KR100194798B1 - Integrated Method of Single Meaning Word in Multilingual Translation System Using Conceptual Structure - Google Patents

Abstract

The present invention relates to a single meaning word integrated processing for integrating a single meaning word in a multilingual translation system using a conceptual structure. The present invention relates to a single meaning that can significantly reduce the translation / generator by integrating a single meaning word independent of each language. In order to provide a word integration processing method, a first step of analyzing a concept tree made by interpreting the input language, to determine whether to pass only the concept, to pass the concept having a lower concept, or to pass the concept and expression together; A second step of referring to the generated grammar by the generator of each language when passing only a concept as a result of the determination of the first step, or passing a concept having a lower concept; And a third step of integrating the single meaning word integrated processing unit when the concept and expression are passed together as a result of the determination of the second step, thereby simplifying the number of translation / generators of the multilingual translation system to 1 / (N-1). It can be effected.

Description

Integrated Method of Single Meaning Word in Multilingual Translation System Using Conceptual Structure

1 is an exemplary configuration diagram of hardware to which the present invention is applied.

2 is a basic configuration example of a machine translation system to which the present invention is applied.

3 is an exemplary configuration of a translator / generator of a multilingual translation system covering five languages using a conventional method.

4 is an example of creating a generated grammar using a conventional method.

5 is an example of creating a production grammar according to the present invention.

6 is a diagram illustrating an embodiment of a multilingual translation system according to the present invention.

7 is a flowchart of one embodiment of a single word integrated processing method in accordance with the present invention.

* Explanation of symbols for main parts of the drawings

1: input device 2: CPU

3: output device 4: monoword multilingual dictionary

The present invention relates to a method for integrating a single meaning word in a multilingual translation system using a conceptual structure.

The multilingual translation system is a system for translating a language into several languages at the same time using a computer. For example, the system translates Korean into English, Japanese and German, and translates English into Korean, Japanese and German.

Machine translation systems generally translate one language into another. The input language is the source language, and the translated language is called the target language. The general machine translation method consists of an interpretation step of interpreting a source language, a conversion step of converting (translating) into a target language, and a generation step of generating a target language. And they translate based on linguistic knowledge. In other words, the parts of speech and sentences in each word are important. To this end, a sentence is translated through six steps: morpheme interpretation, syntax interpretation, semantic interpretation, transformation, syntax / meaning generation, and morpheme generation.

However, the machine translation (language translation) using the concept structure is processed at the same time by using the concept structure based on the meaning without translating at each stage centering on linguistic knowledge. This is a great way to handle conversational sentences with many non-legal elements.

Language translation using the concept structure has three generation rules.

The first is to pass only the concept.

For example, in the Korean-English translation, if the interpretation result is only the concept 'greeting' (greeting), the generator only needs to generate an English expression corresponding to the concept 'greeting' (greeting). In other words, you can create 'hello'.

The second is to pass concepts and expressions together.

In this case, the concept is the same in all languages, but the expression is different for each language. So you have to translate for each language. For example, when the concept of 'month name' and 'march' are crossed, the English side converts 'march' into English, that is, 'march'.

The third is to pass on concepts with subordinate concepts.

For example, the concept of 'suggest-metting' and the sub-concepts of 'local' and 'meeting' are passed. On the English side, the sub-concepts of 'local' and 'meeting' You can arrange them in the correct order to produce the proper representation. In other words, you only need to consider how to arrange subconcepts with other expressions to represent concepts. This way, you can create the expression Can I meet you at your office.

Looking at the three methods above, we can see that the first and third methods have nothing to do with other languages. In other words, it is relevant only to its own language and can exist independently. But the second method involves other languages. In other words, you need to know what the expression in the source language means for translation.

Therefore, the multilingual translation system using the conventional concept structure requires N (N-1) generated grammars and N interpreted grammars in the case of a multilingual translation system dealing with N languages, which makes the multilingual translation system very complicated. have.

The present invention devised to solve the above problems, since the first and the third method is independent of other languages, by separating and processing only the second case, it is possible to make independently for each language, so that a single translator / generator can be greatly reduced. The purpose is to provide a semantic word integration method.

In order to achieve the above object, the present invention, in a single meaning word integrated processing method applied to a multilingual translation system, by analyzing a concept tree made by interpreting the input language, to pass only the concept, the concept having a lower concept A first step of deciding whether to pass along or to pass together concepts and expressions; A second step of referring to the generated grammar by the generator of each language when passing only a concept as a result of the determination of the first step, or passing a concept having a lower concept; And a third step of integrating the single word integrated processor by passing the concept and expression together as a result of the determination of the second step.

Hereinafter, with reference to the accompanying drawings will be described an embodiment according to the present invention;

1 is an exemplary configuration diagram of hardware to which the present invention is applied. In the drawings, 1 denotes an input device, 2 denotes a central processing unit, 3 denotes an output device, and 4 denotes a multilingual dictionary required for single word integrated processing.

After inputting a language to be translated through the input device 1 and performing a language translation using a database storing a single word multilingual dictionary 4 in the central processing unit 2 in which the present invention is interrogated and output, Output through the device (3).

