JPH0736898A - Fixed word processing system in chinese/japanese machine translation - Google Patents

Abstract

PURPOSE:To accurately translate a Japanese sentence including a part corresponding to a Chinese fixed word. CONSTITUTION:An estimation rule table registering whether 'teki' (like) is to be arranged immediately after a Chinese fixed word or not correspondingly to the property of the fixed word, the attribute of a center word and an additional condition is previously prepared. A fixed word and a center word are extracted from Chinese words corresponding to words in a Japanese sentence obatained by molpheme analysis (step S1), the property of the fixed word is obtained (steps S2, S3), and if necessary, the attribute of the center word and additional condition are obtained (steps S7, S10). While referring to the estimation rule table based upon the property of the obtained fixed word, the attribute of the center word and the additional condition, the existence of 'teki' is judged (steps S4, S8, S11, S12).

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to Japanese-Chinese machine translation for automatically translating a Japanese sentence into a Chinese sentence using an electronic computer, and in particular, a structural particle "target" added to a Chinese "fix word".
Is related to the technology that enables accurate translation.

[0002]

2. Description of the Related Art Although machine translation between Japanese and English has almost entered the practical stage, there are still insufficient points regarding machine translation between other languages and Japanese.

By the way, since machine translation is largely influenced by the peculiarities of the source and target languages, the technique of Japanese-English machine translation cannot be directly applied to the Japanese-Chinese machine translation which is the subject of the present invention. There are many.

[0004]

Specifically, there is a drawback in that the translation of the so-called "fixed word" in Chinese grammar is not sufficient. This will be described below.

In Japanese sentences, many expressions such as "B (central word) of N (noun class)", for example, "Kimi no unnie", are used.
The limited word N (“Kim”) is called “fix word” in Chinese grammar. In addition to the above-mentioned "N no B", expressions of Japanese sentences corresponding to definite words include "A (adjective) |
There are patterns such as “B”, “N | B”, and “D (verb) | B”. Note that "|" indicates that the left and right words are adjacent to each other.

[0006] Chinese fixed words are often accompanied by a structural particle "target" immediately after that, but since "target" may not be used in some cases, "target" is added to all fixed words. And the translation is of poor quality.

[0007] Further, a semantic classification to be expressed by a fixed word,
For example, there was an attempt to judge the presence or absence of "target" by classification such as "place" and "quantity", but as a result of systematically investigating and organizing actually used Chinese sentences, even if they have the same meaning classification. In some cases, "target" is added and in other cases, this method is not applicable. For example,
"Vegetable price" and "one correct answer" both belong to the meaning classification of quantity, but the former translation is "vegetable price", the latter translation is "one correct answer", and the former is "target" but the latter Is not attached.

As described above, when translating a Japanese sentence including a portion corresponding to a Chinese fixed word, there is a drawback that an accurate translation cannot be performed.

The present invention has been proposed in view of the above points, and an object thereof is to provide a Japanese-Chinese machine translation capable of accurately translating a Japanese sentence including a portion corresponding to a Chinese fixed word. To provide a fixed word processing method in.

[0010]

In order to achieve the above-mentioned object, the present invention uses a Chinese fixed word property, a central word attribute, and an additional condition in a Japanese-Chinese machine translation for generating a Chinese sentence from a Japanese sentence. On the other hand, a definite word and a central word are extracted from the Chinese word corresponding to the word of the Japanese sentence obtained by morphological analysis and the estimation rule table that associates whether or not to put "target" immediately after the definite word. Steps, steps to acquire the properties of the extracted definite words, steps to acquire attributes for the extracted central words, and additional conditions to the extracted definite words and central words as necessary And a step of determining the presence or absence of “target” by referring to the estimation rule table according to the property of the acquired fixed word, the attribute of the central word, and the additional condition.

That is, as a result of researching a vast amount of data on Chinese definite words, the inventor of the present application uses information on definite words and central words on the Chinese side of the translation destination, not on the Japanese language that is the translation source. The present invention has been discovered by discovering that the presence or absence of the structural particle “target” can be clearly distinguished by narrowing down in the order of the property of the word, the attribute of the central word, and the additional condition.

[0012]

In the Japanese-Chinese machine translation system of the present invention, it is possible to preliminarily deal with whether or not "target" is placed immediately after the fixed word for the property of the fixed word in Chinese, the attribute of the central word, and the additional condition. An attached estimation rule table is provided. Then, the fixed words and central words are extracted from the Chinese words corresponding to the words in the Japanese sentence obtained by the morphological analysis, then the properties of the extracted fixed words are acquired, and the attributes of the central words are extracted as necessary. And get additional conditions. After that, the property of the acquired fixed word, the attribute of the central word,
Whether or not there is a "target" is determined by referring to the estimation rule table according to the additional condition.

