JPH09185626A - Machine translation device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To perform phrase translation to let the effect of auxiliary verb appear for each clause consisting of parallel clauses including the auxiliary verb. SOLUTION: A target language generating part 2d of a translation module 2A is provided with a parallel clause detecting means, auxiliary verb detecting means and auxiliary verb translation generating means. A main memory 3 is provided with an auxiliary verb buffer 3k for storing the auxiliary verb code of auxiliary verb detected for each clause by the auxiliary verb detecting means. Then, concerning a source sentence, in which the parallel clauses including the auxiliary verb exist, inputted to an input part 1, the target language generating part 2d generates a target language for which the ending of each translated clause consisting of the parallel clauses is deformed into ending corresponding to the auxiliary verb code. As a result, the effect of this auxiliary verb is let appear in both the parallel clauses and unnaturality is canceled.

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, and more specifically, it divides a source language sentence into phrases and clauses, and translates each phrase or clause into target language phrases or clauses. The present invention relates to a machine translation device capable of outputting.

[0002]

2. Description of the Related Art In many existing machine translation apparatuses, translation processing is executed in units of one sentence. Therefore, the longer the length of one sentence is, the more likely it is that the syntax will be erroneously parsed, and if a long sentence is translated on a sentence-by-sentence basis, the possibility that a translated sentence that does not make sense will be increased. Also,
Even if the syntax can be parsed correctly, a translated sentence that is logically correct but extremely difficult to read is often output. Therefore, in the current machine translation apparatus, a main usage method is that a user post-edits the output result and provides it for use.

Further, it is better to obtain a translation result in a unit smaller than one sentence in a state where there are few parsing errors, than to obtain a translation result in a sentence unit in a state where there is a high possibility that the syntactic analysis is wrong. As a good point, a machine translation device has been proposed which outputs a translation result in units smaller than one sentence such as a word, a phrase or a clause without outputting the translation result in a sentence unit. Examples of such a machine translation device include the following.

That is, dictionary construction, morphological analysis, and translation word selection are performed in the same manner as in a normal sentence-by-sentence translation apparatus to obtain the syntactic structure of the original sentence and the translated word for each word of the original sentence. After that, when generating a translated sentence, the original sentence is divided into phrases and clauses based on the syntactic structure, and the words in the original language (hereinafter referred to as the original word) that compose the divided original sentence clauses and clauses are supported. The translated words are grouped into each phrase or section of the original sentence, and the translated words are arranged in an appropriate order in each phrase or section (only the translated words belonging to the phrase or section are extracted from the order of the translated words when translated in one sentence unit). The translated phrase is assembled into the translated phrase.

Here, the "phrase" of the translated phrase does not mean "phrase" in the English grammar, but is a word used to control the "phrase" and "section". The same applies to the word "phrase" used below.

[0006]

However, the conventional machine translation device that outputs a translation result in a unit smaller than one sentence has the problems as shown in Examples 1, 2 and 3. [Example 1] Original text "It would push them to the end of the credito
rs' lineand make it unlikely. "(1) It would push them to the end of the credito
rs' line (2) and (3) make it unlikely. The result of phrase translation (a) It puts them at the end of the creditor's line (b) and (c) so that it's unlikely Let's go

In this example, when the original sentence and the translation result are compared in phrase units, the auxiliary verb "would" exists in the original sentence in the original sentence phrase (1), but this is reflected in the translated sentence phrase in (a). Not. The effect of the auxiliary verb “would” appears only in the translated phrase of (c) in the form “to be probably”. The reason for this is that the result of normal translation processing of the original sentence, one sentence at a time, is that "it puts them at the end of the creditor's line and makes it unlikely. This is because the effect of the auxiliary verb appears only at the end of the sentence and that the phrase division position of the original sentence is faithfully reflected in the translated sentence.

However, comparing the original phrase of (1) and the translated phrase of (a), the auxiliary verb “would” exists in the original phrase, but the effect does not appear in the translated phrase. Is unnatural. In terms of meaning, the auxiliary verb "would" of the original sentence phrase is effective for both two clauses of the original sentence that share the subject and are in parallel. In some cases, it is only necessary to change the translation form at the end of the sentence, but in the case of phrase translation where each clause is separated, it is necessary to explicitly change the translation form for each clause. It cannot be said to be a translation.

[Example 2] Normally, when a translated sentence of a target language is generated from an original sentence of a source language by a machine translation device, in addition to a translated word of a source language word, it does not correspond to any word in the source language. Although some translated words are generated and used, the generation position of the translated word that does not correspond to the word in the source language is determined by a very simple rule. For example, when the target language is Japanese, a translated word that does not correspond to a word in the target language is generated in the same phrase as the previous translated word, based on the translated word generation order when translated in sentence units. I'm supposed to.

