JPH06203068A - Machine translation system - Google Patents

Abstract

PURPOSE:To provide a machine translation system capable of performing than heretofore. translation more naturally CONSTITUTION:This system is provided with modules 11, 12 which generate a structural analysis tree by analyzing a text of source language, memory 18 which stores a rule to convert the structural analysis tree to that of target language, a module 13 which converts the structural analysis tree of the source language to that of the target language by referring to the memory 18, a module 21 to detect a word which satisfies a condition decided in advance in the structural analysis tree of the target language, and a module 22 which prepares the text of the target language by eliminating a detected word based on the structural analysis tree of the target language.

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,
In particular, the present invention relates to a machine translation device that can prevent unnatural translations from being generated due to differences in language structure.

[0002]

2. Description of the Related Art In a conventional machine translation device, a source language text input from a keyboard is transferred to a main CPU.
Input to the translation module under the control of the central processing unit. The input source language text is converted into a target language text by using a dictionary including a main dictionary and a user dictionary stored in a memory provided in the machine translation device, a grammar rule, and a tree structure conversion rule. Will be translated.

[0003]

However, in such a conventional machine translation device, since there is a one-to-one correspondence between a word in the source language and a word in the target language, a generated target is generated. There was a problem that the language text became unnatural. FIG. 19 is an example showing such a problem. The example shown in FIG. 19 is an example in which an English sentence is translated into Japanese.

Referring to FIG. 19, in English, "you" appears in two places and "your" appears in one place. In the translated Japanese sentence, the word "you" appears three times in total.

Words such as pronouns such as "you" in English sentences are indispensable in the structure of English. However, this is not always the case in languages other than English. For example, in Japanese, it is common to omit the subject if the subject is clear among the speakers. Therefore, in a language such as Japanese, if an English pronoun or the like is directly replaced by a translated word, the translated sentence may be redundant and unnatural. Therefore, in consideration of such a problem peculiar to each language, it is desired to perform more flexible translation so as not to make a one-to-one relationship between the original sentence and the translated sentence. . If possible, such translation should be easy.

Therefore, an object of the invention described in claim 1 is to provide a machine translation device capable of more flexible translation than ever.

An object of the invention described in claim 2 is to provide a machine translation device capable of easily performing a more flexible translation than the conventional one.

An object of the invention described in claim 3 is to provide a machine translation device with few redundant expressions and capable of easily performing a more natural translation.

An object of the invention described in claim 4 is to provide a machine translation device capable of easily translating English into a Japanese with a natural expression.

The object of the invention described in claim 5 is to prevent words, such as pronouns, which are expressed in English but are unnecessary in Japanese, from appearing in Japanese text, and the natural expression from English can be obtained. It is an object of the present invention to provide a machine translation device capable of easily translating Japanese into Japanese.

The object of the invention as set forth in claim 6 is to prevent words, such as pronouns, judged by the user that they are expressed in English but need not be expressed in Japanese, from appearing in Japanese text. It is an object of the present invention to provide a machine translation device that can easily translate English into a natural expression in Japanese.

[0012]

A machine translation device according to claim 1 analyzes an input source language text,
First means for generating a source language structural analysis tree and a conversion rule for converting the generated structural analysis tree into a predetermined target language structural analysis tree
Storage means, a means for converting a source language structure analysis tree into a target language structure analysis tree by referring to conversion rules, and detecting a word satisfying a predetermined condition in the target language structure analysis tree. And a generating unit for generating a target language text by excluding the detected words based on the target language structure analysis tree.

A machine translation device according to a second aspect is the machine translation device according to the first aspect, wherein the detecting means includes a second storing means for storing an exclusion rule defining a predetermined condition. Determining means for determining whether any of the exclusion rules stored in the second storage means can be applied to each word in the structure analysis tree of the target language,
Means for preparing a predetermined marker in relation to a word for which any of the exclusion rules is determined to be applicable by the judging means, and the generating means excludes the word for which the marker is prepared and excludes the target language. Produces the text of.

