英和機械翻訳における文脈依存性

(1)

Dependence on Context

in case of English-Japanese Machine Translation I ─14

Katsuyuki Shibata^1）

（ReceivedMay31,2011）

英和機械翻訳における文脈依存性

I

- 14

柴田　勝征^1）

（平成 23 年 5 月 31 日受理）

Abstract

This is the fourteenth of a series of articles on the context dependency analysis in case of English-Japanese machine translation system which we call “US system”. In this issue we examine the examples taken from Lesson 14 of three English textbooks for the first year grade Japanese junior high school students; namely, Sunshine 1 (Kairyudo), New Crown 1 (Sanseido) and New Horizon 1 (Tokyo Shoseki).

The main subjects to be examined in this issue are the followings: (1) Whether “be” should be translated into “dearu” or “iru”? (revisited) (2) Whether the definite article ‘the’ should be translated as ‘sono’ or to be neglected when it is followed by an adjective? (3) Selection among “hito”, “otoko”

and “otoko no hito” for the noun “man”, (4) An example case where “that” is translated as “sono”. (5) An example case where “there are” is translated into “ga iru”. (6) Add a counting unit to a numeral (revisited).

Keywords:machinetranslation,contextdependence,contextinheritedfromtheprecedingsentences,USsystem.

1） DepartmentofAppliedMathematics,FacultyofScience,FukuokaUniversity,8-19-1Nanakuma,Jonan-ku,Fukuoka, 814-0180,Japan

福岡大学理学部応用数学科，〒 814-0180福岡市城南区七隈 8-19-1 1. Whether “be” should be translated into “dearu”

or “iru”? (revisited)

Inthefirstarcle[1]ofthisseries,weexplained thatverb“be”isusedasacopula(thecorresponding Japanese is“dearu”) as well as“exist”(the correspondingJapaneseis“iru”).Andwepresented

thenachoicerulebelowwhichselectsthecopula

“dearu”;

33; いる; 2; T0=v; E0<>be ; T0<>?; J-3<>7; TA<>ah;

PJS<>* にいますか？; G635;

confirmingthattheprecedingJapanesetranslation

(2)

doesnotcontain“nii-masu-ka?”

Thistimeweexhibitanotherchoicerulewhich also select the copula“dearu”with a little bit of differentcheckingcondition.

I’minLondon.

私はロンドンにいます。

WatashiwaRondonnii-masu.

Thereisaparadetoday.

今日パレードがあります。

Kyoupareedogaari-masu.

Therearealotofpeopleonthestreet.

通りに多くの人びとがいます。

Tooriniooku-nohito-bitogai-masu.

Excuseme.

すみません。

Sumimasen.

Areyouwaitingforaparade?

あなたはパレードを待っていますか？

Anatawapareedowomatte-i-masuka?

Yes,Iam...(1) はい、待っています。

Hai,matste-i-masu.

For the“be”(“am”) in example sentence (1) above,thepossibilityof“existence”(iru”)isrejected bythefollowingchoicerule

33; いる; 2; T0=v; T0<>jB; J-3<>7; JZ<>7; DJA<>どこに ; TZ<>gB; VR<>いる; J415;

Theruleabovechecksamongotherthingsthat the Japanese for the main verb in the previous sentence was not“iru”(existence). In fact the context data for the preceding sentences are restoredas;

sj = 1 人びと sz = hXz oj = oz = preoj = 1 通り preoz = d+cBy 本 vr = 待つ vz = ?Gui

sj2 = 1 パレード , sz2 = y 列 , oj2 = , oz2 = , preoj2

= 1 ロンドン , preoz2 = OKgB, vr2 = いる , vz2 = OEau

TheJapaneseforthemainverbintheprevious

sentenceisseentobe“matsu”(“wait”)andthe Japanese for the verb in the sentence before the precedingsentenceis“iru”(“exist”).

2. Whether the definite article ‘the’ should be translated as ‘sono’ or to be neglected when it is followed by an adjective?

Inoneofourpreviousarticles[2],weexplained that our US Translation System has about 520 grammarrulesforthetranslationofanounphrase headed by the definite article, which determine whetherthearticle‘the’shouldbetranslatedas‘sono’

orneglectedaccordingtothecontext.

Thistimeweshowasimilargrammarrulewhich translatesanounphrasewithanadjectivebetween thedefinitearticleandanoun;

Whereisourcar?

