:KDW0DNHVD1RQQDWLYH$FFHQW"D6WXG\RI.RUHDQ(QJOLVK
-RQJPL.LPDQG6X]DQQH)O\QQ
'HSDUWPHQWRI(QJOLVK/DQJXDJHDQG/LWHUDWXUH
.DQJZRQ1DWLRQDO8QLYHUVLW\5HSXEOLFRI.RUHD
NLPMP#NDQJZRQDFNU
'HSDUWPHQWRI/LQJXLVWLFVDQG3KLORVRSK\
0DVVDFKXVHWWV,QVWLWXWHRI7HFKQRORJ\
VIO\QQ#PLWHGX
$EVWUDFW
:H UHSRUW D VHW RI UHVXOWV LQ WKH VHFRQG ODQJXDJH /
DFTXLVLWLRQ RI (QJOLVK SKRQRORJ\ E\ ILUVW ODQJXDJH /
VSHDNHUV RI .RUHDQ 6SHFLILFDOO\ ZH IRFXV RQ VLJQLILFDQW
GLIIHUHQFHV LVRODWHG EHWZHHQ / VSHDNHUV¶ SURGXFWLRQ RI
LVRODWHG ZRUGV LQ (QJOLVK DQG WKHLU SURGXFWLRQ RI WKH VDPH
ZRUGV LQ VHQWHQFH DQG SKUDVDO FRQWH[WV 5HVXOWV LQGLFDWH
VLJQLILFDQWO\ PRUH DFFXUDWH SURGXFWLRQ RI ZRUGV LQ LVRODWLRQ
WKDQLQWKHSURGXFWLRQRIWKHVDPHZRUGVLQSKUDVDOFRQWH[WV
7KH SDUWLFXODU SKRQRORJLFDO SKHQRPHQD IRFXVHG RQ FRQFHUQ
ERWKVWUHVVUHGXFWLRQDQGSODFHPHQW:HDOVRFRQVLGHUVHYHUDO
RWKHU DVSHFWV RI VHJPHQWDO SKRQRORJ\ :H DUJXH WKDW WKH
GLVFUHSDQF\ LQ UHVXOWV REVHUYHG EHWZHHQ WDVNV PD\ DFFRXQW
IRUPDQ\RIWKHVHHPLQJO\GLVSDUDWHUHVXOWVLQGLFDWHGLQRWKHU
VWXGLHV RI / SKRQRORJ\ :H GLVFXVV VHYHUDO SRVVLEOH
H[SODQDWLRQVIRUWKHVHGDWDLQWHUPVRIZKLFKSURGXFWLRQWDVN
PRVWFORVHO\SURYLGHVDPHDVXUHPHQWRIGHYHORSLQJOLQJXLVWLF
FRPSHWHQFHDQGZKLFKPLJKWUHIOHFWWKHUROHRIHLWKHUJHQHUDO
OHDUQLQJ VWUDWHJLHV RYHUJHQHUDOL]DWLRQ RU UHYHUVLRQ EDFN WR
WKH / JUDPPDU XQGHU FRQGLWLRQV RI VWUHVV RU ZKHQ WKH /
JUDPPDULVQRWIXOO\GHYHORSHG
,QWURGXFWLRQ
:KDW PDNHV D QRQQDWLYH DFFHQW" 6HYHUDO IDFWRUV PD\
FRQWULEXWHWRWKLV$QDFFHQWPD\LQYROYHVHJPHQWDOLQVHUWLRQ
GHOHWLRQ DQGRU VXEVWLWXWLRQ 2Q WKH RWKHU KDQG DQ DFFHQW
PD\EHGXHWRGLIIHUHQFHVEHWZHHQWKH/DQGWKH/SURVRGLF
SDWWHUQV ZKLFK PD\ LQ WXUQ UHIOHFW GLIIHUHQFHV KDYLQJ WR GR
ZLWKGXUDWLRQDPSOLWXGHDQGRUSLWFK$PRQJWKHVHGLIIHUHQW
SRVVLELOLWLHVPRVWUHVHDUFKKDV IRFXVHGRQLVVXHVFRQFHUQLQJ