A machine translation apparatus according to a third aspect is the machine translation apparatus according to the second aspect, wherein the exclusion rules are a plurality of semantically overlapping words included in the structural analysis tree of the target language. It includes an exclusion rule that specifies at least one of the words as an exclusion target.

A machine translation device according to a fourth aspect is the device according to the second aspect, wherein the source language is English and the target language is Japanese.

A machine translation device according to a fifth aspect is the device according to the fourth aspect, wherein the exclusion rule specifies a specific pronoun included in the structural analysis tree of the target language as an exclusion target. Includes exclusion rules.

The machine translation device according to claim 6 is the device according to claim 5, further comprising means for inputting a specific pronoun.

[0018]

In the machine translation device according to the first aspect, the source language structure analysis tree is generated from the input source language text, and the generated structure analysis tree is stored in the first storage means. It is converted into a structure analysis tree of the target language according to the conversion rule. When a word satisfying a predetermined condition is detected in the target language structure analysis tree, the target language text is generated by excluding the detected word.

In the machine translation device according to the second aspect, a source language structure analysis tree is generated from the input source language text, and the generated structure analysis tree is stored in the first storage means. It is converted into a structure analysis tree of the target language according to the conversion rule. It is determined whether or not any of the exclusion rules stored in the second storage means is applicable to each word in the converted structure analysis tree. If there is an applicable exclusion rule, a predetermined indicator is prepared in relation to the word, and the word in which the indicator is prepared is excluded to generate the text in the target language.

In the machine translation device according to the third aspect, the source language structure analysis tree is generated from the input source language text, and the generated structure analysis tree is stored in the first storage means. It is converted into a structure analysis tree of the target language according to the conversion rule. It is determined whether or not any of the exclusion rules stored in the second storage means is applicable to each word in the converted structure analysis tree. The exclusion rule includes an exclusion rule that specifies at least one of a plurality of semantically overlapping words included in the structure analysis tree of the target language as an exclusion target. If there is an applicable exclusion rule, a certain sign will be prepared in relation to that word,
The target language text is generated excluding the words for which the indicator was prepared.

In the machine translation device according to the fourth aspect, an English structural analysis tree is generated from the input English text, and the generated structural analysis tree is the conversion rule stored in the first storage means. Is converted into a Japanese structural analysis tree. It is determined whether any of the exclusion rules stored in the second storage means is applicable to each word in the structure analysis tree of the conversion word. If there is an applicable exclusion rule, a predetermined marker is prepared in association with the word, and the word for which the marker is prepared is excluded to generate Japanese text.

In the machine translation device according to the fifth aspect, an English structural analysis tree is generated from the input English text, and the generated structural analysis tree is the conversion rule stored in the first storage means. Is converted into a Japanese structural analysis tree. It is determined whether or not any of the exclusion rules stored in the second storage means is applicable to each word in the converted structure analysis tree. When there is an applicable exclusion rule, for example, when a specific pronoun is found, a certain sign is prepared in relation to that specific pronoun and the specific pronoun is excluded to generate Japanese text. To be done.

In the machine translation device according to the sixth aspect, an English structural analysis tree is generated from the input English text, and the generated structural analysis tree is the conversion rule stored in the first storage means. Is converted into a Japanese structural analysis tree. It is determined whether or not any of the exclusion rules stored in the second storage means is applicable to each word in the converted structure analysis tree. If there is an applicable exclusion rule, for example, if a specific pronoun previously entered by the user is found, a predetermined sign is prepared in association with that specific pronoun to exclude that specific pronoun. Japanese text is generated.

[0024]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A machine translation device according to an embodiment of the present invention will be described below.

Before explaining the embodiments, the concept of machine translation will be briefly described. Referring to FIG. 1, the analysis processing performed in machine translation has various analysis levels. The machine translation is a process for inputting the source language shown in the upper left of FIG. 1 to sequentially perform the processing of each level to finally obtain the target language shown on the right side of FIG. For example, when a source language is input, first, a level L1 dictionary lookup process, a level L2 morphological analysis process, a level L3 syntactic analysis process, and so on are performed, and finally a level L10 morpheme generation process is performed. The target language is generated.

