c c 8 Ã d
MI-Trigger-based Language Modeling
!" " # $ $ % &'( "! ) * * + ,-.& /01"$"2"2 2" 3 . (" $$ +%( 3 .
Abstract 0d 0 d d d d d
dd ! d d "# $ "##$ 0 d % & &'
d d ## 0 d d d # 0 d % d d d d d 0 ( % d d ) d d d
d 0 d £eywords ü * 0 * d + d c , # *
1 Introduction & & &) !" 4" ! & "" 3 )
)"2 4$ " &) !"" $$ "$&&$ .$$$ $ & 4" "$3 25 $& ""$ 4 & "6"7 6) &72 Ã8 && !. $ " ) )" ) " & ( " "$ " 6" 7 6) & $ 7 & ! "& "! 4 " $"2 "!$&"" &"
^ # -d .0
( "& $$ ) ""$ 2 ) &! ""$ ) " 63 93 7 6$9" 72 "!$&$ & 4 " Ã: ! 2 &2. ; $ <" $ &! ""$ ) ) "" " &! ""$ ) 2 " ) ""$ $ 3 " 3! & 2 =! $ & 3 )
) )" ) . ) $ ! " & $ $". "$ " " $ $ 3 )
à 22 6$9" 7. $& $ 3 )
> $( Ã 226"9 7. "&"$ Ã 226 9 72 " " & $" $ " 3 2 " $ Ã8& 2 &2.
? ") & $ & " "$ & "&( " $ 3!2#" "& @&"" $ 6 Ã) A& $ A "" $/27 B ( 3". &/ $" " " Ã6 $/7 2$ ! Ã6) 7$ " 3> $( $& ( 3"Ã 6& $ 7 672 " " ) &&/) " " Ã $ 3& &!"" @&"2 " $ ") " 3& &!"" "& $ 3 $" 3! $ & ( 3 " Ã6) 97 3> $ " Ã6 ""972 Ã: . &2.
" $ & & & $ &&!" " " "2Ã*" &*.
C" $$ " 3 "$ 3 & 4 $ 3 $/ $ <" "!2 & " $ $& " Ã= :.
? Ã3 ! " 2 &2.
C2 # " . ) " $ $ "2 8 $ . &" 4"" & "" 3 )
# " $ $ "2 Ã &2.
&! " "$ & 3# " $ $ " " $ )$ $ )" 3! / " " $ $ ""$ 3 " & 2 8 .) )&&$"" & ""3 )
)"2 $ & ""3 )
)"$ 4 "" $ & " $ 2 D& % &" "& & & ( 3
"$$ ""&&!" "
$ $ "/". ! ( 3(" $ $ "2B " $ .% &"$ &!$ "" $ $!) %) )) ) $ &! & " $$ & % & & "
! /" $ " && $$ 3 !
) ))2 D& ) $ " $( & % &" $ "" $ $!. &" $ $!"& &"3 4& &!2 " " ""! & ""3 )
)"( "&" $ " ) & $$ "$ # " & 2
0
;
- "$ " "&&)" $+ " $$ 2 $ )" & $ "$3 $?)& $C "$3 " " " 2 $ "$3 " 3 " & & $2 $ ( " 4 & " &$ " %E%# $ $( "2B &&!. " !" "( $;2 - Concept of Trigger Pair = " 3( "$.) ( $ " " 3 "$$$ 4 $ ) ""$ ""$ ) 2 ) · " &! ""$ ) ) . Ã · { " $" 6 7. ) · 3 )2 D · $$" $ . " . $ " " 3 3&! " $ 2 · $ 3 &" 4 ) " F $ "2 B "&$!. ) )&& $$ & "" 3 )
"& )" & " $ 3 4 & ) " F $ "2 8) 3& 3 " & & G ) 3& " 3 "&( )" & $ G + ) " G D )&&"$"" " &! 4)" $"2 ´ Selecting Trigger Pair @( ) $ "$ Ã *. ! ) H = 3 "& )". 3 "$ ""& 2 / 3 "- ($ 3& !2 % .&/ $ " 3 &) 3 $" $(
/ + . 3 ) " + ) 3)"$$ " $ " "$ $ / 2 . ) " ( & 3& $"" &) !"$& ""
" & $ " 3& 3 " ) & " " 3 " & &2
^ # -d .0
´ 1 Window Size "3(") !$& 3 "" "$ )) "- . )$ " 4 " $ 3 )
2 $ " 3& ))"- .) "/)) $ " $ 3 )
))" $" ) 3 3& "2 . )
$"$ &" $ D = ÃD= &" " $
.+.222. 2 " $ " " " $&. "$ ) 4 $
"$ " $ 2D $ $ & & 4! $ &" $ D= &2 # & & 4! " " ( 3 ""3& $$ " $ & "32 $ & & 4! $ & "3 )$ & !