私達の自動車はどこにありますか？

Watashitachinojidoushawadokoniari-masu ka?

It’soverthere.

それ（その自動車）はあちらにあります。

Sore（sonojidousha）waachiraniari-masu.

Let’sgotothe next place...(2) 次の場所に行きましょう。

Tsugi no bashoniiki-mashou.

ThegenerationruleforthisJapanesetranslation ofthenounphrasein(2)aboveisasfollows.

2; 851; 1; 0; E0=the; E0<>the ; E0<> the; DJ1<>まさに ; E2<>the ; DJ2<>SJ $; DJ2<>OJ $; DJ1<>SJ $; DJ 1<>OJ $; J0#Z; 5520;

Therulechecksamongotherthingsthatneither thenounnortheadjectiveisnotcontainedinthe subjectnorintheobjectoftheprecedingsentence.

Iftheseconditionsareverified,theruleannihilates the Japanese for“the”. Note that Japanese languagehasnoarticleasgenerallyusedinEnglish.

Something similar to English article is used only when the noun in question should be stressed in somereason.

(3)

3. Selection among “hito”, “otoko” and “otoko no hito” for the noun “man”.

English noun“man”means“a male person”as well as“a human being in general”. These two meaningscorrespondtodifferentJapanesewords

“otoko”and“hito”. Further, we have a polite expression“otokonohito”for“amaleperson”.See thefollowingexample;

Alice:Areyouanegg,Humpty?

アリス「あなたは卵ですか、ハンプティー？」

Arisu“Anatawatamagodesuka,Hanputii?”

Humpty:No.I’maman...(3) ハンプティー「いいえ。私は人です。」

Hanputii“Iie.Watashiwahitodesu.”

ThechoicerulefortheJapanesecandidatesfor

“man”intheexamplesentence(3)aboveisthis:

11; 男; 2; E0=man; LT; E-1=a; J-1=8; T-2=v; E- 3=I; BS=*Are you an ; 5705;

The rule checks that the“man”in question is embedded in a sentece as“I am a man”and the previoussenteneis“Areyouan....?”Thentherule rejectsthepossibilityformentionningthegenderof the“man”.

4. An example case where “that” is translated as

“sono”.

Inthefiftharticle[2]ofthisseries,weexplained thatwhen‘that’precedesanountomodifyit,weare confrontedwiththeproblemofchoosingbetween

‘ano’ and ‘sono’ for such ‘that’. We showed there thatifthenouninquestionisexplicitlyseenbythe peopleindiscussion,theJapanese“ano”isadequate for“that”. This time we show the case where Japanese“sono”isadequatefor“that”;

Alice:Pleaselistentothissong,Humpty.

アリス「この歌を聞いてください、ハンプティー。」

Arisu“Konoutawokiitekudasai,Hanputii.”

HumptyDumptysatonawall,HumptyDumpty hadagreatfall.

ハンプティー＝ダンプティーは壁に座りました、

ハンプティー＝ダンプティーはどしんと転げ落ちました。

Hanputii-Danputiiwakabenisuwari-mashita, Hanputii-Danputii wa doshin-to koroge-ochi- mashita.

Humpty:Oh,no!Idon’tlikethatsong...(4)

ハンプティー「まさか！私はその歌が好きでは

ありません。」

Hanputii“Masaka! Watashi wasono^{uta ga} suki-de-waari-masen.”

TheJapanesefor“that”appearingintheexample sentence(4)isselectedbythefollowingchoicerule;

88;*その; 2; E0=that; E1=song; BS2=*isten to this song; v440;

The rule above verifies that“that”in question modifies“song”which follows it and the English sentencebeforetheprecedingsentencecontainsthe charactersequence“istentothissong”.Inthatcase,

“thatsong”isinterpretedtomentionthesongwhich wasmentionedearlier,andconsequently“that”is translatedtobe“sono”.

5.Anexamplecasewhere“thereare”istranslated into“gairu”.

Inthesixtharticle[3]ofthisseries,weexplained that when there are/is (a) human being(s) or animal(s),wesay“gairu”inJapanesewhilethe existenceofinanimate(s)orplant(s)isexpessedas

“gaaru”.Thereweshowedthecasewhere“gaaru”

waschosen.Thistimewepresentthecasewhere“ga iru”shouldbeselected.

Ken:Therearesomestrangeanimalsinthezoo.

健「動物園に（いく頭かの）珍しい動物がいます。」

Ken“Doubutsu-enni（ikutou-kano）mezurashii doubutsugai-masu.”

