JPH11143876A - Bidirectional machine translation device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To facilitate a translation processing between two languages by a dictionary and a single analysis program. SOLUTION: This device consists of two sets of language processing means 10 and 11 for executing each processing of structure analysis of an input language sentence, a conversion into an output language and generation/outputting of an output language sentence in accordance with an object-oriented programming; and the input language of one kind is translated into the output language of the other kind and, at the same time, the input language of the other kind is translated into the output language of one kind.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a bidirectional machine translation apparatus for performing bidirectional translation between different kinds of natural languages.

[0002]

2. Description of the Related Art Various automatic translators using translation software have been proposed as systems for automatically translating, for example, English to Japanese or Japanese to English.

For example, as shown in FIG. 5, such a conventional automatic translator is connected to an input unit 1 and uses grammatical information and dictionary information from an English analysis dictionary 2 to input English as an input language. It has an English analysis unit 3 for analyzing the grammatical structure of a sentence.

The English-language analyzing unit 3 uses the grammatical information and the dictionary information from the English-Japanese conversion dictionary 4 to convert the result of the syntax analysis into a translation destination language. It is connected.

The English-Japanese conversion unit 5 is connected to a Japanese generation unit 7 for generating an output sentence from the conversion result with reference to the Japanese generation dictionary. Is output from the output unit 8.

Therefore, in such a conventional automatic translator, the English analysis unit 3 analyzes the grammatical structure of an English sentence as an input language input from the input unit 1.

For this reason, here, the English sentence, which is merely a symbol string, is decomposed into words, and the basic form of the English sentence is obtained while referring to the English analysis dictionary 2.

[0008] This English analysis processing is called morphological analysis, and is executed as preprocessing for syntax analysis.

Next, the English-to-Japanese conversion unit 5 uses the grammar information and the English-to-Japanese conversion dictionary 4 to convert the result of the syntax analysis into Japanese, which is the output language, which is the translation destination language.

On the basis of the data structure generated by the conversion process, a Japanese language generation unit 7 executes a Japanese language generation process using a Japanese language generation dictionary 6, and obtains a Japanese language as a final result. The sentence is output from the output unit 8.

This Japanese generation processing is the reverse of morphological analysis, and determines the form of each word from the basic form of the word and the accompanying information included in the result of the above conversion processing.

Thereafter, the word order and the like are determined according to the tree structure information and rules included in the conversion result, and a final Japanese sentence is created and output.

[0013]

However, in such a conventional automatic translator, for example, when performing each translation from English to Japanese and from Japanese to English, the above-described syntax analysis and conversion processing are performed. , Three types of subsystems for language generation are separately required for each of the English and Japanese translation processes, and therefore, since the device needs to be composed of six types of modules, the device management is extremely difficult. There was a problem that it became complicated.

According to the above-described automatic translation procedure, tens to hundreds of grammar rules are required, and the results obtained may differ depending on the order in which these rules are applied when executing each process. There was a problem that.

If the results are to be controlled in a unified manner, the analysis program becomes more complicated.

The present invention solves the above-mentioned conventional problems. A module for performing a translation process between two languages using a dictionary, a grammar rule, or the like includes, for example, two modules for English data and Japanese data. It is an object of the present invention to provide a bidirectional machine translation device which can easily add and modify dictionary data, and can easily execute a translation process between two languages by a dictionary and a single analysis program.

[0017]

In order to achieve the above object, a bidirectional machine translation apparatus according to the first aspect of the present invention provides a syntax analysis, conversion into an output language, and output of an input language sentence according to object-oriented programming. It comprises two sets of language processing means for executing each process of generating and outputting a language sentence, translating one type of input language into another type of output language, and converting the other type of input language into the one type of Is translated into the output language.

The two-way machine translation apparatus according to the second aspect of the present invention is such that one type of language is English and the other type is Japanese.

[0019]

DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram conceptually showing a bidirectional machine translation apparatus according to the present invention. Is an input unit for inputting a language such as Japanese or English.