X Ã " 3 + X Ã
2
# & @! " ! $ & "32 ( ) ( 3& "
. $ &3 3&! ""$
3 3&! "" $
à . $ & !
à , . &
( .
XÃ
" "
XÃ
.
à , . & +
à ,
,
à B & $". $ & & 4! " " &&! $ $( ! 3! & &) $ & & 4!. $( !" " 3 &2 ""3 $ " &$ " " $ " $ . ) ( $ ! " ")" $ & & 4!2 " "& 3 " & ( 3 " & " & " ""&! ( 2 8 . $" " $ $ & & 4! " $ &&! " I 8 )" $". )$ " 3 ; && $ $ " + && )"2
0
" $
# & " $ # & & 4! & 4! +? ; C; + ; ? ?
C ;
?; C ;C 3& # & & 4 " &" $ D= &" " $ "
B 3& ) $ & & 4! " &) " J . $ " ""$ &! ") ( J+.?.C. 2BJ;.. . . . $ & & 4!$ " ""&&!2 D $$& "$ 4"" &! & ")" "!28) ( ." "") "$ 4)) "- 2
´ - Selecting Trigger Pair D ( " ©
© ( ))
© , © " 4)0
))02 #" $& 3 )
) ( "
·
© , © $$" " )
à · { . ) " $ & 2
H"& ) ! "" "" "$ $ $ & 3 )
) ( "
à · {
·
" " $""$ Ã# *2 5
" " " & $ " F &!.)$ " " 3" " "
"· $ 0
"· $ " ·
$
"·
$ " · $
Ã+ )
à . " $$ $ F $! ) à . $$
))2
^ # -d .0
H& $""$ " ""& ." &! " 2#" &!$ & $"")
{ . $ & "" ) && $ & . 3 $ Ãi { 2 H $$ $ ) 6 7Ã &
)"Ã
( " 3 ) 67Ã ) $$ $ ) 6i 7Ã$ 3 ) 6 7Ã" 2 % ( & "". "$ ) 6i 7 " &/ &! 3 $ $ " . " ( &! ! 3 $ 2 ) $ $ &! ) " 3 " &. Ãi
{ !3 3& 2
·
. & ( " 4 $ 3 ( 3! $
" ( & ÃH: 3 )
)
· "·
$ 0 "· $
"· $ "· $" $
1
"·
1 " ·
K
$
$
" ·
$
"·
"· $" $ " ·
$
" · $" $
$
" ·
$
" · $" $
Ã? 53("&!.@F ? / " >3 3&!$" 2" "".) " " F " & $ "2 & )/". Ã*" & *2
C " " F Ã#$2 &2.
" ( " $ &&! ! ""$ ) "2 ù Measuring Trigger Pair #"
à · { " & $ 3! ( &
· Ã · L " ") @F ?. & Ã · L & $"
$ & " 3 )
))" .)$$ 3 $ "&&)"
0
"· $ 0
"· $ "· $ 6 " $
ÃC '": 4/ &@" Ã:@ .) $ Ã ·
à ·
à ·
·
à Ã
Ã
Ã
à à ·.
à ·.
à ·.
·.
Ã; )
à · à F $ " )" ·
$" " $( &!.
$$
à ·. " F $! ) à ·.
$$ ))2 ( & " & ºm ºm
à · L " Ã
·
L · .2 2:" 2
.
à ·L 2
3( F ". & ) ))" ( &
à · L & $" $
·
$ & 2 """ !.
à · L . " ! )" ·
. ) $ " & & "
5 MI-Trigger-based Modeling
( 2
à · L " $
à · { 2
+
^ # -d .0
H" "$"" 3( . ) $ "$ 3 " " " 3& )) "- . " & $ " " ( & " "" & 2"" $.) )&& "$3 & 3)3& 3 " &2H" " &! 3! & . ) $ && : "2 3& $$ : &. ) $" ( 3 $" 2 53("&! " ))"- 2H") ( "$ 4))"- . ) )&& 4 ))"- "Ã .+.222. 2 H $ " $ ) " : " $ Ã ) ! " $ Ã ) !2 D& : & " "& . : & " & & ) ""$ 3 " 4 $ ( 3 $ 3 $ " ! " " $ 2D ( " """ " I8 $" + &&)"3!$ 4& $ ". ( !). $ && "$$ $ ")" $ $$$ $ " "$"2 4 & " ") B " +. )$ ") 6ë
ë7Ã6 9 7 " "
$ & ) " $ " + 6i 7Ã6$9" 7 " " $ & ) " $ " + " $( &! I 8 $"2 H &&!3)" " ".) )&&)$$&"" ºm ""& ! 3 ("2 ºm = (" " " $ "& 3 " " $ ) !2 ºm
:" & ""$$ F $!)"2Ãi "
" & Ã
{ 2
{
0
?