Machine translation is roughly divided into the following two types, depending on which level of analysis processing is performed. First
Is a pivot method in which an intermediate language shown at level L6 is analyzed and a target language is generated therefrom.
Secondly, the internal structure of the source language is obtained by performing analysis to any of the above-mentioned levels L2 to L5, and then converted to the internal structure of the target language at the same level as the obtained internal structure of the source language. Later, it is a transfer method that generates a target language.

The intermediate language used in the pivot system is a concept that does not depend on either the source language or the target language. Therefore, if the source language is analyzed once to obtain an intermediate language, a plurality of languages can be generated from this intermediate language, which is advantageous for multilingual translation. However, in such a pivot method, it is still unsolved whether the intermediate language which is the basic concept can be really set.

The transfer method is a compromise solution to such a problem of the pivot method, and many current systems adopt this transfer method. The following description is for the transfer method, and the machine translation device of the embodiment described later also uses the transfer method.

The contents of each analysis process shown in FIG. 1 will be described below. (1) Dictionary lookup, morpheme analysis In this process, the process of dividing the input sentence into morpheme strings (word strings) using the longest matching method or the like while referring to the dictionary in which the morphemes are stored First done. Then, a process of obtaining grammatical information such as a part of speech and a translated word for each of the obtained words and further analyzing tense, personal name, and number are performed.

(2) Syntactic Analysis In this process, a process of assembling and deciding a sentence structure (structural analysis tree) such as inter-word crossing based on the part-of-speech / variant, etc. is performed. In this process,
No judgment is made as to whether the structure of the obtained sentence represents the correct meaning.

(3) Semantic Analysis From a plurality of structure analysis trees obtained as a result of the syntactic analysis processing, semantically correct and improper ones are discriminated and only the correct ones are adopted.

(4) Context Analysis In the context analysis process, a process of understanding the topic of the input sentence and removing abbreviated parts and ambiguities contained in the input sentence is performed.

In the translation module of the machine translation apparatus according to the embodiment of the present invention described below, it is assumed that analysis processing up to level L3 is performed. That is, the functional block of the translation module of the machine translation device of this embodiment has a structure as shown in FIG. This translation module
Referring to FIG. 2, dictionary lookup / morpheme analysis module 11 for performing dictionary lookup / morpheme analysis processing on the source language text provided from input module 10.
A syntactic analysis module 12 for performing syntactic analysis on the morphologically analyzed text; a syntactic conversion module 13 for converting the syntactic analysis result to generate a structural analysis tree of the target language; and a syntactic conversion module. And a translation sentence generation module 14 for generating a translation sentence of the target language based on the structure analysis tree of the target language generated by 13. The contents stored in the memory 6 are referred to in the processes performed by the respective modules 11 to 14, and the contents of the respective processes will be described in detail in the embodiments.

The translated sentence generation module 14 applies the translated word conversion rule to the structure analysis tree of the target language generated by the syntax conversion module 13, and sets a predetermined value in the omission flag for the omissible word. The translation word conversion rule application module 21 and the abbreviation flag set by the translation word conversion rule application module 21 are referred to, and the translation text of the target language is omitted while omitting the optional words based on the structure analysis tree of the target language. And a translation generation module 22 for generating. The translated sentence in the target language generated by the translated word generation module 22 is output to a printer, an external file, or the like via the output module 17.

The memory 6 includes a dictionary memory 18, a buffer memory 19 and a translation conversion rule memory 20. The memory 6 also includes a memory (not shown) for storing grammar rules and tree structure conversion rules.

FIG. 3 is a block diagram of a machine translation apparatus which is an embodiment of the language processing apparatus of the present invention. With reference to FIG. 3, this machine translation device is connected to a main CPU (central processing unit) 1, a bus 7 to which the main CPU 1 is connected, a main memory 2 connected to the bus 7, and a bus 7. A display device 3 including a CRT (cathode ray tube) or LCD (liquid crystal display device), a keyboard 4, the above-mentioned translation module 5 connected to a bus 7, and a translation dictionary / grammar connected to the translation module 5. And the above-mentioned memory 6 storing a knowledge base such as rules and tree structure conversion rules.

