ (2.6) ^CIT$ ____ (2.6) ^FRANCISC O$____ (2.6) ^VW$ ____(2.5) ^Clay $____ (2.5) ^Clay ton$____ (2.5) ^Cric ket$____ (2.4) ^CIA$ ____ (2.4) ^Clim ate$____ (2.3) ^CEO$ ____ (2.3) ^jih ad$____(2.3) ^Jac kson$____(2.2) ^Jac k$____(2.2) ^Jac ob$____(2.2) ^Jac obs$____(2.2) ^Sunni$ ____ (2.2) ^IPO$ ____ (2.1) ^Jay $____(2.1) ^Jay s$____(2.1) ^Jag uar$____(2.1) ^Punjab $____ (2.0) ^Cliv e$____ (2.0) ^Benjam in$____ (2.0) ^CNBC $____ (2.0) ^Jo$ ____(2.0) ^jac ket$____(2.0) ^jac kets$____(2.0) ^Jim $____(1.9) ^Jim my$____(1.9) ^JP$ ____(1.9) ^enjoy $____ (1.9) ^enjoy ed$____ (1.9) ^enjoy ing$____ (1.9) ^enjoy s$____ (1.9) ^Anna$ ____ (1.9) ^Madonna$ ____ (1.9) ^Vienna$ ____ (1.9) Filter 27 (bias = -0.51) #
W
p
H
P
X
f
w
d
4
j
Z
l
Q
-
A
r
2
z
q
R
^
c
/
s
Y
v
"
o
M
x
J
D
G
0
b
g
v
V
d
G
a
Y
q
i
N
Z
u
S
t
m
D
j
f
4
B
M
.
H
T
X
c
3
x
5
\163
C
U
)
o
%
e
I
\$
k
-
V
u
p
L
C
l
Q
o
b
.
x
D
F
m
9
w
8
d
"
t
B
U
r
s
7
z
P
E
c
/
X
J
R
O
'
M
q
e
6
r
7
t
8
T
x
o
S
k
\194
w
4
A
5
y
Q
n
F
B
V
m
X
i
9
p
s
H
"
q
j
u
d
\195
$
N
G
U
E
a
Non-zero for 20.3% of words.
^Wiki pedia$____ (3.1) ^Mike $____ (2.7) ^Nike $____ (2.6) ^Wire less$____ (2.6) ^hike $____ (2.5) ^hike s$____ (2.5) ^GPS $____(2.4) ^Mikh ail$____ (2.2) ^wipe d$____ (2.2) ^wipe $____ (2.2) ^NYC$ ____ (2.2) ^Wins ton$____ (2.1) ^GPs $____(2.1) ^ships $____ (2.1) ^relationships $____ (2.1) ^championships $____ (2.1) ^chips $____ (2.1) ^Ask$ ____ (2.1) ^G8$ ____(2.0) ^Agre ement$____ (2.0) ^GB$ ____(2.0) ^Like $____ (2.0) ^talks $____ (2.0) ^walks $____ (2.0) ^Talks $____ (2.0) ^Wind ows$____ (2.0) ^Wind $____ (2.0) ^Wind sor$____ (2.0) ^gre at$____(2.0) ^gre ater$____(2.0) ^gre en$____(2.0) ^gre w$____(2.0) ^gre atest$____(2.0) ^gre enhouse$____(2.0) ^risks $____ (2.0) ^asks $____ (2.0) ^tasks $____ (2.0) ^masks $____ (2.0) ^Hawks $____ (1.9) ^Seahawks $____ (1.9) Filter 28 (bias = -0.47) #
<BOS>
y
Q
h
E
d
w
u
W
m
X
p
I
c
N
D
S
L
2
C
K
x
O
g
e
l
"
n
^
U
5
P
1
a
v
H
g
f
Y
x
R
L
T
y
k
K
I
N
j
F
-
B
V
n
;
a
G
.
C
\$
Q
o
r
e
z
,
'
m
&
%
)
w
i
5
\194
8
I
u
Q
M
l
f
A
m
X
v
p
y
7
c
O
U
a
o
2
-
S
x
W
B
t
F
q
n
s
Z
k
R
'
h
R
y
V
h
-
D
s
d
b
p
k
L
z
t
w
c
J
,
E
F
Q
q
'
l
U
H
G
f
Z
1
Y
o
9
x
$
T
\195
A
"
n
Non-zero for 13.3% of words.