Frequency ? C
;+
C
C
C +?
;
uistance C ? + B $$ $ F $! ) Ã6ë9ë7 " $" $
Frequency +C
+
?
C
;
B + $$ $ F $! ) Ã6i " $" $
uistance 9 7
" $ %3
+C
+ +
?
C C
;
+
C
^ # -d .0
: " Ãÿ
3& + 3 " " " $ " : & H( %3 H( %3 H( %3
: Ã ÿ ; : Ã ÿ ; " : Ã ÿ ; " " M .+ ?;; M+ .++ ?C? MC .C+ ;? M+.? +;? M++.+? ? MC+.C? +; M?.C ?;+ M+?.+C + MC?.CC ;;+ MC. C+ M+C.+ + MCC.C ;? M. ? M+.+ +CC MC.C ; M.; C + M+.+; ++?C MC.C; M;. ?; M+;.+ + MC;.C C M. +? ; M+.+ ? MC.C + M . M+ .? ; MC . CC M . +C M?.? ; M. ? M . + +C+C M? .?+
M .+ C M +. ? C+ M?+.?? M+.? M ?. C ;+; M??.?C C M?.C ? M C. ;; M?C.? ?; MC. C M . C M?.?
; M. ? M . ; C M?.?;
M.; C M ;. ? M?;.? + M;. C+ M . CCC M?.?
;+ M. C?+
0
M .+ C+ M? .C M . C + M+.+ ? MC.C
++ NJ +? 3& ? 3 " " ( & : & : %3 : %3 " " M.
+++ M . +? M .+ C? M . + C M+.? ??+C M +. ? ? M?.C + M ?. C + MC. C? M C. ? M. + M . C M.; M . ;
M;. C M ;. M. + +C M . M . ? NJ 3& C %3 " " & : & $ $" 3( ) $". +: &"Ã)$ : &") ))"- " 3&" ""2@ $ & " ))"- ". ) ( & " ) ! ) !2: ( . " $ ". $: &$& " " 3 3 " ")& ( & " && ) " " 2 4 ". &! 3 " && " $& 2 @4 " $" 3 " " 3 " &)&&3 $$ $2
^ # -d .0
B "&$! " . ) $ " " II : . $ && 3 " & " II : &. )& II " " " " )) "- 2 B 4 & . : & " " " $ : 3 " & ) ))"- 2 H&& &" 3& I8 $" + &&)"2 <" / : & " 4 & 2 D & 3 + ÿ + ÿ &&Ã) "4 " $ " ) 3 + # " )" & 4$ ""3& " $ ) "2H" "" .&!) " $$$ & "? " / 2 " "&" 2; && ) "2 " & $ 3! ( & . 3 " &&) " / " "2 B &&!. $ 3 " && : " " 3! & 2 " ) !. ) 3& : & )$ $& " 3 " && " )$ " 3! & 2 " "$" 3 : & ") 3& +. 3& ? 3& C2 " 3& "") 3 " " $ ". ( & & " $( &!2 $ : 3 " &" " "" & )$ &! & $" $ $ ) ) )" $ & . ) & " $3 ) 2 (
© ©+ 222 © . ) $ " & $3 3&! & Ã "&&)" B " $ : 3 " &.
& Ã
& Ã
+ 4Ã .
#
0 Ã 2 { Ã
2
" $ : 3 " &.
& Ã & Ã + 4Ã .
##
0 Ã 2 { .
2
2
à ) " ))" "-
2 " " $ 3 )
)" ©
© 2 2 " $ @F " & $3 3&!
0
;
" ) & " $ " ( & $3 : " : &2
6 PINYIN-to-Character conversion H" &$ : & $. %E%# $ $( " "!" " $"$ 2 $. %E%# $ $( " " 3 $ 3 "$ 3& " # " & $ "" " 3
"3> $" ! " $ " & " $ 2 " ( & /" $ " " "$ "/". $& ºm & " ) $ &Ã O2O2 Ã# 22 &2. ; ) " " & "Ã / 2 &2.
?2 " $ " "! & . 3 $$ $! " & +P ( ) & ) $ "2 ºm & 3 " $$&& 4$."! $$ " $& "Ã8" :22 &2. ÃE #22 &2.
Ã8" D22.
C2 " $ " 3& "&( & & 4$ & 3!3& $ &! $$ $!$ 3 $ P2 ºm " "$ & $Ã# O2 2 &2.
à *2.
+Ã# O22
?Ã :2E2 ;2" $" " & $"$ % " " " " "$ & $ & &". 22 8::. & & $ ""3& $ $ "& "2 $$ $! $ 3
P2 ºm !3 $ " 3 & " " "$ & Ã O2O2.
