JPH06325082A - Machine translation system - Google Patents

Abstract

PURPOSE:To simultaneously set plural pieces of translation environment and to set translation environment to the structure of plural documents. CONSTITUTION:This system is provided with an input means 11 for inputting an original sentence (a), a translation environment setting means 12 fetching translation environment specified by the original sentence (a) to set it to a translation environment buffer 13, a translation environment referring means 14 referring to translation environment set by the translation environment buffer 13, a translation means 15 executing translation by referring to word information in a translation dictionary 18 and grammar in a grammar memory 19 based on translation environment which is referred to and an output means 16 outputting a translated sentence (b). The translation environment setting means 12 fetches translation environment from the original sentence (a) to set it to the translation environment buffer 13. The translation means us executes translation by referring to word information in the translation dictionary 18 and grammar in the grammar memory 19 based on translation environment in the translation environment buffer 13. Consequently, it is not necessary to register plural kinds of grammar in the grammar memory 19.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a machine translation device capable of dynamically changing the setting of a translation environment by a user.

[0002]

2. Description of the Related Art Conventionally, in a machine translation device for translating a source language into a target language, first, a morpheme analyzing means analyzes a morpheme of an input sentence expressed in the source language (hereinafter referred to as "source sentence"), Word information about each morpheme (hereinafter,
It is called "word information". ) Is obtained from the translation dictionary. Next, the syntax analysis unit refers to the word information and the grammar registered in the grammar memory to analyze the structure of the original sentence, and the conversion generation unit converts it to the structure of the target language to output the output sentence expressed in the target language. (Hereinafter, referred to as “translation”) is generated and the translation is completed.

By the way, in the machine translation apparatus having the above construction, the translation accuracy and the translation processing speed are in a contradictory relationship, and if the translation accuracy is tried to be improved, the translation processing speed will decrease.
If the translation processing speed is increased, the translation accuracy will be reduced. Therefore, a plurality of types of grammars are registered in the grammar memory, a grammar to be used for translation processing is selected according to an operator's instruction, and the operator's intended translation purpose (hereinafter, referred to as "translation environment") is handled. The translation is performed at the translation accuracy and the translation processing speed (Japanese Patent Laid-Open No. 64-5).
No. 8060).

[0004]

However, in the conventional machine translation device, it is necessary to register a plurality of types of grammars corresponding to the translation environment in the grammar memory.
The capacity of the grammar memory becomes large. Also, since one grammar is required for one translation environment, it is not possible to set multiple translation environments at the same time. That is,
Arbitrary units for translation, such as words, phrases, sentences,
Paragraphs, documents, etc. (hereinafter referred to as "document structure")
It is not possible to combine one or more translation environments for each.

Not only cannot the translation environment be set for the structures of a plurality of documents, but also the translation environment cannot be set during the grammar processing by the syntax analysis means and the conversion generation means. Further, the operator has to select a necessary grammar when the setting of the translation environment is changed, which is troublesome.

The present invention solves the problems of the conventional machine translation apparatus described above, can reduce the capacity of the grammar memory, can set a plurality of translation environments at the same time, and can improve the structure of a plurality of documents. A machine translation device that can set a translation environment, can set a translation environment in the middle of grammar processing by a syntax analysis unit and a conversion generation unit, and can easily change the translation environment setting The purpose is to provide.

[0007]

Therefore, in the machine translation apparatus of the present invention, an input means for inputting an original sentence and a translation environment setting for extracting a translation environment specified in the original sentence and setting it in a translation environment buffer Means, a translation environment reference means for referring to the translation environment set in the translation environment buffer, and a translation means for translating by referring to the word information of the translation dictionary and the grammar of the grammar memory based on the referenced translation environment. , And output means for outputting the translated text.

In another embodiment of the present invention, the translation environment setting means extracts the translation environment together with its effective range from the original sentence and sets it in the translation environment buffer. In still another embodiment of the present invention, the input means has means for designating a translation environment in addition to the translation environment designated in the original text. In still another embodiment of the present invention, the translation environment setting means has means for setting a translation environment based on the translation result of the translation means.