^TAR P$____(2.7) ^ERA$ ____ (2.5) ^YOR K$____(2.4) ^ETA$ ____ (2.3) ^WTA$ ____ (2.3) ^NASCAR $____ (2.3) ^III$ ____ (2.3) ^Wils on$____ (2.2) ^TOR ONTO$____(2.2) ^IRA$ ____ (2.2) ^XVI$ ____ (2.2) ^MRI$ ____ (2.1) ^Wilk inson$____ (2.1) ^HIV $____(2.1) ^Vegas $____ (2.0) ^Fabregas $____ (2.0) ^gas $____(2.0) ^gas oline$____(2.0) ^gas es$____(2.0) ^Ras mussen$____(2.0) ^Ras hid$____(2.0) ^II$ ____(1.9) ^NYSE $____ (1.9) ^ATLANTA$ ____ (1.9) ^eggs $____ (1.9) ^WTO$ ____ (1.9) ^IAE A$____(1.9) ^Ips wich$____(1.9) ^Tas k$____(1.8) ^CHICAG O$____ (1.8) ^TIM E$____(1.7) ^Tak e$____(1.7) ^Tak ing$____(1.7) ^PAR IS$____(1.7) ^legis lation$____ (1.7) ^regis tered$____ (1.7) ^legis lative$____ (1.7) ^regis ter$____ (1.7) ^YOU $____(1.7) ^Ris k$____(1.7) Filter 29 (bias = -0.48) #
B
p
.
i
b
O
v
P
<BOS>
d
N
y
q
s
M
E
V
I
\194
G
Z
1
n
U
x
2
9
,
F
-
h
z
c
3
k
S
X
a
Q
o
I
f
i
F
W
x
Y
.
1
L
T
r
\194
e
2
y
C
b
H
m
z
\$
&
%
w
h
k
N
V
j
X
l
4
S
o
p
s
E
n
S
m
O
C
Y
d
h
.
3
D
G
v
o
c
r
t
"
Z
R
/
4
'
W
U
J
z
g
P
9
f
N
\194
x
l
e
F
8
I
Y
f
V
B
\194
y
$
c
g
F
Q
N
S
K
7
r
X
e
G
a
4
P
"
U
n
v
t
p
D
o
_
w
Non-zero for 25.6% of words.
^BCS$ ____ (3.1) ^Fabio$ ____ (2.7) ^biog raphy$____ (2.7) ^autobiog raphy$____ (2.7) ^BAE$ ____ (2.6) ^Silvio$ ____ (2.6) ^NHS$ ____ (2.6) ^PARIS$ ____ (2.5) ^TORONTO$ ____ (2.5) ^Big$ ____ (2.5) ^HBOS$ ____ (2.5) ^RBIs$ ____ (2.5) ^big$ ____ (2.4) ^bigg est$____ (2.4) ^bigg er$____ (2.4) ^Robbie$ ____ (2.4) ^Debbie$ ____ (2.4) ^Lockerbie$ ____ (2.4) ^biol ogical$____ (2.3) ^biol ogy$____ (2.3) ^movie$ ____ (2.3) ^Movie$ ____ (2.3) ^cannabis$ ____ (2.2) ^SOURCE$ ____ (2.2) ^viol ence$____ (2.2) ^viol ent$____ (2.2) ^viol ations$____ (2.2) ^viol ated$____ (2.2) ^Mir$ ____ (2.2) ^Virg inia$____ (2.1) ^Virg in$____ (2.1) ^Antonio$ ____ (2.1) ^Davis$ ____ (2.1) ^Elvis$ ____ (2.1) ^Travis$ ____ (2.1) ^IS$ ____(2.1) ^DENVER $____ (2.0) ^Ohio$ ____ (2.0) ^1.25$ ____ (1.9) ^Bashir$ ____ (1.9) Filter 30 (bias = -0.59) #
<BOS>
t
X
y
9
p
Z
T
7
m
Q
c
6
g
8
o
J
z
N
h
2
k
4
i
3
s
5
A
V
O
F
U
I
a
j
G
n
l
^
S
F
k
e
-
\$
v
N
Y
E
z
2
u
f
m
5
R
4
'
8
T
L
g
3
C
S
c
W
i
6
)
K
G
Q
V
X
U
,
\195
(
l
Y
k
S
B
j
f
7
U
G
w
h
a
5
b
g
r
4
m
l
v
6
t
\194
K
3
n
0
q
1
P
8
N
$
z
O
\195
L
c
X
.
