JPS6180359A - Translation system - Google Patents

Abstract

PURPOSE:To select the style of a translating sentence on the basis of user's specification by selectively specifying a sentence style and translating an input sentence so as to coincide with the specified style. CONSTITUTION:A translator T is provided with a CPU 1 for execution translation processing, a main memory 2 for storing a program for the processing of the CPU 1, a display device 3 for displaying contents necessary for the translation, a keyboard 4 for inputting contents necessary for the translation processing, and a translation module 5 for storing information necessary for the translation process. At the translation of an input sentence to be translated, the style of the translating sentence is selected through the keyboard 4 and the input sentence is translated in accordance with the specified style.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of the Invention The present invention relates to a translation method for a system in which a machine and a human cooperate to produce correct translation results or answers to questions.

BACKGROUND ART When translating an input sentence to be translated, a prior art translation system uses a predetermined, fixed writing style. For example, a sentence such as [Th1s is a pen, J] is uniquely translated into ``This is a pen.'' The writing style of the sentence was determined. The reason for this is that the writing style used in the translation system is based on an English-Japanese dictionary.

Problems to be Solved by the Invention In the prior art translation systems as described above, the style of the translated text is uniquely determined, so when used in commercial texts or letters, the expressions are too formal and inappropriate. However, there was a problem in that the translated text had to be corrected afterwards before it could be used in practice, resulting in a long working time.

SUMMARY OF THE INVENTION Therefore, an object of the present invention is to provide a translation method that allows the user to select the writing style of a translated sentence according to a user's designation, thereby shortening the working time.

Means for Solving the Problems The present invention is characterized in that a writing style is selectively specified when translating an input sentence to be translated, and the input sentence is translated in the specified writing style. This is a translation method that

Effects The present invention allows the user to select the style of the translated text when translating the input text that is to be translated, and uses the translated text as it is, which reduces the work time. .

Embodiment FIG. 1 is a block diagram showing the configuration of a translation device T that can implement the present invention. The translation device T is connected to a central processing unit (CPU) 1 that performs processing such as translation processing, and includes a main memory 2 in which programs and the like for processing by the CPU 1 are stored, and a main memory 2 for displaying information necessary for translation processing. A display device (CRT) 3, a keyboard 4 for inputting information necessary for the translation process, and a translation module 5 in which information necessary for the translation process is stored are provided.

FIG. 2 shows an embodiment of the translation module 5.

Translation monoyl 5 has four buffers r A y B
y C+D and a table 6 having a dictionary for translation, grammar rules, and production rules. These buffers A
-D stores data at each analysis level in the translation process, as will be described later.

Here, general machine translation processing will be explained.

Machine translation generally has levels as shown in Figure 3.

When the source M edict is input from the upper left, the dictionary lookup of ■, ■
Morphological analysis of , syntactic analysis and parsing of ■ proceed. Machine translation can be broadly divided into two types depending on the level of analysis. One is to analyze concepts that do not depend on each language. This concept is called an intermediate language. This method of generating a target language is called the pivot method. The target M word is a translated sentence. The other method is called the transfer one-way method, in which the internal structure of the source M word is obtained by analyzing any one of the following: morphological analysis of ■, syntactic analysis of ■, semantic analysis of ■, and context analysis of ■. After converting to the same level structure of the target language, a tarded language is generated. Here, the contents of each analysis are listed in Table 1.

(Leaving space below) Table 1 In this embodiment, it is assumed that translation is performed up to the level of syntactic analysis. In the present invention, as shown in FIG. 4, a translation module is considered which consists of three parts: a dictionary lookup morphological analysis section GA, a syntactic analysis section GB, and a tarded language generation section GC.

FIG. 5 shows the contents of each buffer A to D in FIG.
is a pen.

This example shows how to translate an English sentence into Japanese. First, the original text is read into buffer A in FIG. 2 as shown in FIG. 5(1). Dictionary lookup morphological analysis unit GA in Figure 4
The necessary information for each word is extracted and placed in buffer B in FIG. Part of speech information, which is a part of it, is as shown in FIG. 5 (2).

This is a multipart speech word, but this is a multipart word.
It is uniquely determined by the syntax analysis unit GB in the figure, and the modification relationship is entered into the buffer C in FIG. 2 as shown in FIG. 5 (3). According to this and the grammar rules of Table 6 in Figure 112, the second
The analysis shown in the table is performed.

(Left space below) Table 2 This rule expresses, for example, that [a sentence is made up of a subject and a predicate].