(4)

Mike:Arethereanybeavers?

マイク「ビーバーがいますか？」

Maiku“Biibaagai-masuka?”

Ken:Yes,there are...(5) 健「はい、います。」

Ken“Hai,i-masu.”

Mike:Arethereanykoalas?

マイク「コアラがいますか？」

Maiku“Koaragai-masuka?”

Ken:No,therearen’t.

健「いいえ、いません。」

Ken“Iie,i-masen.”

Mike:Howmanyelephantsarethere?

マイク「象は何頭いますか？」

Maiku“Zouwanantoui-masuka?”

Ken:Therearetwo...(6) 健「２頭います。」

Ken“Ni toui-masu.”

Thechoicerulefor“thereare”in(5)aboveisas follows;

33; *u; 1; T0=E; FT; J2=0; T1<>y; SO=au; J855;

Thisrulechecksthattheverbinquestionis“there are/is”(T0=E;)atthetopoftheEnglishsentence (FT), and that the attribute sets of the subject andtheobjectoftheprecedingEnglishsentence containseither‘a’(animal)or‘u’(human).When theseconditionsaresatisfied,therulechosesthe Japanesecandidatewith‘u’attribute.

Infactthecontextdataisrecordedasfollows;

sj=1 象sz=?xay 頭oj=oz=preoj=1 動物園preoz=d+By つvr=いるvz=OEau

sj2=1 コアラ ,sz2=?xay 匹 ,oj2=,oz2=,preoj2

=,preoz2=,vr2=,vz2=rvr=irurv2=

Thesubjectoftheprecedingsentenceis‘elephant’

andthesubjectofthesentencebeforethepreceding sentenceis‘koala’.Bothhaveevidentlytheanimal attribute‘a’.

6. Add a counting unit to a numeral (revisited).

Intheeitharticle[4]ofthisseries,weexhibited examplesentences;

Ken:Howmanyracketsdoesshehave?

Mike:Shehasonlyone.Itisold.Shewantsa newracket.

andexplainedthatinAsianlanguagesasJapanese, Chinese and Korean, we use counting units for concretecountingthingslikehumans,books,houses, fruits,etc,whileinEuropeanlanguagesasEnglish, French, German, Russian, etc, they don’t attach countingunitstonumeralswhentheycountthings.

Consequetly,wehavetoaddacountingunittothe numeralwhenwetranslateanEnglishsentenceinto Japanese.Incasethenumeralisdirectlyfollowed byacountablenoun,theattributeinformationofthe nounregisteredinthesystemdictionarygivesthe adequatecountingunitforthenoun.Butsometimes thenounphraseattheendofasentenceisomitted andthesentenceendswithanumeralasshownin thesecondsentenceabove.

This time, see the last sentence (6) in the precedingsection.TheJapanesefornumeral“two”

isprocessedbythefollowinggrammarrule.

5; 31; X; 1; LT; T1=n; T0=E; T1<>yz; SZ=?; SZ=y; OZ

<>?; J0#-格; J0#VR; J0#DES;

T1#+SZ $; T1#KAZ1; J1#KAZ2; J1#-の: no; E865;

Thecommand“T1#+SZ $”appendstheattribute set(SZ)ofthesubject(‘elephant’)ofthepreceding sentence to that of the first component of the numeral(‘two’).Thecommand“T1#KAZ1”picks up the counting unit ‘tou’ from the appended information,andthecommand“J1#KAZ2”attaches that‘tou’tothenumeral“ni”andmakesitinto“ni tou”.

References

［１］ Katsuyuki SHIBATA: Dependence on ContextincaseofEnglish-JapaneseMachine TranslationI-1.FukuokaUniversityScience

(5)

Reports,vol.35No.1.pp.75-79,2005

http://www1.rsp.fukuoka-u.ac.jp/chosho/

Cntxt1-1.html

［２］ KatsuyukiSHIBATA:DependenceonContext incaseofEnglish-JapaneseMachine

TranslationI-5.FukuokaUniversityScience Reports,vol.37No.1.pp.93-103,2007

Cntxt1-5.html

［３］ KatsuyukiSHIBATA:DependenceonContext incaseofEnglish-JapaneseMachine

Cntxt1-6.html

［４］ KatsuyukiSHIBATA:DependenceonContext incaseofEnglish-JapaneseMachine

Cntxt1-8.html