JPH038082A - Machine translation processing system - Google Patents

Abstract

PURPOSE:To improve translation rate by providing the temporary connection relation of syntax even when the analysis of the relation is missed, continuing machine translation and translating the part of the temporary connection relation based on the order of words and postpositional particle information in an original sentence at the time of the translation in the Japanese / Korean machine translation. CONSTITUTION:A form element analysis processing part 4-1 executes the form element analysis of the original sentence and divides the original sentence into words and a syntax analysis processing part 4-2 executes syntax analysis concerning the divided words and synthesizes the divided words to a clause. Afterwards, the connection relation of these synthesized clauses is determined. When the connection relation can not be determined, the temporary connection relation is determined. Based on the connection relation determined in such a way, a conception structure generation processing part 4-3 generates conception structure and based on this conception structure, a generation processing part 4-4 generates a translated sentence. At such a time, concerning the part for which the temporary connection relation is determined, the translated sentence is generated based on the order of the words and the postpositional particle information in the original sentence.

Description

[Detailed Description of the Invention] [Summary] Regarding a machine translation processing method that performs machine translation between Japanese and Korean, it is possible to Even if a relationship fails to be analyzed, temporary relationships are maintained during syntax analysis, and l111i translation continues, and temporary relationships are created based on the word order and particle information of the original text during translation. The purpose of Uke is to translate related parts and improve the translation rate.The morphological analysis processing unit performs morphological analysis of the original text and divides it into words, and the morphological analysis processing unit analyzes the syntax of the words divided by this morphological analysis processing unit. After parsing and combining into clauses, the 4-ukes of these synthesized clauses determine the relationship, and if the 4-ukes cannot determine the relationship, the temporary 4-ukes determine the relationship, and this syntax The 4-uke determined by the analysis processing section is a conceptual structure generation processing section that generates a conceptual structure based on the relationship, and a translation is generated based on the conceptual structure generated by this conceptual structure generation processing section, and at this time, a temporary The fourth receiver includes a generation processing unit that generates a translated sentence based on the word order and particle information of the original sentence for the portion for which the relationship has been determined, and is configured to output the translated sentence generated by the generation processing unit.

[Industrial application field]

The present invention relates to a machine translation processing method for machine translation between Japanese and Korean.

[Problems to be solved by conventional techniques and inventions] Conventionally, j
! ! The i-machine translation system takes the original text as input, performs morphological analysis by referring to the dictionary for the original language, divides it into words, parses it by referring to the parsing grammar, synthesizes it into phrases, and then divides it into words by referring to the parsing grammar. was designed to analyze relationships and create a syntax tree. Here, there is a problem in that if the determination of the relationship between the four lines between clauses fails, the result of morphological analysis is displayed to indicate that machine translation has failed, and it is impossible to proceed further.

However, as in machine translation between Japanese and Korean, both use particles to express case, and even if syntactic analysis fails between two original languages that have similar word order, the characteristics between these two original languages will be lost. It is desired to perform machine translation using .

The present invention takes advantage of the fact that the word order and particle information between Japanese and Korean are similar. The aim is to increase the translation rate by allowing the machine translation to continue, and at the time of translation, based on the word order and particle information of the original text, the temporary 4-uke translates related parts.

[Means to solve problems]

Referring to Figure 1, Ll! Explain the means to solve the problem.

In FIG. 1, a morphological analysis processing section 4-1 performs morphological analysis of the original text and divides it into words.

The syntactic analysis processing unit 4-2 parses the words divided by the morphological analysis processing unit 4-1 and synthesizes them into phrases, and then determines the relationship between the four sentences of these synthesized phrases, and
When the Uke cannot determine the relationship, the tentative 4 Uke determines the relationship, for example, to the right neighbor.

The conceptual structure generation processing section 4-3 generates a conceptual structure based on the four relationships determined by the syntax analysis processing section 4-2.

