JPH03102568A - System for machine translation from japanese to chinese - Google Patents

Abstract

PURPOSE:To enable machine translation from Japanese to Chinese by storing idiomatic syntaxes in a Chinese grammar investigating means. CONSTITUTION:A translated word selection learning means deletes a connected nodifier. A Japanese part of speech applying means 3 applies a part of speech to a Japanese sentence deleting the connected word while referring a Japanese dictionary file for each word. A part of speech rearranging means 4 rearranges parts of speech to the Japanese sentence, to which the parts of speech are applied, according to a Chinese grammar stored in the means 4 itself. A translating means from Japanese to Chinese executes translation to Chinese while referring a Chinese knowledge data base file CKDBF in respect to this arrangement. A Chinese grammar investigating means 6 executes syntax analysis according to the grammar concerning a translated Chinese sentence. A converting means 7 to the Chinese grammar executes grammar conversion to the Chinese sentence, which syntax is analyzed, and outputs the idiomatic syntax. A Chinese output means 8 outputs Chinese characters (KANJI) while referring a Chinese character pattern file CCPF and an alphabet/figure/symbol character pattern file ANSCPF.

Description

[Detailed description of the invention] [Industrial application field] The present invention relates to a machine translation method from Japanese to Chinese.

[Conventional technology]

Conventionally, as a machine translation method between different languages, "Information Processing J 19B5 VOL 26 No. 10" published by Information Processing Society of Japan has been used.
As introduced in , machine translation methods between English and Japanese are known, but there has been no machine translation method from Japanese to Chinese.

[Problem to be solved by the invention]

As mentioned above, in the past, machine translation from Japanese to Chinese was not possible, so there was a problem in that we had to rely exclusively on human translation.

[Means to solve the problem]

Japanese to Chinese machine of the present invention {llN. The translation method consists of a Japanese language input means that inputs Japanese sentences in Hiragana, Katakana, or Romaji and converts them into sentences containing Kana and Kanji, and a translation selection learning means that deletes collocation modifiers from the sentences containing Kana and Kanji. , a Japanese part-of-speech assigning means for assigning a part of speech to a kana-kanji mixed sentence from which the collocation modifier has been deleted, and a part-of-speech arrangement for rearranging the parts of speech according to Chinese grammar for a kana-kanji mixed sentence to which the part of speech has been assigned. a Japanese-to-Chinese translation means for translating the kana-kanji-containing sentence in which the parts of speech have been rearranged into a Chinese sentence; A Chinese grammar checking means for outputting, a Chinese grammar conversion means for converting the output of the Chinese grammar checking means into Chinese grammar, and a Chinese grammar converting means for converting the converted output into a code to produce a Chinese sentence. It is characterized by having an output means.

〔Example〕 Next, the present invention will be explained with reference to the drawings.

Figure 1 is a flowchart showing the most effective machine translation method of the present invention, which includes 1,000 steps of Japanese language input, 2,000 steps of translation word selection learning, 3,000 steps of Japanese part-of-speech addition, and 1,000 steps of Japanese part-of-speech addition. It consists of a rearrangement means 4, a translation means 5 from Japanese words to Chinese, a Chinese grammar research means 6, a conversion means 7 to Chinese grammar, and a Chinese output 1000 steps 8.

Figures 4 to 6 are examples of Japanese character pattern files (hereinafter referred to as JPF), Figure 7 is an example of alphabet/numeric/symbol character pattern files (hereinafter referred to as ANSCPF), and Figure 8 is Japanese Terminology file [hereinafter J
Figure 9 is an example of a Chinese character pattern file (hereinafter referred to as CCPF).
13 illustrates an example of a Chinese knowledge database file (hereinafter referred to as CKDBF).

In Figure 1, a Japanese word processor is used as the Japanese input means l, and a Japanese word processor is used as the translation selection learning means.
Translation selection learning method by application (application number 6).-077
639) to delete collocation modifiers.

The Japanese part-of-speech assigning means 3 assigns a part-of-speech to the Japanese sentence from which collocations have been deleted by referring to a Japanese dictionary file for each word. The Japanese dictionary file is Sanseido Publishing's Japanese dictionary (
(3rd edition).

The part-of-speech rearrangement part 4 rearranges the part-of-speech for the Japanese sentence to which the part-of-speech has been assigned according to the Chinese grammar that it has memorized, and the Japanese to Chinese translation means 5 performs C Translate into Chinese by referring to KDBF.