[0013]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing an embodiment of a Japanese-Chinese machine translation system to which a fixed word processing method in Japanese-Chinese machine translation of the present invention is applied.

Referring to FIG. 1, in this embodiment, a Japanese sentence input unit 1 for inputting a Japanese sentence and an input Japanese sentence are analyzed and distributed to a plurality of independent task modules, and processing results from each module are displayed. A control module 2 that integrates to generate a Chinese sentence, a Chinese sentence output unit 3 that outputs the generated Chinese sentence, a basic sentence processing module 4, a common sentence pattern processing module 5, and a quantity expression processing module 6 that are independent tasks. , An auxiliary word estimation module 7, a fixed word processing module 8, dedicated data 9, 10, 11, 12, 13 used exclusively by each module, and common data 14 commonly used by each module. The dedicated data 9, 10, 11, 12, 13 and the common data 14
Contains various tables and stacks. Further, the module operating as an independent task is not limited to the example shown in the figure, and may be added if necessary.

FIG. 2 shows an example of the estimation rule table 15 used for the processing in the fixed word processing module 8, which is defined for the classification and properties of Chinese fixed words, attributes of central words, and additional conditions. It is associated with whether or not to put the structural particle “target” immediately after the word. The estimation rule table 1
The specific contents of 5 are not limited to the illustrated example.

The operation of the above embodiment will be described below.

First, in FIG. 1, when a Japanese sentence is input from the Japanese sentence input unit 1, the control module 2 analyzes the input Japanese sentence (syntactic analysis, semantic analysis, etc.) and divides it into a basic sentence and a common sentence pattern portion. The basic sentence processing module 4, the common sentence pattern processing module 5, the quantity expression processing module 6, the auxiliary word estimation module 7, and the fixed word estimation module 7 It is delivered to the word processing module 8.

Here, the basic sentence processing module 4 converts the word order according to the Chinese grammar, and the common sentence pattern processing module 5 "is only ~", "is like ~",
A translated word corresponding to a common sentence pattern such as "... Uchini" is generated, a quantity expression processing module 6 performs processing peculiar to expression of a quantity, and an auxiliary word estimation module 7 uses a Japanese case particle in an appropriate Chinese language. Performs processing to convert to auxiliary words. In addition,
Since the processing in these modules is not directly related to the present invention, the details are omitted. For details of these modules, see Japanese Patent Application No. 5-81345.
It is described in Japanese Patent Application No. 5-81346.

The processing in the fixed word processing module 8 which is a characteristic part of the present invention will be described below.

FIG. 3 is a flow chart showing an example of processing in the fixed word processing module 8.

The determinate processing module 8 first uses the result of the morphological analysis of the Japanese sentence in the control module 2,
A fixed word (a word that modifies / limits another word) and a central word (a word that is qualified / limited) are extracted from the information of the corresponding Chinese word (morpheme) (step S1). Note that the fixed words and the central words can be extracted by a normal syntax analysis technique. Here, for example, for the part of the Japanese sentence "his older sister", "he" is extracted as the fixed word, and "sister" is extracted as the central word. It should be noted that the corresponding Chinese word is also known from the dictionary (not shown) used in the morphological analysis. In the present example, the Chinese word for "he" is "other" and the Chinese word for "sister" is "姐姐".

Next, the fixed word processing module 8 acquires the classification (global classification) and the property (fine classification) for the Chinese fixed words extracted above by referring to the dictionary (steps S2 and S3). . Definite term "other" in the above example
Is classified as "pronoun class", and its nature is "personal pronoun".

Next, the fixed word processing module 8 performs matching on the estimation rule table 15 (FIG. 2) based on the classification and properties of the above fixed words (step S4). When the estimation rule table 15 shown in FIG. 2 is collated with the category "pronoun class" and the property "personal pronoun" of the above-mentioned example, it corresponds to the numbers 1 to 4.

Next, the fixed word processing module 8 judges whether or not a plurality of items exist as a result of the collation of the estimation rule table 15 (step S5). In the above example, the number is plural, so the determination is YES.