The generation processing unit 4-4 generates a translated sentence based on the conceptual structure generated by the conceptual structure generation processing unit 4-3, and at this time, the temporary 4-uke is based on the word order and particles of the original sentence for the part for which the relationship has been determined. It generates a translated text based on information etc.

[Function] As shown in FIG.
1 performs morphological analysis of the original sentence and divides it into words, and the syntactic analysis processing unit 4-2 parses the divided words and combines them into clauses. And when the 4th receiver cannot determine the relationship, the temporary 4th receiver determines the relationship, for example, to the right neighbor, and the conceptual structure generation processing unit 4-3 generates a conceptual structure based on the determined 4th receiver. death,
The generation processing unit 4-4 generates a translated sentence based on these generated conceptual structures, and at this time, the temporary 4-uke generates a translated sentence based on the word order and particle information of the original sentence for the portion where the relationship has been determined. That's what I do.

Therefore, even if the 4-uke cannot determine the relationship during syntax analysis, the temporary 4-uke determines the relationship and continues machine translation, and when the translated sentence is generated, this temporary 4-uke determines the relationship. By generating a translated sentence based on the word order of the original sentence, particle information, etc., it is possible to improve the machine translation rate.

〔Example〕

Next, the configuration and operation of one embodiment of the present invention will be explained in detail using FIGS. 1 to 4.

In FIG. 1, an input section 1 is used to input an original text.

The storage unit 2 stores the original text to be translated, the results of translation processing,
It stores the obtained translation.

The editing control unit 3 controls editing of the original text and translated text.

The translation processing unit 4 is composed of 4-1 to 4-4, and is used to translate the original text using a machine II.

The morphological analysis processing unit 4-1 morphologically analyzes the input original text and divides it into words by referring to the original language dictionary 6 that stores information for dividing the original language into words (see Fig. 2). )
reference).

The syntactic analysis processing unit 4-2 refers to the analysis grammar 7 to parse the words into which the original text has been divided, and synthesizes them into phrases (
(See Figure 2 (C)), and furthermore, the 4-uke of these combined clauses determines the relationship (see Figure 2 (D)), and if the 4-uke cannot determine the relationship, the tentative 4-uke determines the relationship. For example, the right neighbor is determined (see FIG. 2 (e)).

The conceptual structure generation processing unit 4-3 analyzes the syntax and determines the 4
Uke generates a conceptual structure based on relationships (Second
(see figure).

The generation processing unit 4-4 generates a translated sentence based on the conceptual structure,
At this time, the temporary 4-uke generates a translated sentence based on the word order of the original sentence, particle information, etc. for the part for which the relationship has been determined (see Figure 3).

Next, a specific example will be explained using FIG. 2.

Figure 2 (a) shows an example of the original Japanese sentence to be translated into Korean: ``I bought a book and he bought a magazine.''

FIG. 2(b) shows an example of morphological analysis and division into words.

FIG. 2(C) shows an example of phrase synthesis.

Figure 2 (d) shows an example of the results of the clause synthesis in Figure 2 (c) in which the relationships between the four ukes of these clauses have been determined.
Here, the clause ``book wo °'' already had ``magazine wo'' in response to the predicate ``bought,'' so the 4th reception failed because it was not shared.

In Figure 2 (E), the 4th sentence in the clause in Figure 2 (D) is about the failure of the relationship ``book'', and the 4th sentence on the right side ``is'' determines the relationship as a temporary dummy, and the machine Don't let the translation stop.

Figure 2 (e) is about the relationship,
This is a generated conceptual structure. This conceptual structure is the second
Focusing on the predicate "buy" at the end of the diagram (E), from the end, "gti magazine", "he" buys the book" (dummy), and "I" are related as shown in the diagram. It is something that The text in the figure is a generation start mark.