2 is an exemplary structure of a machine translation system using a conceptual structure. When a native language sentence is input, the native language is interpreted using a conceptual structure (21). The interpreted results are generated at the same time as being translated into the target language (22). The generated sentence is displayed immediately.

3 is an exemplary configuration of a translator / generator when a multilingual translation system covering five languages performs machine translation in a conventional manner.

In order to translate a language into another language in a multilingual translation system using the concept structure, one language must first be interpreted into a concept tree and then generated using the concept tree in another language. The generator responsible for generation must have a generation syntax for each language.

In FIG. 3, each arrow means a production grammar, and an arrow from language 1 to language 2 indicates a production grammar for translating language 1 to language 2, and an arrow from language 2 to language 1 indicates language 2 Means the generated grammar for translating into language 1. All other arrows have the same meaning.

For example, in the case of a system for translating three languages between the English and Korean languages, Japanese grammar for Korean-Japanese translation and Japanese grammar for English-Japanese translation are needed, and English for Korean-English translation is generated. Grammar, English-language grammar for Japanese-English translation, Korean-language grammar for English-Korean translation, and Korean-language grammar for Japanese-Korean translation are needed.

The reason for this is that, as usual, the second method of writing the three generation rules of language translation using the concept structure is processed separately without being separated. As shown in Fig. 4, the two generated grammars are taken as an example, and it is easy to see that similar parts are duplicated because the second method is not separated. In other words, it can be easily seen that the first and third methods of the three rules of generating grammar are duplicated in two Japanese generated grammars, and the second method can also be easily reduced by expressing the second method in another way.

In order to eliminate such duplication, a method of removing the second method and treating it at once is shown in FIG. That is, the Japanese grammar generated in FIG. 4 can be prepared by excluding duplicates as in FIG.

Based on the examples above, when generating the generated grammar using the existing method, N-1 generated grammar is required to generate one language in the system of translating N languages. When using processing, only one generation syntax is required for a language, and only a single single word integration processing part is added for the second method of generation syntax.

6 is an embodiment configuration diagram of a multilingual translation system according to the present invention, in which 41 is an interpreter for converting each language into an intermediate language expression having a conceptual structure, and 42 is a language expression generated from the intermediate language expression. I'm a generator. 43 is a part of the conceptual structure that integrates only the single meaning expressions that need translation into another language.

Of the three rules of writing grammar, the first and third, which are irrelevant to other languages and only related to one's own language, are in charge of the generator of each language, and the second method related to other languages is collected separately. Done.

Taking a multilingual translation system covering five languages as an example, the difference between the conventional method and the method using the present invention is illustrated in FIG. 3 and FIG. In FIG. 3, 20 translators / generators are used, and in FIG. 6, only 5 generators and 1 translator are possible. The proportion of one translator in the whole system is small compared to the interpreter or generator, so there is little impact on the system. Thus, for a translation system covering five languages, the number of translators / generators can be reduced by 75%.

7 is a flowchart of one embodiment of a single word integrated processing method according to the present invention.

First, the expression is retrieved from the monolingual multilingual dictionary 4 of FIG. 1 using the primitive monolingual words that should be translated by the second method of the production grammar writing rule. Then, the band of the target language is automatically generated using the retrieved expression. Pass the generated target language monoword to the generator.

Specifically, the generator that performs the construction, which is the final stage of translation using the generational syntax, sees each of the concepts in the conceptual structure tree as input, and calls the single meaning word integration processor or only refers to the generational syntax. Will be decided. When a call to an integrated processor is required, a representation of the concept is passed to the integrated processor. So you get a translation.

The single-word integrated processor automatically uses the input expression and source language information to automatically find and pass the target language expression from the table that has already been created.

For example, if there is a table created as shown in [Table 2] below, the native language is Korean, the expression is Korean expression 1, and the language to be translated is German, the band is on the same line. Find the German expression 1 and pass it on.

[Table 2] below is an exemplary configuration of the single meaning word multilingual dictionary used in the present invention.

The multilingual dictionary contains the table representation of the representative expressions and the expressions of each language corresponding to the representative expressions. Thus, while searching for the source language to be translated, it is possible to know the band word expression of the target language.

In the present invention as described above, since the second method of the three generation grammar creation rules is separated and processed separately, the number of translators / generators of the multilingual translation system is 1 / (N since only N interpreters, generators and one translator are required. -1) has the effect that can be simplified.

The present invention described above is not limited to the above-described embodiments and the accompanying drawings, and various substitutions, modifications, and changes can be made in the art without departing from the technical spirit of the present invention. It will be apparent to those of ordinary knowledge.

Claims (2)

Analyze the concept tree created by interpreting the input language in the single word integration processing method applied to the multilingual translation system, and decide whether to pass only the concept, the concept with the subordinate concept, or the concept and the expression together. The first step to do; A second step of referring to the generated grammar by the generator of each language when passing only a concept as a result of the determination of the first step, or passing a concept having a lower concept; And a third step of integrating the single meaning word integrated processing unit when the concept and expression are passed together as a result of the determination of the second step. The method of claim 1, wherein the single meaning word integration processing process comprises: a fourth step of receiving a source language single meaning word and searching for an expression in the single meaning word multilingual dictionary; And a fifth step of automatically generating a band of the target language using the retrieved expression and outputting the target language monoword to the generator.