When the translation process from English to Japanese is performed according to such rules, the following results are obtained. Original text `` keating and other Lincoln exectives are the
subject ofa class-action lawsuit alleging they mis
led investors about the safety of junk bonds issued
by American Continental. ”The result of dividing the original sentence into phrases (1) keating and other Lincoln exectives are the
subject ofa class-action lawsuit (2) alleging they misled investors about the saf
ety of junk bondsissued by American Continental Results of phrase translation (a) Keating and other Lincoln executives are the subject of the class action claims (b) They invested in the security of Jack Bonds issued by American Continental I misled my house

In this case, it is difficult to understand that "insist" in the translated phrase of (a) takes an object and that the object is the entire translated phrase of (b). To improve this, put it at the end of the translated phrase in (b)
The case particle "to" which received the whole translated phrase of (b),
It will be easier to understand if it is generated immediately before "arguing" the translated phrase of (a). However, as described above, since the case particle "to" does not correspond to the word of the original sentence phrase, the translated word "misleading" generated immediately before "to" when translated in sentence units. Since it is designed to be generated in the same phrase as, the translated phrase in (b) is generated.

[Example 3] The original sentence "The plan was opposed by Keating and some of
his former topexecutives, who invested in the comp
any's securities. "Result of splitting into original phrases (1) The plan was opposed by Keating and some of
his former topexecutives (2) the result of who invested in the company's securities (a) the plan was opposed by Keating and some of his former top executives (b) invested in the company's securities

In this example, the original phrase in (2) is a relative clause, but if only the translated phrase in (a) is viewed,
(In this case, the nominative case) is a missing section, so
It is difficult to take meaning as a phrase by this alone.

Therefore, an object of the present invention is to eliminate the above-mentioned problems that hinder the comparison of the original sentence and the translation result in phrase units or to read and proceed only the translated sentence phrase. It is to provide a machine translation device that makes it easier to understand the contents.

[0015]

In order to achieve the above object, the invention according to claim 1 provides a syntactic analysis unit for performing syntactic analysis of a sentence in a source language input from an input unit, and an obtained syntactic structure. A syntax conversion unit that converts the syntactic structure in the target language, a target language generation unit that generates a target language translation string of the input sentence based on the conversion result, and a division unit that divides the input sentence into phrases or clauses according to the division rules. A division position determining unit for determining a division position, and a translated sentence for obtaining a translation string in the target language for each phrase or clause based on the division result by the division position determining unit and the target language generated by the target language generating unit In a machine translation device that has a phrase construction unit and translates an input sentence into phrases or clauses, the target language generation unit is
When there is a parallel clause that shares only a specific case in the input sentence, the auxiliary verb is detected from each of the clauses that make up the parallel clause, and the last word is detected for each of the clauses that make up the parallel clause. Generate a translated word string in the target language converted to the final form according to the auxiliary verb, and reflect the aspect based on the auxiliary verb in any of the clauses that form a parallel clause including the auxiliary verb that shares only the above specific case. The feature is that each translation is performed.

According to the above configuration, when an input sentence having a parallel clause including an auxiliary verb sharing only a specific case is translated for each clause forming the parallel clause, the parallel clause is formed. A translated word string is generated by reflecting the aspect based on the auxiliary verb in each of the clauses.

According to a second aspect of the present invention, in the machine translation device of the first aspect of the invention, the target language generation unit is specified based on the syntactic structure in the target language obtained by the syntactic conversion unit. Parallel clause detecting means for detecting parallel clauses sharing only the case, auxiliary verb detecting means for detecting auxiliary verbs from the respective clauses detected by the parallel clause detecting means, and information representing the auxiliary verb detected by the auxiliary verb detecting means. Auxiliary verb information storage means for storing
With respect to each clause detected by the parallel clause detecting means, an auxiliary verb translation generating means for generating a translated word string in the target language in which the last word is converted to an end form according to the information stored in the auxiliary verb information storage means It is characterized by having.

In the above configuration, when the parallel clause detecting means detects a parallel clause sharing only a specific case in the syntactic structure of the target language, the auxiliary verb detecting means detects the auxiliary verb from each clause forming the parallel clause. Information that is detected and represents this auxiliary verb is stored in the auxiliary verb information storage means. Then, when generating the translated word string of each of the clauses forming the parallel clause, the translated word string in which the last word is converted to the final form according to the information stored in the auxiliary verb information storage means by the auxiliary verb translation generation means Is generated. Thus,
When an input sentence includes a parallel clause including an auxiliary verb that shares only a specific case, each clause forming the parallel clause is translated by reflecting the aspect based on the auxiliary verb.