When the source language sentence is input, the translation module 5 translates the sentence in a predetermined procedure and outputs the target language. Referring to FIG. 4, as a hardware, translation module 5 is connected to bus 7 and translates a source language (English in the case of this embodiment) input via bus 7 according to a predetermined translation program. Then, the translation CPU 15 for outputting to the bus 7 as the target language (Japanese in this embodiment) and the bus 7
Connected to the translation CPU 15, a translation program memory 16 for storing a translation program executed by the translation CPU 15, a buffer A for storing an input source language original sentence for each word, and a buffer A for storing each word. Per word,
Part of speech of each word obtained by referring to the dictionary included in the memory 6,
A buffer B for storing information such as a translated word, a buffer C for storing information about a source language structure analysis tree, and a structure of a target language obtained by converting the source language structure analysis tree stored in the buffer C. A buffer D for storing a parse tree, and a buffer for storing a sentence prepared as a Japanese form by supplementing the Japanese structural parse tree stored in the buffer D with appropriate particles and auxiliary verbs. Including E and.

The operation of the English-Japanese translation by the machine translation apparatus of this embodiment will be described below with reference to FIGS. In addition,
The details of the translation program executed by the translation CPU 15 will be described later.

First, the read original sentence is decomposed into morphemes by morphological analysis, and as shown in FIG.
(See FIG. 4).

Subsequently, a translation C based on the translation program 16
Under the control of the PU 15, the dictionary lookup / morpheme analysis unit 11 shown in FIG. 2 refers to the dictionary stored in the memory 6 for each word of the original sentence stored in the buffer A, and translates each word. And information such as part-of-speech information can be obtained. This information is stored in buffer B shown in FIG. Part of this information includes the part-of-speech information of each word, and the part-of-speech information is stored as shown in FIG. That is, "this" is a multi-part-of-speech word and has two parts of speech, a pronoun and a referential adjective. The part of speech of "is" is a verb. Similarly, the parts of speech for "a" and "pen" are also stored in the buffer B. "Thi
Although s "is a multi-part-of-speech word, what is the part-of-speech in the sentence is uniquely determined by the process corresponding to the parsing unit 12 shown in FIG.

In the process corresponding to the syntactic analysis module 12 shown in FIG. 2 of the translation program, a structural analysis tree showing the inter-word relations between words according to the dictionary and grammatical rules stored in the memory 6 is shown in FIG. 7 is determined. The result of this parsing is stored in the buffer C in FIG.

The structure analysis tree is determined as follows. Among the grammatical rules stored in the memory 6, the grammatical rules for English are made up of the following.

Sentence → main part, predicate main part → noun phrase predicate → verb, noun phrase noun phrase → pronoun noun phrase → article, noun For example, the first rule of these rules is that "a sentence is a main part It is made up of predicates. " The same applies to other rules. The structural analysis tree is determined according to these rules. It should be noted that such grammatical rules are similarly prepared for Japanese, and correspondence is made between English grammatical rules and Japanese grammatical rules.

In the process corresponding to the syntax conversion module 13 shown in FIG. 2 of the translation program, the structure of the input structural analysis tree (see FIG. 7) of the English sentence is the syntax for the Japanese sentence shown in FIG. Converted to a parse tree structure. In this conversion, the “tree structure conversion rule” stored in the memory 6 is used in the same manner as that used by the syntax analysis unit 12 described above. This conversion corresponds to the conversion from the level L3 to the target language level L9 in FIG. The obtained result is stored in the buffer D shown in FIG. The example sentence "Thi used in this description
This means that "is is a pen" has been converted into the Japanese character string "This is a pen" by this conversion.