[0009]

According to the present invention, as described above, the machine translation apparatus includes an input means for inputting an original sentence, and a translation environment setting means for taking out a translation environment specified in the original sentence and setting it in a translation environment buffer. A translation environment reference unit that refers to the translation environment set in the translation environment buffer, a translation unit that translates by referring to the word information of the translation dictionary and the grammar of the grammar memory based on the referred translation environment, and a translated sentence. It has an output means for outputting.

Therefore, when the original sentence is input, the translation environment setting means extracts the translation environment from the original sentence and sets it in the translation environment buffer. The translation means refers to the translation environment in the translation environment buffer via the translation environment reference means, refers to the word information in the translation dictionary and the grammar in the grammar memory, and translates based on the translation environment. Then, the translated text is output by the output means.

In another embodiment of the present invention, a translation environment and its effective range are specified in the original text. The translation environment setting means extracts the translation environment together with its effective range from the original sentence and sets it in the translation environment buffer. In this case, translation can be performed based on the corresponding translation environment in the effective range. In still another embodiment of the present invention, the input means has means for designating a translation environment in addition to the translation environment designated in the original text. In this case, the setting of the translation environment can be dynamically changed.

In still another embodiment of the present invention, the translation environment setting means has means for setting a translation environment based on the translation result of the translation means. In this case, the setting of the translation environment can be dynamically changed.

[0013]

Embodiments of the present invention will now be described in detail with reference to the drawings. FIG. 1 is a block diagram of a machine translation device showing an embodiment of the present invention. In the figure, a is an original sentence input from an arbitrary input medium, and b is a translated sentence output after being translated. The original sentence a is input by characters recorded on a recording paper, characters recorded on a disk, voice, or the like.

In the original sentence a, various translation environments are specified together with their effective ranges based on predetermined rules.
The translation environment is designated corresponding to the type of translation such as the purpose of translation and the translation method, and is used as information for controlling the machine translation device. It should be noted that the translation environment can be specified independently without the effective range. Further, the translation environment can be hierarchically designated as a nest structure.

When the input means 11 inputs the original sentence a recorded on an arbitrary input medium to the translation environment setting means 12, the translation environment setting means 12 determines the translation environment specified in the input original sentence a and its validity. The original sentence a in the range and the effective range is taken out, the translation environment and the effective range are set in the translation environment buffer 13, and the original sentence a in the effective range is translated into the translation means 1.
Transfer to 5 and request translation. Further, the translation environment setting means 12 transfers the entire translated text b transferred from the translation means 15 to the output means 16 when all the translations are completed.

The translation environment and its effective range are as follows:
It is not necessary to specify it in the above, and it is possible to input it separately by the input means 11. The translation environment and its effective range are set, for example, for each sentence, and the translation environment and its effective range specified in each sentence are set in the translation environment buffer 13 at the same timing.

The translation means 15 is a translation environment setting means 1
The translation environment set by 2 is requested to the translation environment reference means 14 and referred to. In this case, the translation environment buffer 13 may have a plurality of translation environments and their effective ranges specified in each sentence, and the translation environment reference means 14 simultaneously refers to each translation environment and its effective range. Will be lost. Therefore, the translation environment setting means 12 also sets the order of reference by the translation environment reference means 14 when setting each translation environment and its effective range. Also, the translation environment can be converted into a database and set in the translation environment buffer 13.

Subsequently, the translation means 15 translates by referring to the word information of the translation dictionary 18 and the grammar of the grammar memory 19 based on the referred translation environment. Then, the translated text b in the valid range is transferred to the translation environment setting means 12.
Normally, the translation environment setting means 12 takes out the translation environment specified in the original sentence a and sets the translation environment.
If necessary, the translation environment can be specified by itself based on the translation result of the translation means 15, and the translation environment setting means 12 can be requested to set the translation environment at any time.