Y
f
\194
r
4
B
7
t
V
K
6
A
g
N
-
F
1
a
$
p
Z
P
X
k
"
o
H
,
G
U
W
l
8
e
3
b
0
_
O
Non-zero for 25.9% of words.
^1957$ ____ (3.1) ^1947$ ____ (3.0) ^1987$ ____ (3.0) ^1955$ ____ (3.0) ^1954$ ____ (2.9) ^1945$ ____ (2.9) ^1985$ ____ (2.9) ^1944$ ____ (2.9) ^787$ ____ (2.9) ^NFL$ ____ (2.9) ^1984$ ____ (2.9) ^1956$ ____ (2.8) ^1967$ ____ (2.8) ^1934$ ____ (2.8) ^1946$ ____ (2.8) ^1986$ ____ (2.8) ^1953$ ____ (2.7) ^1950$ ____ (2.7) ^1965$ ____ (2.7) ^1951$ ____ (2.7) ^1964$ ____ (2.7) ^1943$ ____ (2.7) ^1983$ ____ (2.7) ^1940$ ____ (2.7) ^1958$ ____ (2.6) ^1980$ ____ (2.6) ^NY$ ____(2.6) ^1941$ ____ (2.6) ^1981$ ____ (2.6) ^750$ ____ (2.6) ^1966$ ____ (2.6) ^1933$ ____ (2.6) ^Schwarzenegg er$____ (2.6) ^egg s$____(2.6) ^egg $____(2.6) ^1977$ ____ (2.6) ^1948$ ____ (2.6) ^1988$ ____ (2.6) ^BEIJING$ ____ (2.5) ^1963$ ____ (2.5) Filter 31 (bias = -0.47) #
<BOS>
f
w
h
W
F
Q
P
O
y
z
m
E
u
I
L
X
x
"
p
^
r
\194
H
2
n
G
d
l
k
b
D
j
B
k
D
V
o
C
d
;
.
'
E
Q
-
P
e
R
L
b
l
B
v
Z
0
"
t
F
g
Y
w
r
u
f
T
U
j
)
c
p
h
,
z
X
f
L
p
H
o
Q
S
7
c
q
t
6
y
Z
O
9
s
u
m
8
i
a
k
b
'
\195
g
J
w
2
M
D
C
1
K
Y
x
d
,
y
B
d
k
H
v
1
b
u
t
s
q
4
T
-
l
Z
f
G
r
3
N
"
o
8
J
2
z
U
K
6
x
$
w
L
A
7
\195
D
R
Non-zero for 15.5% of words.
^okay $____ (2.6) ^key $____(2.4) ^key s$____(2.4) ^key board$____(2.4) ^key note$____(2.4) ^blockad e$____ (2.3) ^Pay ne$____(2.2) ^Pay $____(2.2) ^Brookly n$____ (2.2) ^Cad bury$____(2.2) ^Cad illac$____(2.2) ^Ray $____(2.1) ^Ray mond$____(2.1) ^Ray s$____(2.1) ^quickly $____ (2.1) ^buy $____(2.1) ^buy ing$____(2.1) ^buy ers$____(2.1) ^buy er$____(2.1) ^Vau xhall$____(2.0) ^Vau ghan$____(2.0) ^Buy $____(2.0) ^Rud d$____(2.0) ^Rud y$____(2.0) ^Pad res$____(2.0) ^bay $____(2.0) ^Bay $____(1.9) ^Bay ern$____(1.9) ^Bay lor$____(1.9) ^Rad io$____(1.9) ^Rad cliffe$____(1.9) ^low-key $____ (1.9) ^bud get$____(1.9) ^bud gets$____(1.9) ^weekly $____ (1.8) ^Weekly $____ (1.8) ^Blackbu rn$____ (1.8) ^blockbu ster$____ (1.8) ^Cus toms$____(1.8) ^Cus tomers$____(1.8) Filter 32 (bias = -0.77) #
<BOS>
t
L
T
/
g
X
k
Z
h
Q
i
<EOS>
y
K
p
l
j
9
e
\194
E
z
c
6
u
.
r
8
H
7
\163
B
d
b
-
5
v
N
F
W
s
X
f
V
l
2
t
4
C
Y
.
E
o
6
n
q
A
"
c
H
L
3
F
b
,
)
D
Z
U
w
%
9
d
1
'
J
r
7
z
f
z
F
.