In the target M word generation unit GC shown in Figure 4, appropriate particles and auxiliary verbs are added, and the word is converted into Japanese form, as shown in Figure 5 (4).
The data is stored in buffer D in FIG. 2 as shown in FIG.

The method of the present invention will be described below using FIGS. 6 to 9. The sentence used at this time is Tb1s is
a pen.

Consider translating the input sentence. In this embodiment, for example, when a break in a sentence ends with a style such as ``It is ~ deru.'', it is called the first style. I will call the time the second writing style.Figures 6 to 9
In the translation described in the figure, the tJS2 writing style will be specified.

FIG. 6 is a flowchart showing the entire process performed by CP tJ 1. In step n1, the writing style of the translated text is specified. For example, when the writing style is set to the first writing style, the flag is not set. When the writing style of the translated text is set to the second writing style, a flag is set in step n1.

That is, when no special designation is made, the translation device T
It means to translate in style. step n2
Then, it is determined whether there is an input sentence as shown in Figure (1) @7 that is to be translated. Terminate processing when there is no input sentence.

When it is determined in step n2 that there is an input sentence, control moves to step n3. In step n3, the initial morphological analysis section GA shown in FIG. 4 is performed, and the parts of speech and translated words are extracted as shown in FIG. 7(2). In step n4, the syntax analysis unit GB shown in FIG. 4 performs the processing, and the label shown in FIG. 7(1) is determined. In step n5, a translated sentence is generated using the writing style specified in step n1. Steps 03 to n5 are subroutines, and each process is performed in a separate routine. When generation is completed in step n5, control returns to step n1.

Step n of specifying writing style as shown in Figure 6
1 is within the loop of the 70-chart in FIG. 6, so the writing style of the translated sentence can be specified for each sentence. When specifying a writing style for each collection of input sentences that make up a sentence, step n1 can be executed outside the loop of the flowchart shown in Figure 6, or step n
Once you have performed the process in step 1, you can simply set the flag so that the state of the flag is maintained without having to specify the flag again. At this time, when the generation performed in step n5 is completed, control returns to step n2.

FIG. 8 is a flowchart showing the generation operation in step n5 shown in FIG. At step n6, the beginning of the generation table shown in FIG. 7(3) is input to pointer P. In step n7, it is determined whether the label indicated by pointer P is in the original text shown in FIG. 7(1). When it is determined in step n7 that the label indicated by pointer P does not exist in the original text, control moves to step n8, where pointer P is incremented and control is returned to step lower.

When it is determined in step n7 that the label indicated by pointer P is in the original text, control is transferred to step 9. In step n9, it is determined whether the label specified by the pointer P is a predicate and the writing style is the second writing style. When the label is a predicate and the writing style designation is the second writing style, control is transferred to step nlo. In step n1(), the predicate is converted from the first writing style to the second writing style, and control is transferred to step n12.

In step n9, if it is determined that the label specified by pointer P is not a predicate or the writing style is specified to display the first writing style instead of the second writing style, step nl
Control is transferred to 1. In step nil, the translated words are stored in the result buffer 7 shown in FIG. 7(4) in locations 1 to 4. In step n12, it is determined whether the label specified by the pointer P is a predicate. When it is determined in step n12 that the label indicates a predicate, control moves to step n13. At step n13, aspect processing is performed. Modal processing is the process of changing the ending of a sentence when it is a negative sentence, such as ``It is not...'', or changing the ending of a predicate sentence, such as ``It will be...'' when the sentence is a future car. Say what you want.

When the process of step n13 is completed, control moves to step n14.

At step n12, it is determined that the label specified by pointer P is not a predicate, and at step n14
control is transferred to In step n14, FIG. 7 (3)
The particle corresponding to the label indicated by the pointer P shown in is placed in the result buffer 7 shown in FIG. 7 (4).
Let ~ store in 4. In step n15, it is determined whether all sentences have been generated. When it is determined in step n15 that the generation has not been completed yet, the control moves to step n8, and after incrementing the pointer P, the control returns to step n7. When it is determined in step n15 that the generation has ended, the process shown in the 70-chart of generation shown in FIG. 8 ends.