The portion of the translation program that performs the processing corresponding to the translated sentence generation module 14 of FIG. 2 is obtained by adding an appropriate particle "ha" or auxiliary verb to the obtained Japanese character string "this is a pen". , And stores it in the buffer E shown in FIG. 4 in the grammatical Japanese form as shown in FIG. This process corresponds to the level L9 shown in FIG.
Corresponds to the conversion from level L10. The obtained Japanese sentence “This is a pen.” Is output from the translation module 5 shown in FIG. 4, stored in the main memory 2, and displayed on the display device 3.

A feature of the present invention is that the syntax conversion module 13
When a translated sentence is generated from the Japanese structural analysis tree generated by, a word that is frequently used in English but does not need to be translated in Japanese, for example, a word that is the subject of a subordinate clause and is the same as the subject of the main clause. It is a place to generate a translated sentence so that it does not translate something like. The details of the processing performed by the translated sentence generation module 14 will be described below.

FIG. 10 shows the translated sentence generation module 1 of FIG.
4 is a flowchart of the process performed in 4. In particular, step S001 corresponds to the process performed by the translation conversion rule application module 21 of FIG.

Referring to FIG. 10, first, step S001.
Then, the buffer D that stores the conversion result of the syntax analysis tree obtained as a result of the processing of the syntax conversion module 13 (see FIG. 2) is scanned, and a processing of setting an omission flag for the omissible word is performed. This process is performed by checking whether or not any of the translation word conversion rules stored in the translation word conversion rule memory 20 of FIG. 2 can be applied to each word of the input text. The process performed for a certain word in each translation conversion rule is to confirm whether or not a certain condition applies to this word. In this embodiment, a case where three rules are used as the translation conversion rules will be described.

The first translation word conversion rule is that an abbreviation flag is set when the subject of the subordinate clause is the same as the subject of the main clause.
That is. The second translation conversion rule is to "set an abbreviation flag for the same pronoun as the subject". The third translation conversion rule is that "a flag is set for pronouns designated by the user not to be translated." Regarding the third rule, the word set by the user is stored in the translated word conversion rule memory 20 by using the keyboard 4 shown in FIG. The process of checking whether or not these three rules are applicable is sequentially performed in the rule application process of step S001. The processing for each rule will be described below with reference to the flowchart.

FIG. 11 is a flowchart of the processing performed for the first translation conversion rule. Referring to FIG. 11, first, in step S101, the pointer is positioned at the beginning of the parse tree conversion result buffer D.

Then, in step S102, it is determined whether or not the pointer reaches the end of the conversion result. If it has arrived, the process for rule 1 ends. If the word at the position pointed to by the pointer is not the subject, the step S1
Judgment at 03 is made. If it is not the subject, the process proceeds to step S108, and 1 is added to the pointer. After step S108, the control returns to step S102.

If the word pointed by the pointer is the subject, control proceeds to step S104. In step S104, it is determined whether or not the section to which the word at the position currently pointed to belongs is the main section. If it is the main clause, the processing proceeds to step S105, and if it is the dependent clause, the processing proceeds to step S106. That is, if the word at the position pointed to by the pointer belongs to the subordinate clause, control is performed in step S.
It will proceed to 106.

In step S105, the word at the position pointed to by the pointer is determined to be the subject of the main clause. Therefore, the word at the position pointed to by the pointer is stored in a buffer (reserved in the buffer memory 19 in FIG. 2) for remembering the subject. After step S105, the control proceeds to step S108.

When the control proceeds to step S106, it is determined whether or not the word at the position pointed by the pointer is the same as the word (subject) stored in the buffer at step S105. If they are the same, the control proceeds to step S107, and if they are not the same, the control proceeds to step S10.
Go to 8 respectively.

In step S107, since the subject of the main clause and the subject of the subordinate clause are determined to be the same word, it is determined that rule 1 can be applied. Therefore, the omission flag indicating that the first translation conversion rule is applicable is set. The post-processing of step S107 proceeds to step S108.

By thus completing the processing shown in FIG. 11, the first word of each word in the buffer is
It is determined whether or not the translation conversion rule of is applicable, and if such a rule is present, an abbreviation flag is set for it.