For example, when it is found that the original sentence a is a technical document as a result of translation, the translation environment setting means 12 is requested to set the translation environment for the information in the translation environment buffer 13 as context information. be able to.
Furthermore, the translation environment can be specified by natural language based on the translation result. For example, information such as “No need to translate from here.” Can be set as a translation environment and no further translation can be performed.

When the translation of the effective range is completed in this way, the translation environment setting means 12 sets the translation environment set in the translation environment buffer 13 for the untranslated portion, its effective range and the original text of the effective range. a is transferred to the translation means 15 to request translation. When the translation means 15 translates all of the original sentence a by repeating this, the translation environment setting means 12 transfers the entire translated sentence b transferred from the translation means 15 to the output means 16. The output means 16
Outputs the translated text b.

Next, the operation of the translation environment setting means 12 in the machine translation apparatus having the above construction will be described. First, when the original sentence a is input to the translation environment setting unit 12 by the input unit 11, the translation environment setting unit 12 takes out the translation environment specified in the original sentence a and analyzes the effective range of each translation environment. Then, the translation environment and its effective range are set in the translation environment buffer 13 for each sentence, and the translation portion is extracted from the original sentence a in a predetermined order and transferred to the translation means 15 to request translation.

When the translation environment setting means 12 requests the translation, the translation means 15 requests the translation environment reference means 14 to refer to the necessary translation environment in the translation environment buffer 13 at any time. When each translation environment is hierarchically designated as a nest structure, the translation means 15 translates from the innermost effective range, but the translation environment corresponding to the translation location can be referred to.

At this time, the translation environment reference means 14 refers to the translation environment buffer 13 and transfers the referred translation environment to the translation means 15. Then, the translation environment reference means 1
When the translation environment is transferred from 4, the translation means 15 translates the translated text b in the effective range by referring to the word information of the translation dictionary 18 and the grammar of the grammar memory 19 based on the transferred translation environment. Transfer to the environment setting means 12.

Next, the translation environment setting means 12 receives the translated text b transferred from the translation means 15, and replaces the portion extracted from the original text a with the translated text b. At this time, the translation environment of the translation location of this time is deleted from the translation environment buffer 13. In this way, repeat the translation for each translation environment,
When the original sentence a is completely translated, the entire translated sentence b is transferred to the output means 16 and the translation process of the machine translation device is terminated.

FIG. 2 is a flow chart showing the operation of the translation environment setting means in the embodiment of the present invention. Step S1 The translation environment setting means 12 (FIG. 1) receives the original sentence a from the input means 11. Step S2 The translation environment specified in the original sentence a is taken out and the effective range of each translation environment is analyzed. Step S3: It is determined whether or not all the translations have been completed. When it is completed, the process proceeds to step S4, and when it is not completed, the process proceeds to step S5. In step S4, the entire translated text b is transferred to the output means 16. Step S5 The translation means 15 sets the translation environment corresponding to the translation part of the innermost effective range. Step S6 The translation environment reference means 14 takes out the translation part from the original sentence a and transfers it to the translation means 15. Step S7: The translation environment is set according to the request from the translation means 15. In step S8, the translated text b is received from the translation means 15 and the translated part extracted from the original text a is replaced with the translated text b. Step S9 Delete the translation environment of the translation location this time from the translation environment buffer 13.

FIG. 3 is a flow chart showing the operation of the translation environment reference means in the embodiment of the present invention. Step S11: Wait until there is a request for reference to the translation environment. In step S12, the data is transferred to the translation means 15 (FIG. 1) with reference to the translation environment. FIG. 4 is a flow chart showing the operation of the translation means in the embodiment of the present invention. Step S21 Wait until a translation request is made. In step S22, the translation environment reference means 14 is requested to refer to the translation environment, and the word information and the grammar are referred to and translated based on the translation environment. In step S23, the translation result is transferred to the translation environment setting means 12.