2 $$ $!$ 3 $&" P2 " " $. ) )&& &! : 3 " &" &$ %E%# $ $( "2 B " $ ". %E% $ " # " " $ "Ã C # " $ $ " ! "$& # " 43/" " "2 %E%# $ $( " " : &" ") 3& 2 D) - " $ " $
?2P
?2P +
C2CP
2P ?
C2;P
2 P C
2P
2?P
2+P
2P
2?P
2P
^ # -d .0
;
C2 P
2CP
C2P
2+P
C2P
2 P
C2?P
2P 3& %E%# $ $( " " : &"
" 3& : &" ( 3 $ " : &" " ))"- 2 . & "" 3 )
)" "& 3 & " $ ) !2 " &" %E%# $ $( " $ $ 2P2 H" ) " " 3( . && : &" &! $& 3 " && "2 5 ! "/ ) " " 3& 3 " : &"&$& G 8 .) )&& 4 $ 3 " " : & %E%# $ $( " "2 D " : &2 %E%# $ $( " " ") 3& 2 %3 : " * $* : : & 2?P .
2;P +.
+2P C.
C2+P .
2P .
2?P ..
2P .+.
2;P .C.
2P ..
;2P ..
;2 P +..
;2+P +..
;2 P ?..
;2+P ?..
;2+P C..
;2?P C..
;2CP ..
;2P ..
;2;P 3& $ 3 " "
0
: & %E%# $ $( " " D $ "
3& $ " "F$/&! 2;P 2?P " 3 : " $ " " . . " " "&)&! 2P ;2;P " 3 : " $ " " .. ..2 . 3 " . " "& & " 3 $& : &2
7 Conclusion " " " ) : 3 " & $ $ & "" 3 )
)" 3! " $$ 2 " " & $ 3! ( & " 3! & 2 = " $ Ã " $ Ã : 3 " &" $"$ ) )) "- 2 " ºm : &" ( 3 $ : &" " )) "- 2 . " 3 & & "" 3 )
)" " $ ) !2 ºm 3 " : &$ 3 / " 3& "- )&"$" &) 2 ºm &" $ $! " " & & " 3 "& & &! && $ ""2
8 Reference = :2 ÃO
? 6B 4$ '" (" 4$ & ! 472c .Q& .%3 +.2+?? # O2 2 &2Ã
6#( " $# " 4#" " $72 +c 3^ 2 ) .2?? # 22 &2 Ã ; 6 #" #( " H& # " # $ <" $ !3&" # " # $ 72 &456 3 c . . ) .2C?;CC+ # O22 Ã
? 6H : $ & : & # " 7 c cd 8 2Q&2;.2;C #$ 2 &2 Ã
6@ $ # # & ) % ) : " @" 3 3& "@&"= "72 c d Q& .2 C 8 && ! :2 Ã 6 4" " &"$ & ( &72 9+)d 3! #2= - &&.O2# .:28 && ! *2*3".
^ # -d .0
8& 2 &2 Ã: $
? 6 $ & H3! 4$ & * & "72 c Q& .%3 . " ? + 8" :22 &2 Ã 6 H $ & H $ #( $ !3&" # $ "72 &454 3 c 2 . ) . 2C?C 8" D22 Ã
C 6# " " # $ #( " !" 3 " $ : $7 cd cd 8 2 Q&2.%2+.2++;+? 3 ! " 2 &2Ã
C 6H &!"" O " # %" " #&&$ & 72 c 3^22 ; O2O2 &2 Ã 6 ( & % ) # " : # " D !" 72 $
" &# !". ) . 2 C; C; O2O2 Ã 6 $# $ #( " !" # " '" ! $$# $ 3& $" $ 72c cd 8 2Q&2 .%2+.2 + :2E2 Ã ; 6* ( H3! $ # " 3! * & 4 $F 72c cd 8 2Q&2?. 2 +C : 2 &2 Ã
6 " % & '" : & "$"72 *** 4'2 C : ! 2 &2 Ã ; 6$ $ 4 & $" (" & ) $72 * . 3! 2* 33 2 2 H$ $ "".
*" &*2ÃH&
C6H ( "$ & : &H: 4@! H $72 d# 0d# : &&'( "! / 2 &2Ã
?6H@( & "& H&" : " > "& 72 9 Q&2?C. %2 +. 2+C +C *2 &2 Ã
? 6H "$ & : B D = " # " 472# $ ""# " R5 & "2Q&2C.%2C.2?? ? *2Ã
+6HH&$ "$ &5- )! $ $# $ #( " # " 72 +c 3^ 2 ) .2?; ? E #22 &2Ã
6*& 3 " D $ : # " $ "H '$ H $7 c cd 8 2Q&2. %2+.2 ;