Using the first translation conversion rule, the following result is obtained. FIG. 15 shows a result of parsing the English sentence shown in FIG. 19 and converting it into a Japanese tree structure. The conversion result is stored in the buffer D as shown in FIG. Applying the above-mentioned first translation conversion rule to this buffer D, “y” in (* 1)
The omission flag is set for ou ". In order to make the above determination, information indicating which part of the tree structure the main part is stored in the buffer D is required. However, such information is also stored together in this buffer D. In the case of Fig. 15, the information that the third "You" is the subject of the main clause is held, and from the right. Since the fourth "you" matches this, the omission flag is set for this "you" in step S107 of FIG.

FIG. 12 is a flowchart of a process for checking whether or not the second translation conversion rule can be applied.
First, in step S201, the pointer is positioned at the head position of the conversion result buffer D.

Then, in step S202, the pointer is
A determination is made as to whether the last position of the conversion result has been reached. If the pointer reaches the end of the conversion result, the control proceeds to step S203, and if it does not reach the end, the control proceeds to step S209.

In step S209, it is determined whether or not the word at the position currently pointed to by the pointer is the subject. This judgment is made based on the grammatical information in the tree conversion result buffer as shown in FIG. If the current word is the subject, control proceeds to step S210, and if it is not the subject, control proceeds to step S211.

In step S211, 1 is added to the pointer. The control after step S211 is step S20.
Return to 2.

In step S210, the word at the position currently pointed to by the pointer is stored in a buffer prepared in advance for storing the subject. This buffer is secured in the buffer memory 19 of FIG. After step S210, control proceeds to step S211.

By repeating the processing of steps S202 and S209 to S211, the word serving as the subject is stored in the buffer, and when the pointer reaches the end of the buffer, control proceeds to step S203.

In step S203, the pointer is returned to the head position of the tree conversion result buffer D again.

Then, in step S204, it is determined whether or not the pointer points to the last position of the conversion result. If the end has been reached, this process ends. If the end has not been reached, control proceeds to step S205.

In step S205, it is determined whether the word at the position currently pointed to by the pointer is a pronoun. If it is not a pronoun, the control is step S20.
If it is a pronoun, the control proceeds to step S206.

In step S208, 1 is added to the pointer. After step S208, the control is step S204.
Return to.

In step S206, it is determined whether the word at the position currently pointed to by the pointer is the same as the word stored in the buffer, that is, the same as the subject. If they are the same, the control proceeds to step S207, and if not, the control proceeds to step S20.
Go to 8.

In step S207, since the second translation conversion rule can be applied to the word currently at the position pointed by the pointer, the omission flag is set. After step S207, the control is step S208.
Proceed to.

By repeating the processing of steps S204 to S208, the omission flag based on the result of applying the second translation conversion rule is set for the pronoun that is the same as the subject of the sentence.

In the example shown in FIG. 15, a flag is set for the word "your" with (* 2). In this case, the possessive "your" is treated as the same as "you".

When the second translation conversion rule is applied, the following result is obtained in the example sentence shown in FIG. Referring to FIG. 15, the fourth word “you” from the left.
As a result of setting the abbreviation flag in r ”, the part corresponding to this word is deleted in the translated Japanese sentence. That is, the example sentence shown in FIG. 19 is converted into Japanese according to the second translation conversion rule. The translation results in what is shown in Fig. 17. In Fig. 17, the first translation conversion rule is also applied, and the result shown in Fig. 17 is more natural than that shown in Fig. 16. It is close to Japanese.

FIG. 13 shows a flowchart of the processing for checking whether or not there is a word to which the third translation conversion rule can be applied. Referring to FIG. 13, first, in step S301, a process of setting the pointer to the head position of tree conversion result buffer D is performed.

Subsequently, in step S302, it is determined whether or not the position pointed by the pointer has reached the end position of the contents stored in the tree conversion result buffer D. If it has, this process ends. If not, the control proceeds to step S303.

In step S303, it is determined whether the word at the position pointed to by the pointer is the word designated by the user in advance. If they do not match, the control proceeds to step S305, and if they match, the control proceeds to step S304.