Next, the translation environment specified in the original sentence a and its effective range will be described. First, the effective range of the translation environment is specified according to the following rules. (A) A special symbol [] indicating a block can be described in the original sentence a, and the range enclosed by the special symbol [] is the effective range. Further, the block can have a nest structure. (B) The translation environment can be specified by enclosing it with a special symbol (). In addition, the translation environment is described in the special symbol (), a number that can be uniquely identified is added to the translation environment, the special symbol () is described at the rear of the original sentence a, and The same number can be written in the symbol (). In this case, the two special symbols () indicate the effective range of the translation environment.

If no number is added to the translation environment,
The effective range is within the innermost block to which the translation environment belongs. (C) When a plurality of contradictory translation environments are specified due to overlapping effective ranges of translation environments, the translation environment of the innermost block is given priority and is valid. The setting of which translation environment is prioritized can be changed as necessary.

Next, the designation of the translation environment will be described.
The translation environment is described as (specified environment name) by enclosing the specified translation environment with the special symbol () as described above. Then, when it is desired to translate a predetermined word or part of speech into a specific term, it can be described as (specified item = specified value). In this case, the translation environment buffer 13
The specified item and the specified value are set in correspondence with. Therefore, when the translating means 15 refers to the designated item in the form of substituting into the variable, the designated value of the translation environment buffer 13 can be obtained.

For example, the designated item can be a part of speech such as a definite article, and the designated value can be a term such as "that", "that" or the like. Note that one or more (specified items = specified values) must be described in advance by (specified environment name).
It is defined by combining.

After the machine translation device is activated, the combination of (designated item = designated value) can be changed to dynamically define and designate a new translation environment. Next, an example of the actual original sentence a will be shown. However, A to E indicate the input character strings that are the translations of the original sentence a. Example 1 When the translation environment in the original sentence a is specified as A [(untranslated) B [(translated) C] D] E, it is translated as follows.

That is, since the designation of the translation environment (translation) is valid in the range of the innermost block to which it belongs, it means that "input character string C will be translated." Further, since the translation designation of (non-translation) is effective in the range of the innermost block to which it belongs, it means that "the input character strings B, C and D are not translated".

Although the contradictory translation environments are specified for the input character string C, they are effective in the range of the innermost block. Therefore, it means that "the input character strings B and D are not translated, and the input character string C is translated." Example 2 In the original sentence a, the translation environment is A (1, no article) B (2, noun priority) C (1) D (2)
If E is specified, it is translated as follows.

That is, the translation environment of (no article) is
Since it is valid in the range surrounded by (1, no article) and (1), it means that "the article is not translated in the input character strings B and C." Further, since the translation environment of (noun priority) is effective in the range surrounded by (2, noun priority) and (2), "input character string C,
In D, the noun is given priority and the part of speech is estimated. Will be instructed. Therefore, it means that the article is not translated in the input character strings B and C, and the noun is given priority in estimating the part of speech in the input character strings C and D. In this case, a plurality of translation environments are specified for the input character string C, but they do not conflict with each other. Therefore, all translation environments are valid. Example 3 In the original sentence a, the translation environment is (title) A (no article) B (2, noun priority) C (1)
When specified as [(untranslated) D (2) E], it is translated as follows.

That is, since the translation environment of (title) is effective for all input character strings A to E, it is instructed to "translate the input character strings A to E under the context condition of title." Will be there. Further, since the translation environment of (no article) is effective in the range surrounded by (1, no article) and (1), “input character string B, C
In, do not translate the article. Will be instructed.

The translation environment of (noun priority) is
Since it is valid in the range surrounded by (2, noun priority) and (2), it means that "in the input character strings C and D, the noun is prioritized to estimate the part of speech."
Further, since the translation environment of (non-translated) is effective in the range of the innermost block to which it belongs, it is instructed that "the input character strings D and E are not translated". Therefore, “the input character strings A to E are translated under the context condition of title, the articles are not translated in the input character strings B and C, and the noun is given priority in estimating the part of speech in the input character strings C and D. The character strings D and E are not translated. " In this case, a plurality of contradictory translation environments are specified for the input character string D. Therefore, the translation environment (no translation) in the innermost block is valid. Example 4 In the original sentence a, the translation environment is specified as (title) A [(definite article = “that”) BC] DE, and the translation environment of (title) is (“goal” = noun phrase) (phrase) = None) (capitalized word = no translation) is defined in the translation environment buffer 13.