Next, when it is determined that there are a plurality of collation results (YES in step S5), the fixed word processing module 8
Determines whether the central word attribute is necessary to narrow down the matching result (step S6). In the above-mentioned example, since the attribute of the central word is set in the hit part of the estimation rule table 15 of FIG. 2, it is determined that the attribute of the central word is necessary for narrowing down.

Next, when it is determined that the attribute of the central word is necessary (YES in step S6), the fixed word processing module 8 acquires the attribute of the central word by referring to the dictionary (step S7). The attribute of the central word "姐姐" in Chinese in the above example is "a name for a person".

Next, the fixed word processing module 8 performs matching again by newly adding the attribute of the central word to the estimation rule table 15 (step S8). In the above example, the number 3 in FIG. 2 is applicable because the attribute “name for person” of the central word is added as a condition.

Next, after the above-mentioned re-comparison (step S8) and when it is judged that the attribute of the central word is not necessary (NO at step S6), the fixed-word processing module 8 sets the additional condition for narrowing down the comparison result. It is determined whether it is necessary (step S9). In the above example, the collation result has already been narrowed down to one, so it is determined that the additional condition is not necessary.

As an example of the case where the additional condition is required, there is a case where "his work" is given as a part of the input Japanese sentence. In this case, the Chinese mnemonic is "other" for "he"
Then (as in the example above), the central word is "craft" for "work". In this example, the attribute of “craft” is shown in FIG.
Since there is no corresponding one among the numbers 1 to 4, additional conditions are required to further narrow down.

Next, when it is determined that the additional condition is necessary (YES in step S9), the fixed word processing module 8 confirms the additional condition (step S10). The additional condition is obtained from the attributes of the fixed word and the central word in the semantic analysis using the dictionary. The additional condition in the above-mentioned second example is a "territorial relationship".

Next, the fixed word processing module 8 performs collation again with the addition rule newly added to the estimation rule table 15 (step S11). In the above-mentioned second example, the condition that the “territorial relationship” is added as a condition,
This corresponds to number 4 in.

Then, the above re-collation (step S1
1) If a plurality of items do not exist (NO in step S5) and if it is determined that the additional condition is not necessary (NO in step S9), the slang word processing module 8 determines from the matching result of the estimation rule table 15 The presence or absence of "target" is confirmed (step S12). In the above-mentioned first example, "target" is absent because number 3 corresponds, and in the second example, "target" is present because number 4 corresponds.

Finally, the control module 2 integrates the processing results returned from the modules 4, 5, 6, 7, and 8 to generate a completed Chinese sentence, and the Chinese sentence output unit 3
Output via. For the above-mentioned first example "His elder sister", "other elder sister" is generated, and for the second example "His work", "other work" is generated.

[0035]

As described above, in the Japanese-Chinese machine translation system of the present invention, the information of the fixed word and the central word on the Chinese side of the translation destination is used instead of the source Japanese language. Since the presence or absence of the structural particle “target” is determined in the order of the property of the fixed word, the attribute of the central word, and the additional condition, the quality of translation related to the fixed word can be improved.

[Brief description of drawings]

FIG. 1 is a configuration diagram showing an embodiment of a Japanese-Chinese machine translation system to which a determinate processing method in Japanese-Chinese machine translation of the present invention is applied.

FIG. 2 is a diagram showing an example of an estimation rule table.

FIG. 3 is a flowchart showing an example of processing contents of a fixed word processing module.

[Explanation of symbols]

 1 ………… Japanese sentence input unit 2 ………… Control module 3 ………… Chinese sentence output unit 4 ………… Basic sentence processing module 5 ………… Common sentence pattern processing module 6 ………… Quantity expression Processing module 7 ………… Auxiliary word estimation module 8 ………… Definite word processing module 9 to 13… Dedicated data 14 ……… Common data 15 ……… Estimation rule table

Claims (1)

[Claims] 1. In Japanese-Chinese machine translation for generating a Chinese sentence from a Japanese sentence, with respect to the characteristics of Chinese fixed words, attributes of central words, and additional conditions,
An estimation rule table that associates whether or not "target" is placed immediately after a fixed word, and a step of extracting fixed words and central words from Chinese words corresponding to words in Japanese sentences obtained by morphological analysis; , A step of acquiring the properties of the extracted fixed words, a step of acquiring attributes for the extracted central words as necessary, and a step of acquiring additional conditions for the extracted fixed words and central words as necessary And a process of determining whether or not there is a "target" by referring to the estimation rule table based on the acquired property of the nominated word, the attribute of the central word, and the additional condition. method.