For example, in the case of English-Japanese translation, the English sentence input to this input unit 1 is parsed, and the information obtained therefrom is sent to a Japanese analysis / synthesis unit 11 for Japanese-English translation. In the case of (1), an English analysis / synthesis unit as a language processing means for synthesizing and outputting an English sentence using the information sent from the Japanese analysis / synthesis unit 11.

In the case of Japanese-to-English translation, the Japanese language analyzer 11 parses the Japanese sentence input to the input unit 9 and sends the obtained information to the English analysis / synthesis unit 10. In the case of (1), it is a Japanese analysis / synthesis unit as another language processing means for synthesizing a Japanese sentence using the information sent from the English analysis / synthesis unit 10.

Reference numeral 12 denotes an output unit for an output language.

The parsing of the English sentence and the parsing of the Japanese sentence are executed with reference to the English data and the Japanese data, respectively.

In the present invention, the above-mentioned English analysis / synthesis unit 10
In addition, the Japanese analysis / synthesis unit 11 is divided into two modules not in units of processing but in units of language.

Accordingly, translation processing is basically performed only by these two modules. That is, each module analyzes a sequence of words in an input language, analyzes a structure as a sentence, and expands a syntax analysis module in a conventional apparatus. It has a function as a plate.

In this case, the processing method of the present invention is different from the conventional one in that the two modules, that is, the English processing module and the Japanese processing module simultaneously perform processing while sending data to each other.

Therefore, in the process of translating English into Japanese, the English processing module analyzes the input English sentence, and the result (including the partial result) is immediately sent to the Japanese processing module. This Japanese processing module assembles Japanese sentences as they are based on the received information.

On the other hand, in the process of translating Japanese into English,
The Japanese processing module analyzes the input Japanese sentence, sends the result to the English processing module, and the English processing module assembles the English sentence.

The present invention employs object-oriented programming centering on a language to be processed, instead of focusing on procedures. Here, words, noun phrases, verb phrases, and the like to be processed are all treated as objects, and each object is independently executed in order to increase the efficiency of the entire apparatus. That is, they are executed in parallel processing.

An object corresponding to a word and an object corresponding to a phrase (hereinafter abbreviated as a phrase object including both objects) have two types of data, an attribute list and a connection list. The attribute list defines attributes of words and phrases, and the connection list specifies attributes of other phrases connected to the phrase.

Here, as the attribute (noun-Japanese),
For example, PERSON-represents people (eg, students, girls), QUANT-represents quantities (eg, many, most), N
UM-numerals (eg, three, five books), EXPL-form nouns (eg, Taro has returned), PROPER-proper nouns (eg, Japan, Taro).

As the attribute list, for example, "d
HEADER = “dogs”, POS = NP, P corresponding to a noun phrase composed of plural English nouns “ogs”
LUR information is included.

Further, as the connection list, CENTE
R = the phrase being processed itself, L1-the condition of the phrase connected from the left to the phrase being processed, and R1-the condition of the phrase connected from the right during the processing, as shown in FIG. 2, for example, CENTER: POS
= VP, INTRANS, L1: POS = NP.

This connection list indicates that when the phrase being processed is a verb phrase and there is a noun phrase on the left side, the two can be combined. In conventional devices, this type of information was described as a grammar rule.

As another example of the connection list, as shown in FIG. 3, CENTER: HEADER = "spit"
e ", L1: HEADER =" in ", R1: HEAD
ER = “of” information is included.

In this connection list, the heading of the phrase being processed is “spite”, “in” is on the left side, and “o” is on the right side.
This indicates that if there is a phrase headed by f ", it can be combined. In a conventional apparatus, this kind of information requires special processing together with a idiom.

Next, the operation will be described. First, when processing is started, the phrase object sends a message to an adjacent object, and based on the connection list,
Check if it can be combined.

When each item in the connection list is satisfied,
After merging, a new object is created. At this time, a message is sent to the adjacent object to notify that effect.

Next, the above-described processing is performed on the newly created phrase object. Also, if such a process is started from a column of phrase objects corresponding to the input word, and is continued as long as operable, eventually a phrase object including all of the phrase objects corresponding to the input word is generated. , The parse was successful and corresponds to the sentence that is the result of the phrase object.