VHJPHQWDO IDFWRUV )HZ VWXGLHV KDYH LQYHVWLJDWHG WKH /
DFTXLVLWLRQ RI VWUHVV DQG SURVRG\ KRZHYHU VHH %URVHORZ
$UFKLEDOG DQG &DUVRQ > @ DQG IHZHU LQYROYH DQ\ DFRXVWLF
DQDO\VHV $V D FRQVHTXHQFH ZH DUH OHIW ZLWK D YHU\
LQFRPSOHWH XQGHUVWDQGLQJ RI ZKDW DFFRXQWV IRU D QRQQDWLYH
DFFHQW
,QWKLVSDSHUZHFRQWULEXWHWRWKHVWXG\RIDFFHQWWKURXJK
TXDQWLWDWLYHDFRXVWLFDQDO\VHVRIDODUJHGDWDEDVHFRQVLVWLQJRI
UHFRUGHGVSHHFKVDPSOHVIURP/VSHDNHUVRI.RUHDQOHDUQLQJ
(QJOLVK DV DQ / 2XU DQDO\VHV IRFXV RQSURVRGLF DVSHFWVRI
WKH / VSHHFK VDPSOHV LQ D V\VWHPDWLF DQG FRQWUROOHG ZD\
:H EHOLHYH WKDW WKH FRQFOXVLRQV VXJJHVWHG E\ WKHVH GDWD
FRQWULEXWH WR RXU XQGHUVWDQGLQJ RI WKH GHYHORSPHQW RI /
SKRQRORJ\
$W RWKHU OHYHOV WKH UHVXOWV RI WKLV FXUUHQW VWXG\ FDQ
XQLTXHO\ LQIRUP VSHHFK HQJLQHHULQJ DSSOLFDWLRQV DV ZHOO DV
RWKHU VXFK HQGHDYRUV 7KH GDWD UHSRUWHG KHUH UHSUHVHQW DQ
H[SDQVLRQ RI D UHVHDUFK SURMHFW WKDW LV FXUUHQWO\ EHLQJ
LPSOHPHQWHGLQDQRQQDWLYHVSHHFKUHFRJQLWLRQV\VWHP>@
7KH/VWXG\
7KH GDWD ZDV FROOHFWHG IURP QDWLYH VSHDNHUV RI .RUHDQ
OHDUQLQJ (QJOLVK DV DQ / DV ZHOO DV VL[ QDWLYH VSHDNHUV RI
(QJOLVK 7KH VXEMHFWV ZHUH UHFRUGHG ZKLOH UHDGLQJ DORXG
SKRQRORJLFDOO\ FRQWUROOHG (QJOLVK WH[WV $OO WKH / OHDUQHUV
VSRNH WKH VWDQGDUG GLDOHFW RI .RUHD 7KH / OHYHO RI
FRPSHWHQFH LQ (QJOLVK YDULHG DOWKRXJK PRVW RI WKH VXEMHFWV
ZHUH DW WKH LQWHUPHGLDWH OHYHO DV GHWHUPLQHG E\ WKHLU
DFDGHPLF VWDQGLQJ LQ D SURQXQFLDWLRQ FODVV 7KH IROORZLQJ
GHVFULEHV WKH PHWKRGRORJ\ FKRVHQ IRU WKH PDMRU SDUW RI WKLV
VWXG\WKDWLQYROYHVVXEMHFWV$VPDOOVXEVHWRIWKHGDWDZDV
FROOHFWHGE\YDULRXVRWKHUPHWKRGVIRUDEDODQFHGVWXG\
7KH PHWKRGRORJ\ RI WKLV VWXG\ FRQVLVWHG RI IRXU SDUWV
)LUVW D QDWLYH VSHDNHU RI (QJOLVK ZDV FKRVHQ DV WKH PRGHO