When a Korean translation is generated based on the conceptual structure shown in FIG. 2(b) and with reference to the generative grammar 8 and the translation dictionary 9, it is generated as shown in FIG. 3(a). At this time, as shown in Figure 2 (
4) As a dummy, the 4th receiver is the “book” that tentatively determined the relationship.
Regarding the translated text, based on the fact that "hon wo" in the phrase synthesis in Figure 2 (c) is ■ in the word order and the particle information " is °, the translated text is as shown in ■ in Figure 3 (b). I am trying to generate .

As mentioned above, when the 4-uke of the original sentence fails when determining the relationship, the temporary 4-uke determines the relationship and continues machine translation, and when generating the translated sentence, it uses the word order of the original sentence, particle information, etc. By generating translations of related parts of this temporary four-way translation, it is possible to significantly improve the translation rate of machine translation.

FIG. 4 shows a specific example of translating the Japanese phrase "The investigator left for Tokyo, New York, and Seoul" into Korean.

Figure 4 (a) shows the original Japanese text.

Figure 4 (b) shows how the phrase synthesis/4-uke has determined the relationship.Here, the double line indicates that the 4-uke failed to determine the relationship, so the provisional 4-uke has determined the relationship (dummy). Show how the decision has been made.

FIG. 4(C) shows FIG. 4(B) as a conceptual structure.

When a Korean translation is generated by referring to the generative grammar 8 and the translation dictionary 9 based on this conceptual structure in Figure 4 (c), the third
It is generated as shown in Figure (b).

At this time, Figure 4 (c) As a dummy, the relationship is tentatively determined for the translation of “New York〈へ〉” and “Tokyo Kuhe〉゛”, and Figure 4 (B) “To New York” during clause synthesis.
Based on the fact that the word order of "To Tokyo" is ■, ■ and that the particle t17 report is "he," the translated sentences are generated as shown in ■ and ■ in FIG. 3 (b).

〔Effect of the invention〕

As explained above, according to the present invention, even if the relationship cannot be determined during parsing, the temporary relationship is determined and machine translation continues as it is, and when the translation is generated, This provisional dependency uses a structure that generates a translated sentence based on the word order and particle information of the original sentence for the part where the relationship has been determined, so even if syntactic analysis fails, the translation process continues, so the translated sentence is It is possible to reduce the number of situations in which no information is output at all, and to significantly increase the translation rate. Furthermore, if syntactic analysis fails, the translation is generated by determining the relationship based on the word order of the clauses used in the original sentence, particle information, etc., and therefore the quality of the translation can be improved.

[Brief explanation of drawings]

Fig. 1 is a diagram of the principle configuration of the present invention, Fig. 2 and 7. FIG. 4 is a detailed explanatory diagram of the present invention, and FIG. 3 shows an example of translated text related to the present invention. In the figure, 4 is a translation processing unit, 4-1 is a morphological analysis processing unit, 4
-2 is a syntactic analysis processing unit, 4-3 is a conceptual structure generation processing unit,
4-4 is a generation processing unit, 6 is a dictionary for the original language, 7 is an analysis grammar,
8 represents a generative grammar, and 9 represents a translation dictionary.

Claims (1)

[Claims] A machine translation processing method for performing machine translation between Japanese and Korean includes: a morphological analysis processing unit (4-1) that performs morphological analysis of an original sentence and divides it into words; The processing unit (4-1) parses the divided words and synthesizes them into clauses, and then determines the dependency relationships of these synthesized clauses, and if the dependency relationship cannot be determined, creates a temporary dependency relationship. A syntactic analysis processing unit (4-2) that determines the syntactic analysis, and a conceptual structure generation processing unit (4-2) that generates a conceptual structure based on the dependency relationship determined by the syntactic analysis processing unit (4-2).
4-3) and the conceptual structure generated by this conceptual structure generation processing unit (4-3), and at this time, the word order of the original sentence and the particle A machine translation processing method characterized by comprising a generation processing unit (4-4) that generates a translated sentence based on information, and configured to output the translated sentence generated by the generation processing unit (4-4). .