The flow of chart 70 in FIG. 8 will be specifically explained below assuming that the input sentence to be translated is shown in FIG. 7 (1). At step n6, the head of the generated table is input to pointer P. The head of the generation table is the subject, as shown in FIG. 7(3). At step n7, it is determined whether the label indicated by pointer P is in the original text of FIG. 7(1). At this time, since the subject is included in the original sentence, control moves to step n9. In step n9, since the label is the subject, control moves to step nil. In step n, 11, Fig. 7(2)
The translated word "kore" corresponding to the subject t l'+ i 5-11= stored in the dictionary buffer shown in is stored at 1 in the storage area of the result buffer. At step n12, it is determined whether the label is a predicate or not, but since the subject is not a predicate at this time, control is transferred to step n14. In step n14, the particle "wa" included in the generation table shown in FIG. 7(3) is stored in the result buffer at 2. At step n15, it is determined that the generation has not yet been completed, and at step n8, the pointer P is incremented and points to the object word, which is the next label.

At step n7, since the original sentence does not contain the object, pointer P is incremented at step n8. At this time, pointer P indicates the complement. Since the original sentence contains a complement in step n7, control moves to step n9. In step n9, since the label is a complement, control is transferred to step nll, and "one pen" corresponding to the translated word in the dictionary buffer is stored at 3 in the store area of the result buffer. step n12
In this case, it is determined that the label is not a predicate, and step n14
control is transferred to

In step n14, since there is no particle corresponding to the complement in the generation table, nothing is stored in the result buffer in this step. At step n15, it is determined that generation has not yet been completed, and at step n8, pointer P is incremented. At this time, pointer P indicates statement g. At step n7, since the predicate indicated by pointer P is included in the original text, control moves to step n9.

As for the writing style, as mentioned above, the second writing style is specified, and the label pointed to by pointer P is a predicate, so the control is as follows.
Move to step nlo. In step nlo, "desu." is selected from the translated words in the dictionary buffer, and the translated word is stored in 4 in the result buffer. In step n12, since the label is a predicate, the process moves to step n13, where modal processing is performed. Since there is no particle in step n14, the process moves to step n15. At step n15, it is determined that the generation has been completed, and the process is terminated.

FIG. 9 is a 70-chart for explaining the conversion process of step nIO shown in FIG. step n16
It is determined whether the part of speech of the predicate is a verb. If it is determined in step n16 that the part of speech is a verb, control is transferred to step n17. In step n17, the conjugated form of the verb is changed to a conjunctive form, and "masu" is added to the translated word.

In step n18, the translated words converted in step n17 are stored in the result buffer.

When it is determined in step n16 that the part of speech is not a verb, control moves to step n19. step n
In step 19, it is determined whether the part of speech is an adjective. When the part of speech is determined to be an adjective, step n20
control is transferred to In step n20, "desu" is added to the translated word, and in step n18, the translated word is stored in the result buffer.

When it is determined in step n19 that the part of speech is not an adjective, control moves to step n21. In step n21, it is determined whether the part of speech is an adjective verb. When it is determined in step n21 that the part of speech is an adjective verb, control moves to step n22.

In step n22, "desu" is added to the stem of the predicate, and in step nlB, the result is stored in the result buffer.

When it is determined in step n21 that the part of speech is not an adjective verb, control moves to step n23.

In step n23, it is determined whether the translated word is ".". At step 23, the translated word is ``.

”, control moves to step n24. In step n24, the translated word is converted to "desu.", and in step n18, the result is stored in the result buffer.

If it is determined in step n23 that the translated word is not ``naru.'', it is determined that the ``light'' processing has been performed in error, and the processing is terminated.

As mentioned above, in this embodiment, normally the sentence breaks are [
It is. ] Although translation is done in the first style, such as ``desu.'', it can also be translated in the second style, such as ``desu.''.

In the above-mentioned example, the sentence break is "is." or "is."
is. Although the translation was made in a writing style such as "," it may be changed to another sentence 15, such as a literary style or a colloquial style.

Effects As described above, according to the present invention, the translated text can be translated (since the style of the translated text is selected when translating the input text, the translated text can be used as it is). This can be achieved by shortening the work time.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the configuration of a translation device T that can implement the present invention, FIG. 2 is a block diagram showing an embodiment of the translation module 5, and FIG. 3 is for explaining machine translation processing. , FIG. 4 is a diagram showing the specific N13 of the translation module, FIG. 5 is a diagram showing the contents to be transferred to each buffer A to D, and FIG. 6 is a diagram showing the entire process performed by the CPUI. 7 is a diagram showing the contents of the dictionary buffer, etc., FIG. 8 is a 70-chart showing the generation operation at step n5 shown in FIG. 6, and FIG.
7 for explaining the process of conversion of the illustrated step nlO.
It is 0-char-1. 5... Translation module, 6... Table, A-D.
・・Buffer, T・・Translation device

Claims (1)

[Claims] A translation method characterized in that a writing style is selectively specified when translating an input sentence to be translated, and the input sentence is translated into the specified writing style.