VSHDNHU 7KLV QDWLYH VSHDNHU ZDV \HDUV ROG PDOH DQG D
QDWLYH RI WKH VWDWH RI 8WDK 86$ +LV GLDOHFW ZDV VWDQGDUG
$PHULFDQPLGZHVWHUQ
7KHVHFRQGSDUWRIWKLVWDVNLQYROYHGWKHUHFRUGLQJRIDOO
WKHVWLPXOLE\WKHQDWLYHVSHDNHUQV7RGRWKLVWKHQVUHDG
WKHH[SHULPHQWDOVWLPXOXVLWHPVLQDTXLWHURRP7KHUHDGLQJV
ZHUHUHFRUGHGDQGGLJLWL]HGDWDN+]VDPSOLQJUDWHXVLQJ
&RPSXWHUL]HG6SHHFK/DEE\.D\(OHPHWULFV,QF
7KH WKLUG SDUW RI WKH WDVN SURWRFRO LQYROYHG GLVWULEXWLQJ
WKH PRGHO VSHHFK ILOHV WR DOO WKH SDUWLFLSDQWV LQ WKH VWXG\ LQ
ERWK &' IRUP DQG DV GRZQORDGDEOH ILOHV IURP DQ LQWHUQHW
ZHEVLWH 7KH SDUWLFLSDQWV ZHUH WROG WR OLVWHQ WR WKH PDWHULDOV
DQG WR SUDFWLFH UHDGLQJ WKHP IRU DQ DYHUDJH RI D ZHHN 7KH
OHDUQHUVZHUHDVNHGWRLPLWDWHWKHPRGHOVSHHFKDVFORVHO\DV
SRVVLEOH ,Q DGGLWLRQ SULRU WR UHFRUGLQJ WKH OHDUQHUV ZHUH
JLYHQ H[SOLFLW SKRQHWLF LQVWUXFWLRQ WKDW IRFXVHG RQ SRWHQWLDO
SURQXQFLDWLRQGLIILFXOWLHVZLWKWKHVWLPXOL
,Q SDUW IRXU WKH QDWLYH VSHDNHUV RI .RUHDQ ZHUH
UHFRUGHG 7KH OHDUQHUV UHDG HDFK VWLPXOXV LWHP LQ D TXLHW
URRP 7KH UHFRUGLQJV ZHUH DJDLQ GLJLWL]HG LQ N+]
VDPSOLQJ UDWH E\ &RPSXWHUL]HG 6SHHFK /DE E\ .D\
(OHPHWULFV([DPSOHVRIWKHVWLPXOXVLWHPVDUHLQ7DEOH
7DEOH([DPSOHVRI6WLPXOL1RWH&RQV
&RQVRQDQWV,QWR,QWRQDWLRQ
3URSHUW\ ([DPSOH
9RZHO E>HD@WE>L@WE>DL@WE>H@WE>D@WE>R@WWE>RX@JKW
&RQV >S@HHS>W@HDW>N@H\>E@HHS>G@HHG>J@HH
6\OODEOH >U@LFH>O@LFHED>WK@ED>WKH@>VS@DLQ>VSU@DLQ
6WUHVV >D@GG>D@GGLWLRQ:+>,@7(K>RX@VHZK>L@WH+>28@6(
5K\WKP
>'HOLYHU@>ERRNV@>)ULGD\@>'HOLYHU@>ERRNVE\@
>)ULGD\@>'HOLYHUWKH@>ERRNVE\@>)ULGD\@
,QWR 7HDFKLQJODQJXDJHVLVKDUGHUWKDQOHDUQLQJWKHP
INTERSPEECH 2004 - ICSLP
8
th
International Conference on Spoken
Language Processing
ICC Jeju, Jeju Island, Korea
October 4-8, 2004
ISCA Archive
http://www.isca-speech.org/archive
10.21437/Interspeech.2004-600