Chinese Grammar Survey Sendan 6 analyzes the translated Chinese sentences using grammar. It also remembers common JTJ All sentences in preparation for Chinese sentences that cannot be parsed. Middle-5? The language grammar conversion means 7 converts the parsed Chinese sentence into grammar, or outputs a conventional sentence. Chinese output 1,000 steps 8 uses a Japanese word processor, and CCPF
and ANSCPF to output Chinese characters (Chinese characters).

Next, a specific example of machine translation will be explained.

"Hiragana" from the Japanese keyboard of Japanese input method 1
So, I rode the plane from Shiyanbai to Pekin. If you input ``'', it will be converted to Japanese character code using JPF and ANSCPF using Japanese input 1,000 steps as shown below.

wa = J n 44 J mtt r ta→Jn+sJm
+a, L/→Jn. +zJm+, is → JnzaJm2s
＋ し→Jn+2Jm、■、や→Jn3eJrn3e
+→Jnt6Jm+6+ is→Jn26Jm2+z→
Jn2Jm2, ka→JnaJms,ra→JI1soJm
39+pe→J n 7 oJ m t O Iki→J
n 7 J m r gn→JntsJm<6+ ma→
Jnz101? m 3+ 1→Jn6oJmeo+hi→J n
27 J m 27 1ko → Jn lo J m to
, U→JrzJm3, ki→JntJmx→Jn
2■Jm2z+→JnzsJm2■, Tsu→J n +
a Jm 28 1 → Jn+aJm+a+ →An
34A m 3 1 When this "Hiragana ゜' sentence is converted into Japanese Kanji using JHCF, it becomes as follows.

Jn++Jmi4Jn+6Jm+aJnt■Jrn+2
(I) −+Kn<aKm+3 (I), J n2s
Jm2e (ha)→jiao conversion, JnuJmuJn3eJm
36Jni6Jm4eJn2sJm2aJn2Jm2(
Lyan yes) → Kn3Kms+Kn3zKm32 (Shanghai), J n6JmaJ n3oJm3p (from) → No conversion, JnyoJm+oJntJmtJn4eJmu (
→K n 4 K m 4K n s K
m 5 (Beijing) r Jn31Jm31JnsoJmeo
(until) → No conversion, Jn2■Jm2tJ'n+oJr
n+oJ nxJm3JnrJmy (plane) -+
K n12Kmt■K n +sKm+aKn+sKm
+, (airplane), J'n22Jm2■ (ni) → Asahi conversion,
Jn2sJm2sJnuJmuJn+sJm+s (mounted)''K n 3gKmg9J n ls J m+a
J n reJm+e (ride), An3+Arr
z4 (.) → No conversion In the end, [I took a plane from Shanghai to Beijing. ] becomes. Next, while Japanese requires a conjunctive modifier, Chinese does not necessarily require a conjunctive modifier. When the conjunctive modifier is deleted using selective learning method 2, it becomes “I took a plane from Shanghai to Beijing.”゜ (see Figure 2), and a Japanese dictionary file is used for each word to assign a Japanese part of speech. By giving the part of speech {=J using method 3, I took a plane from Shanghai to Beijing.

Noun Noun Case particle Noun Adverb Noun Verb, and then the parts of speech rearrangement means 4 match the parts of speech of Chinese grammar.

The difference in the arrangement of parts of speech in Japanese and Chinese is shown in Figure 14.

When rearranged according to the rules shown in Figure 14, I took a plane from Beijing to Shanghai.

Noun Case Particle Noun Adverb Noun Noun Verb Then, using CKDBF, this Japanese sentence is translated by the translation means 5 which translates it into Chinese, as follows.

KnuKm43 (me) →C n 3e C m3s
(I), JneJmaJn3eJm3e (from) →
C n 5 cm s (Jyk) +Kn3+Km
3+Kn32Kmu (Shanghai) ″Cn2oCm2oC
nnCmu (Shanghai) r J n3+ J m3+
J nso J mso (until) →Cn7Cmy (
to) r Kn+4Km+aKn+sKm+5 (Beijing)
→Cn 20m20n+50m+5 (Beijing), Kn+2
Km+2Kn+eKm:+aKn+oKm+o (airplane) →CnvCmeCn+2Cmu (X machine), Kn
3, Km3oJn+eJm+aJn+eJm+e (got on) → Cn2aCm2aCnzeCmu (sat finished),
An3tAm34 (.) → No conversion (Ana<Arn3t
=． ) In other words, “My situation has arrived in Shanghai and I am sitting in Beijing.