Further, the invention according to claim 3 is a syntax analysis unit for performing syntax analysis of a sentence in a source language input from an input unit, and a syntax conversion unit for converting the obtained syntax structure into a syntax structure in a target language. A target language generation unit that generates a translation string in the target language of the input sentence based on this conversion result; a division position determination unit that determines the division position when dividing the input sentence into phrases or clauses according to the division rules; Based on the result of the division by the division position determination unit and the target language generated by the target language generation unit, a target phrase construction unit that obtains a target word string in the target language for each phrase or clause In a machine translation device that translates phrases or clauses, the translated phrase construction unit receives the words related to the dependency relationship as a result of the division by the division position determination unit. When the word of is divided into different clauses, a supplementary symbol indicating that the clause containing the related word originally exists immediately before the corresponding word in the translated word sequence of the clause containing the above received word is added. It is characterized by generating.

According to the above configuration, when the division position determining unit divides the related word and the receiving word in the input sentence into different clauses, the related word is included in the translated word string of the clause including the receiving word. A translated word string that clearly indicates that the clause including the word is originally present is generated.

Further, the invention according to claim 4 is the machine translation device according to claim 3, wherein the translated phrase construction section inputs based on the translated string in the target language obtained for each clause. A void in which a dependent word and a dependent word in a sentence are divided into different clauses and a clause including the dependent word has a structure in which the clause including the dependent word is completely omitted from the clause including the dependent word. When the hollow clause is detected by the clause detecting means and the hollow clause detecting means, the clause including the related word is omitted immediately before the word of the receiving word in the translated word sequence of the clause including the receiving word. A supplementary symbol generating means for generating a supplementary symbol expressing the auxiliary symbol and an adjunct word generating means for generating an adjunct word indicating the dependency relation immediately after the supplementary symbol generated by the supplementary symbol generating means. I am trying.

In the above structure, when the hollow clause detecting means detects that the dependent word and the dependent word of the dependency relation in the input sentence are divided into different clauses, the supplementary symbol generating means causes An auxiliary symbol indicating that the clause including the related word is omitted is generated immediately before the corresponding word in the translated word sequence of the clause including the receiving word, and the auxiliary word is generated immediately after the auxiliary symbol by the auxiliary word generating means. Is generated.
In this way, when the input word and the related word are divided into different clauses, it means that the section containing the related word has been omitted from the translated word string of the side clause containing the received word. Supplementary symbols are inserted and translation for each section is performed.

Further, the invention according to claim 5 is a syntax analysis unit for performing syntax analysis of a sentence in a source language input from an input unit, and a syntax conversion unit for converting the obtained syntax structure into a syntax structure in a target language. And a target language generator that generates a translation string in the target language of the input sentence based on this conversion result, and a division position determination that applies a division rule to determine the division position when dividing the input sentence into phrases or clauses. Section, a translation phrase construction unit that obtains a translation string in the target language for each phrase or clause based on the division result by the division position determination unit and the target language generated by the target language generation unit, In a machine translation device that translates an input sentence into phrases or clauses, the translated sentence phrase construction unit is separated by the division position determination unit and the case element is missing. The position of the missing case element in the translation column of, is characterized in that generating the translation of the antecedent relationship lyrics in the relative clause generated by the target language generator.

According to the above configuration, a translated word string in which the translated word of the antecedent is inserted as the translated word of the relative word is generated in the translated word string of the relative clause which is separated by the division position determination unit and lacks the case element. .

Further, the invention according to claim 6 is the machine translation device according to the invention according to claim 5, wherein the translated phrase construction unit detects the division position based on the syntactic structure obtained by the syntactic analysis unit. When a relative clause detecting means detects a relative clause in which each element separated by a part is missing, and the relative clause is detected by the relative clause detecting means, a translated word string of the input sentence generated by the target language generating portion Based on, the antecedent acquisition means for obtaining the translated word of the antecedent of the relative in the relative clause, and the translation of the antecedent obtained by the antecedent acquisition means, the missing case in the translation string of the relative clause. The relative word translation generating means for generating the relative word at the position of the element, and the case of the relative word translated immediately after the relative word translated by the relative word translation generating means. It is characterized in that it comprises a 置語 generating means after generating the 置語.