When the translation environment is designated as described above, translation is performed as follows. Ie (title)
The translation environment is developed as defined, and "in the input character strings A to E," goal "is a noun phrase in the context condition of title, no punctuation is added, and uppercase words are not translated. Will be instructed.

The translation environment of (definite article = “that”) is “the definite article is“ that ”in the input character strings B and C”. Will be instructed. It should be noted that the present invention is not limited to the above-described embodiments, and various modifications can be made based on the spirit of the present invention, and they are not excluded from the scope of the present invention.

[0039]

As described in detail above, according to the present invention, the machine translation apparatus has an input means for inputting an original sentence and a translation for extracting a translation environment specified in the original sentence and setting it in a translation environment buffer. An environment setting unit, a translation environment reference unit that refers to the translation environment set in the translation environment buffer, and a translation that translates by referring to the word information of the translation dictionary and the grammar of the grammar memory based on the referenced translation environment. And means for outputting the translated text.

When the original sentence is input by the input means, the translation environment setting means extracts the translation environment from the original sentence and sets it in the translation environment buffer. The translation means refers to the translation environment in the translation environment buffer via the translation environment reference means,
Referring to the word information of the translation dictionary and the grammar of the grammar memory,
Translate based on the translation environment. Then, the translated text is output by the output means.

Therefore, it is not necessary to register a plurality of types of grammars corresponding to the translation environment in the grammar memory, and the capacity of the grammar memory can be reduced. Further, not only a plurality of translation environments can be set at the same time, but also a translation environment can be set for a plurality of document structures. Moreover, since the setting of the translation environment is automatically changed, the work of the operator can be simplified.

In another embodiment of the invention, the translation environment and its scope are specified in the source text. The translation environment setting means extracts the translation environment together with its effective range from the original sentence and sets it in the translation environment buffer. In this case, translation can be performed based on the corresponding translation environment in the effective range. In still another embodiment of the present invention, the input means has means for designating a translation environment in addition to the translation environment designated in the original text. In this case, the setting of the translation environment can be dynamically changed, and the translation environment can be specified by voice, special characters, etc., so that it is easy to provide useful information for translation such as gestures, facial expressions, and intonation. Obtainable.

In still another embodiment of the present invention, the translation environment setting means has means for setting the translation environment based on the translation result of the translation means. In this case, the setting of the translation environment can be dynamically changed, and the translation environment can be set during the grammar process.

[Brief description of drawings]

FIG. 1 is a block diagram of a machine translation device showing an embodiment of the present invention.

FIG. 2 is a flowchart showing an operation of a translation environment setting means in the embodiment of the present invention.

FIG. 3 is a flowchart showing an operation of a translation environment reference means in the embodiment of the present invention.

FIG. 4 is a flowchart showing the operation of the translation means in the embodiment of the present invention.

[Explanation of symbols]

 11 input means 12 translation environment setting means 13 translation environment buffer 14 translation environment reference means 15 translation means 16 output means 18 translation dictionary 19 grammar memory a original sentence b translated sentence

Claims (4)

[Claims] 1. An input means for inputting an original sentence, (b) a translation environment setting means for extracting a translation environment specified in the original sentence and setting it in a translation environment buffer, and (c) the translation environment buffer. Translation environment referencing means for referencing the translation environment set in, and (d) translation means for translating by referring to the word information of the translation dictionary and the grammar of the grammar memory based on the referenced translation environment, (e) A machine translation device having an output means for outputting a translated sentence. 2. The machine translation device according to claim 1, wherein the translation environment setting means extracts the translation environment together with its effective range from the original sentence and sets it in the translation environment buffer. 3. The input means includes means for designating a translation environment in addition to the translation environment designated in the original sentence.
The machine translation device described in. 4. The machine translation device according to claim 1, wherein the translation environment setting means has means for setting a translation environment based on a translation result of the translation means.