In step S305, the pointer is incremented by 1, and control is returned to step S302.

In step S304, since the word to which the third translation conversion rule is applicable is found in the tree conversion result, an omission flag indicating that this word can be omitted is set. Processing is performed.
After step S304, the control proceeds to step S305.

In order to apply the third translation conversion rule, the user needs to specify a specific pronoun in advance. Regarding this, for example, it is conceivable that the translation system may make an inquiry to the user at the stage of initial setting before starting the translation processing so that the user can set a "pronoun not translated".

For example, it is assumed that the user has designated "you" as a pronoun that does not translate. All variants of "you" are treated as being designated as "untranslated pronouns" by the designation of "you". Then, as shown in FIG.
In the case of the contents of the tree conversion result buffer shown in, (*
1) The omission flag is set for the three words (* 2) (* 3). That is, as a result of applying the third translation word conversion rule, a translation sentence that does not include the word "you" is obtained as shown in FIG. This translated sentence is closer to natural Japanese than that shown in FIGS. 16 and 17.

Returning to FIG. 10 again, the rule application process of step S001 ends by performing all the processes for each rule as described above.

After step S001, step S002
Then, a determination is made as to whether or not there is a word for which Japanese has not yet been generated among the stored contents of the tree conversion result buffer D. If the word has reached the end and there are no more target words to be generated, this processing ends. If there is a word for which Japanese has not been generated, steps S003 and S00 are performed according to the part of speech of the word.
4, the processing proceeds to S005 and the like, and Japanese is generated according to the pronoun, the noun, the verb, and the like. After this, for example, the post-processing of step S003 proceeds to step S006.

In step S006, step S001
As a result of the rule application processing performed in step 1, it is determined whether or not the omission flag is set for the word to be processed. That is, it is determined whether or not the word to be processed is translated into the translated sentence. If the omission flag has not been set, control proceeds to step S007. If the omission flag is set, control returns to step S002.

In step S007, a translated word stored in the dictionary buffer is assigned to the word to be processed and stored in the output sentence buffer. After step S007, the control returns to step S002.

When the omission flag is set for a certain word, the process of step S007 is not performed. Therefore, this word will not be translated into the translated text.

Although the present invention has been described by taking the machine translation device for English-Japanese translation as an example, the present invention is not limited to the machine translation device for English-Japanese translation. INDUSTRIAL APPLICABILITY The present invention can be applied to general translation between languages in which an unnatural translated sentence may be obtained when words are mechanically corresponded to each other due to different language structures.

[0086]

As described above, in the machine translation device according to the first aspect, when a word satisfying a predetermined condition is detected in the structure analysis tree of the target language, the detected word is excluded. The target language text is generated.
Even words included in the text of the source language may not be translated into the translated text of the target language, and thus it is possible to perform more flexible translation than before.

According to another aspect of the machine translation apparatus of the present invention, among the words in the converted structure analysis tree, any of the exclusion rules stored in the second storage means is associated with the applicable word. A predetermined indicator is prepared and the word is excluded to produce the target language text. A plurality of words to be excluded from the translated text can be defined in advance in the form of an exclusion rule, and a more flexible translation than before can be easily performed.

In the machine translation device according to the third aspect, among the words in the structural analysis tree after conversion, at least one of a plurality of words that are semantically overlapping is designated as an exclusion target. Then, since the target language text is generated by excluding the words designated as the exclusion target, a more natural translation with less redundant expressions can be easily performed.

In the machine translation device according to the fourth aspect, an English structural analysis tree is generated from the input English text and further converted into a Japanese structural analysis tree. If each of the words in the converted structure analysis tree to which the exclusion rule is applicable is excluded, the Japanese text is generated. It is possible to prevent words that are expressed in English but that are unnecessary in Japanese from appearing in Japanese text, and it is possible to easily translate English into a natural expression in Japanese.