B
-
i
U
h
G
k
g
t
u
,
d
S
s
M
E
p
L
j
Z
H
a
4
\195
x
b
n
w
5
Q
W
R
6
r
T
A
W
j
w
F
z
P
K
H
o
f
"
k
O
r
Q
h
a
i
.
C
N
n
X
g
\194
u
v
d
B
l
c
p
$
-
Y
J
/
e
q
s
Non-zero for 23.8% of words.
^Wiz ards$____(4.1) ^Who $____(3.8) ^Wha t$____(3.5) ^Wha tever$____(3.5) ^WHO $____(3.1) ^Califo rnia$____ (3.1) ^Vio lence$____(3.0) ^biz arre$____(2.9) ^WTO $____(2.9) ^Halifa x$____ (2.8) ^Wyo ming$____(2.8) ^Crawfo rd$____ (2.8) ^Befo re$____ (2.8) ^befo re$____ (2.8) ^Silvio $____ (2.7) ^ANGELES$ ____ (2.7) ^Via com$____(2.7) ^JERUSALEM$ ____ (2.7) ^Wea ther$____(2.7) ^awkw ard$____ (2.6) ^likeliho od$____ (2.6) ^Betw een$____ (2.6) ^Woo ds$____(2.6) ^Woo d$____(2.6) ^Woo dward$____(2.6) ^Woo dy$____(2.6) ^bio logical$____(2.6) ^bio fuels$____(2.6) ^bio graphy$____(2.6) ^bio logy$____(2.6) ^betw een$____ (2.6) ^Netw ork$____ (2.6) ^Netw orks$____ (2.6) ^BMW $____(2.5) ^unifo rm$____ (2.5) ^unifo rms$____ (2.5) ^Leic ester$____ (2.5) ^Leic estershire$____ (2.5) ^HBO S$____(2.5) ^HBO $____(2.5) Filter 33 (bias = -0.54) #
N
g
W
-
B
j
K
m
Q
p
,
G
/
d
O
v
q
s
"
l
X
'
a
V
H
b
A
k
2
c
3
z
I
C
<BOS>
x
8
i
9
J
W
g
\194
U
"
P
:
C
o
r
x
k
X
)
(
A
Q
F
v
G
Y
R
S
n
N
Z
6
u
O
c
&
p
/
s
q
\195
D
z
Z
f
g
t
H
d
V
p
G
v
Y
P
A
s
w
,
X
F
4
x
1
o
3
S
7
'
2
O
L
l
n
T
5
-
M
e
h
k
0
I
f
g
F
u
x
T
B
G
N
Y
9
H
e
m
K
h
Q
y
6
D
8
U
X
A
W
c
5
i
,
z
2
C
q
-
7
1
b
d
S
l
Non-zero for 11.4% of words.
^Norf olk$____ (3.1) ^None $____ (3.0) ^None theless$____ (3.0) ^Nobe l$____ (2.8) ^Nige ria$____ (2.8) ^Nige rian$____ (2.8) ^Nige l$____ (2.8) ^Nige r$____ (2.8) ^Now$ ____ (2.6) ^NW$_ ___ (2.3) ^Wolf $____ (2.3) ^N.Y. $____ (2.1) ^D-N.Y. $____ (2.1) ^Winf rey$____ (2.1) ^No.$ ____ (2.0) ^toge ther$____ (2.0) ^altoge ther$____ (2.0) ^Hoga n$____ (2.0) ^NBA$ ____ (2.0) ^Kobe $____ (2.0) ^Wome n$____ (1.9) ^Newa rk$____ (1.9) ^Alge ria$____ (1.9) ^Wagn er$____ (1.8) ^Toge ther$____ (1.8) ^Howe ver$____ (1.8) ^WAS HINGTON$____(1.8) ^Whe n$____(1.8) ^Whe re$____(1.8) ^Whe ther$____(1.8) ^Whe eler$____(1.8) ^Nine $____ (1.8) ^NYC$ ____ (1.7) ^Nobo dy$____ (1.7) ^Wiga n$____ (1.7) ^New$ ____ (1.7) ^hydroge n$____ (1.7) ^ATLANTA$ ____ (1.7) ^Norw ay$____ (1.7) ^Norw egian$____ (1.7) Filter 34 (bias = -0.51) #
P
-
C
.
p
h
k
u
U
v
r
l
K
j
R
M
c
H
z
m
G
\194
,
e
F
q
Q
W
I
Y
O
<BOS>
s
L
B
w
f