191, and the syntax of this Chinese sentence is analyzed by the Chinese grammar examining means 6.

Here, we will introduce the main forms of construction with reference to Chinese grammar.
It is shown in Figure 5.

If the basic constructions shown in Figure 15 are roughly divided into Chinese grammatical structures, there are three types (simple constructions, compound (embedded) constructions, and compound (combined) constructions).

It is said that about 90% of the above-mentioned grammatical structures (three types) can be analyzed, and since idiomatic structures etc. cannot be analyzed, they must be stored in the Chinese grammar research means 6. Examples of idiomatic constructions include constructions that express natural phenomena such as ``low rain, low snow, low fog, shellfish, lightning strikes, and fog''.

Now, when the sentence of this example is analyzed, it becomes as shown in Figure 3. Next, the difference between Chinese and Japanese grammar is the relationship between the predicate (寞语) and the verb (■逼). This relationship is shown in FIG.

When the conversion means 7 to Chinese grammar in Figure 1 converts according to the rules shown in Figure 16, =10-゜'I arrived in Shanghai and sat in Beijing.

Therefore, after grammar conversion, CCPF and ANSCPF
is used to output a Chinese sentence with 1,000 stages of Chinese output.

For example, Cn3aCrn+s (I), CnsCm5 (8), C
n2oCm2oCn. Cm++ (Shanghai), CnyCm
+ (arrived), C n 2 C m 2 C n l 5
C m + s (Beijing) r CnzaCm2sC
n2. C m 29 (sitting), Cn9Cm9Cn+
2Cm+2() desk), A. n34Amx+ (.) When this Chinese character code is output in Chinese characters, it becomes '゛I'm in Shanghai and I'm sitting in Beijing.

Therefore, if you enter a Japanese sentence in Japanese hiragana, it will automatically be machine translated into Chinese.

〔Effect of the invention〕

The present invention has the above-described structure, in particular, the structure in which idiomatic constructions are stored in the Chinese grammar research means, so that machine translation from Japanese to Chinese is possible.

[Brief explanation of drawings]

Figure 1 is the flowchart of the present invention, Figure 2 is a diagram showing an example of translation word selection learning, Figure 3 is a diagram showing an example of syntactic analysis of a Chinese sentence, and Figures 4 to 6 are Japanese character patterns. File(
JPF=Japanese Pattern File
), Figure 7 is a 7 rufabe l-/number/monthly character pattern file (ANSCPF = A.Ipha).
bet/Number/Symbolic Chara
cter Pat. Figure 8 shows the Japanese terminology file (JHCF=Ja.pan
ese Ha. nclerCharacter Fi
Figure 9 is a diagram showing an example of terminology for Chinese Character Pattern File (CCPF).
erPattern File), and Figures 10 to 13 are Chinese Knowledge Database Files (CK
DBF =ChineseKnowledge Dat
a Base File), Figure 14 is a diagram showing the difference in the order of parts of speech in Japanese and Chinese, Figure 15 is a diagram showing an example of the syntax of a Chinese sentence, and Figure 16 is a diagram showing the Chinese edict and Japanese sentence. It is a diagram showing the relationship between predicates and verbs in wooden language. ■...Japanese input means, 2...Translation selection learning means, 3...Japanese part-of-speech assigning means, 4...
... Means for rearranging parts of speech, 5 ... Means for translating from Japanese to Chinese, 6 ... Means for investigating Chinese grammar, 7
... - Conversion means to Chinese grammar, 8... Chinese output means.

Claims (1)

[Scope of Claims] Japanese input means for inputting Japanese sentences in "hiragana", katakana, or Roman letters and converting them into sentences containing kana and kanji; and translation word selection learning means for deleting collocation modifiers from the sentences containing kana and kanji. , a Japanese part-of-speech assigning means for assigning a part of speech to a kana-kanji mixed sentence from which the collocation modifier has been deleted, and a part-of-speech arrangement for rearranging the parts of speech according to Chinese grammar for a kana-kanji mixed sentence to which the part of speech has been assigned. a Japanese-to-Chinese translation means for translating the kana-kanji-containing sentence in which the parts of speech have been rearranged into a Chinese sentence; and a Japanese-to-Chinese translation means for translating the sentence containing kana and kanji in which the parts of speech have been rearranged into a Chinese sentence; a Chinese grammar checking means for converting the output of the Chinese grammar checking means into Chinese grammar; and a Chinese output means for converting the converted output into a code and outputting a Chinese sentence. A machine translation method from Japanese to Chinese characterized by having the following.