As described above, the parsing is performed based on the information for each phrase, so that the conversion (translation) between the two languages is also performed for each phrase, and the converted phrase object is written in the same manner as the parsing. Can be assembled. At this time, after selecting a translated word corresponding to the word from the input sentence from the dictionary, it is checked whether or not the input sentence can be combined, and if possible, the combination is performed.

In the conversion process, the corresponding phrase object of the output language is specified from the value of the "TRAN" feature which is the translated information of the phrase object of the input language combined by the syntax analysis, and these are combined.

At this time, the value of the "TRAN" feature of the phrase object generated by the combination in the input language is used as a pointer to the phrase object generated by the combination in the output language.

This conversion process is performed simultaneously with the syntax analysis. Each time a word in the input sentence is combined with another word, the corresponding word (translation) between the two is extracted from the dictionary and the possibility of combination between the translations is checked. When the parsing ends, the conversion ends.

An output sentence, which is the final result, is created from the data structure generated by the conversion process. Basically, the reverse process of the morphological analysis is performed. The form of each word is determined from the basic form of the word object and the accompanying information included in the conversion result. After that, the word order and the like are determined according to the tree structure information and rules included in the conversion result, and a final output sentence is created.

In the above embodiment, the translation between Japanese and English has been described, but it goes without saying that the present invention can be widely used for translation between Japanese and other foreign languages such as German.

[0046]

As described above, according to the first aspect of the present invention, two sets of executing each processing of parsing an input language sentence, converting it into an output language, and generating and outputting an output language sentence according to object-oriented programming. A language processing means for translating one type of input language into another type of output language, and translating the other type of input language into the one type of output language. A module that performs translation processing between two languages according to language and grammar rules, for example, has two modules for English data and Japanese data.
In this case, the dictionary data can be easily added and corrected, and the translation between two languages can be easily executed by the dictionary and the single analysis program.

Since almost all grammatical information is described as dictionary data, the analysis program is simplified. Although a large amount of information is described in the dictionary, detailed description is possible because grammatical information can be described in word units.

Further, according to the second aspect of the present invention, since one type of language is English and the other type is Japanese, bidirectional translation between English and Japanese can be realized. At this time, since the difference between English and Japanese results in the difference between dictionary data, there is an advantage that both English and Japanese translations can be processed with a single analysis program.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an interactive machine translation apparatus according to an embodiment of the present invention.

FIG. 2 is an explanatory diagram showing a configuration of a connection list in the present invention.

FIG. 3 is an explanatory diagram showing a configuration of another connection list in the present invention.

FIG. 4 is a flowchart showing a flow of a translation process by the two-way machine translation device of the present invention.

FIG. 5 is a block diagram showing a conventional two-way machine translation apparatus as in FIG.

[Explanation of symbols]

 10 English analysis / synthesis unit as language processing means 11 Japanese analysis / synthesis unit as language processing means

────────────────────────────────────────────────── ───

[Procedure amendment]

[Submission date] November 14, 1997

[Procedure amendment 1]

[Document name to be amended] Statement

[Correction target item name] 0023

[Correction method] Change

[Correction contents]

The parsing of the English sentence and the parsing of the Japanese sentence are executed with reference to the English data 13 and the Japanese data 14 , respectively.

[Procedure amendment 2]

[Document name to be amended] Statement

[Correction target item name] 0041

[Correction method] Change

[Correction contents]

[0041] Then, in the conversion process, by parsing a translation word information phrase objects of the combined input language from the value of "TRAN" feature to identify phrase objects corresponding output language, to bond them.

[Procedure amendment 3]

[Document name to be amended] Drawing

[Correction target item name] Fig. 1

[Correction method] Change

[Correction contents]

FIG.

Claims (2)

[Claims] 1. An apparatus according to claim 1, further comprising two sets of language processing means for executing each processing of parsing an input language sentence, converting it into an output language, and generating and outputting an output language sentence in accordance with object-oriented programming. A bidirectional machine translation device, which translates the other type of output language into the one type of output language while translating the other type of output language. 2. The two-way machine translation device according to claim 1, wherein one type of language is English and the other type of language is Japanese.