In the machine translation device according to the fifth aspect, an English structural analysis tree is generated from the input English text and further converted into a Japanese structural analysis tree. Among the words in the converted structure analysis tree, those for which the exclusion rule is applicable, for example, specific pronouns, are excluded to generate Japanese text. It is possible to prevent words such as pronouns that are expressed in English but are unnecessary in Japanese from appearing in Japanese text, and it is possible to easily translate English into a natural expression in Japanese.

In the machine translation device according to the sixth aspect, an English structural analysis tree is generated from the input English text and further converted into a Japanese structural analysis tree. Among the words in the converted structure analysis tree, those for which the exclusion rule is applicable, for example, specific pronouns input by the user in advance, are excluded to generate Japanese text. It is possible to prevent words, such as pronouns, that the user has judged to be unnecessary in Japanese from being expressed in English, from appearing in Japanese text, making it easy to translate English into a natural expression in Japanese. Can be done

[Brief description of drawings]

FIG. 1 is a diagram schematically showing the concept of machine translation.

FIG. 2 is a block diagram of a machine translation device which is an example of a language processing device of the present invention.

FIG. 3 is a detailed functional block diagram of a translation module 5 shown in FIG.

FIG. 4 is a hardware block diagram of a translation module.

5 is a diagram schematically showing the stored contents of a buffer A. FIG.

FIG. 6 is a diagram schematically showing stored contents of a buffer B.

FIG. 7 is a diagram schematically showing stored contents of a buffer C.

FIG. 8 is a diagram schematically showing stored contents of a buffer D.

FIG. 9 is a diagram schematically showing stored contents of a buffer E.

FIG. 10 is a schematic flowchart of a translated sentence generation process.

FIG. 11 is a schematic flowchart of an application process of a first translation conversion rule.

FIG. 12 is a schematic flowchart of application processing of a second translation conversion rule.

FIG. 13 is a schematic flowchart of application processing of a third translation conversion rule.

FIG. 14 is a diagram schematically showing translated word conversion rules.

FIG. 15 is a schematic diagram showing stored contents of a tree conversion result buffer D after applying a translation conversion rule.

FIG. 16 is a diagram schematically showing a translated sentence as a result of applying the first translated word conversion rule.

FIG. 17 is a diagram schematically showing a translated sentence as a result of applying first and second translated word conversion rules.

FIG. 18 is a schematic diagram showing a translated sentence as a result of applying a third translated word conversion rule.

FIG. 19 is a schematic diagram showing a translation result by a conventional machine translation device.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 main CPU 2 main memory 3 CRT 4 keyboard 5 translation module 6 memory 11 dictionary lookup / morphological analysis module 12 syntactic analysis module 13 syntactic conversion module 14 translation sentence generation module 15 translation CPU 16 translation program memory

Claims (6)

[Claims] 1. A means for analyzing an input source language text to generate a structure analysis tree of the source language, and a means for converting the structure analysis tree into a structure analysis tree of a predetermined target language. A first storage unit for storing a conversion rule; a unit for converting the structure analysis tree of the source language into a structure analysis tree of the target language by referring to the conversion rule; and a structure of the target language In the parse tree, detection means for detecting a word satisfying a predetermined condition, based on the structure analysis tree of the target language, for generating the text of the target language by excluding the detected word A machine translation device including a generation unit. 2. The detecting means includes a second storing means for storing an exclusion rule defining the predetermined condition, and each word in the structure analysis tree of the target language,
Determination means for determining whether any of the exclusion rules stored in the second storage means is applicable; and a word for which any of the exclusion rules is determined to be applicable by the determination means. Associated with a means for preparing a predetermined marker, the generating means generating the text in the target language by excluding words for which the marker has been prepared. apparatus. 3. The exclusion rule includes an exclusion rule that specifies, as an exclusion target, at least one of a plurality of semantically overlapping words included in the structure analysis tree of the target language. The machine translation device described. 4. The machine translation device according to claim 2, wherein the source language is English and the target language is Japanese. 5. The machine translation device according to claim 4, wherein the exclusion rule includes an exclusion rule for designating a specific pronoun included in the structure analysis tree of the target language as an exclusion target. 6. The machine translation device according to claim 5, further comprising means for inputting the specific pronoun.