In the above structure, when the relative clause detecting means detects a relative clause which is separated by the division position deciding unit and lacks case elements, the antecedent obtaining means obtains a translated word of the antecedent of the relative clause. It is generated by the relative word translation generating means as a translation word of the relative word at the position of the missing case element in the translation word string of the relative clause. Further, the postposition word generation means generates the postposition word immediately after the translation word of the relative word. In this way, the translated word of the antecedent is inserted as the translated word of the relative clause into the translated word string of the relative clause that is separated by the division position determination unit and lacks the case element, and translation for each clause is performed.

The invention according to claim 7 is the machine translation apparatus according to the invention according to claim 6, wherein the translated sentence phrase constructing unit modifies the translated word of the relative word generated by the relative word translation generating means. It is characterized by including an auxiliary modifier generation means for generating a word and improving the readability of the clause in which the translation of the relative word is generated.

[0028]

BEST MODE FOR CARRYING OUT THE INVENTION The present invention will be described in detail below with reference to the embodiments shown in the drawings. FIG. 1 is a block diagram showing an embodiment of a machine translation apparatus of this embodiment.
In the present embodiment, the machine translation apparatus will be described by taking the case where the source language is English and the target language is Japanese as an example.

In FIG. 1, this machine translation device is configured by connecting an input unit 1, a translation control unit 2, a main memory 3, a storage unit 4, an output unit 5 and a control unit 6 via a bus line 7. There is. The input unit 1 is composed of an input device such as a keyboard, a mouse, an electronic pen, a tablet, a scanner, and a character recognition device, and a communication device connected to a communication line. Is entered. The translation control unit 2 includes a translation module 2A and an output format shaping unit 2B, and performs translation processing in phrase units on an original sentence in the source language input from the input unit 1 to generate a translated sentence in the target language. . The main memory 3 is provided with various buffers for holding data generated in the process of the translation control unit 2 generating a translated sentence. The storage unit 4 stores data such as a dictionary for translation used when the translation control unit 2 generates a translated sentence, grammatical rules, and other rules necessary for translation. The output unit 5 is a CRT (cathode ray tube), LCD (liquid crystal display device).
It is constituted by a display device such as a printer, a printing device such as a printer, or a communication device connected to a communication line, and outputs a generated translated sentence or the like. The control unit 6 is mainly C
It is configured by a PU (Central Processing Unit) and controls the input unit 1, the translation control unit 2 and the output unit 5 to execute various processes such as a translation process.

In general, machine translation has analysis levels as shown in FIG. When the original text in the source language is entered in the upper left of the figure, the dictionary lookup for level L1 and level L2
Morphological analysis, level L3 syntax analysis, and so on. Machine translation can be roughly divided into the following two, depending on which analysis level is used. One is to analyze an intermediate language, which is a concept that does not depend on any of the source language and the target language of level L6, and from that, generate context in the target language of level L7, generate meaning of level L8, generate syntax of level L9. , Is a pivot method for generating a translated sentence in a target language by proceeding with morpheme generation and generation at level L10. The other is to analyze the level L2 morphological analysis, level L3 syntactic analysis, level L4 semantic analysis, and level L5 context analysis to obtain the internal structure of the source language. This is a transfer method in which after converting the obtained internal structure of the source language into the internal structure of the target language at the same level, a translated sentence of the target language is generated.

Hereinafter, each of the above analyzes will be sequentially described. (A) Dictionary Lookup / Morpheme Analysis Look up the dictionary stored in the dictionary memory 4a of the storage unit 4 to divide the input sentence into each morpheme string (word string).
Then, the grammatical information such as the part of speech of each divided word and the translated word are obtained, and further the tense, the personal name, the number, etc. are analyzed. (B) Syntactic analysis The structure (structure analysis tree) of the input sentence such as the dependency between each word obtained as a result of the above morphological analysis is determined. (C) Semantic analysis From a plurality of syntax candidates obtained as a result of the above-mentioned syntax analysis, semantically correct ones and non-semantic ones are discriminated. (D) Context analysis Understand topics and remove omissions and ambiguities. It is assumed that the translation module 2A in this embodiment executes the translation process by the transfer method for performing the analysis up to the syntax analysis of at least the level L3 in FIG.

In order to execute the translation processing by the transfer method described above, the translation module 2A, as shown in FIG. 1, has a dictionary lookup / morphological analysis unit 2a, a syntax analysis unit 2b, a syntax conversion unit 2c, and a target language generation unit. Has 2d.
Further, each translated phrase is constructed from a target language word generated by the target language generation unit 2d and a division position determination unit 2e that determines a division position when the sentence is divided into phrases by referring to the data being translated. It has a translated phrase construction unit 2f.

In the main memory 3, the original text buffer 3a, the dictionary lookup buffer 3b, the syntax buffer 3 before conversion are stored.
c, post-conversion syntax buffer 3d, translation word buffer 3e, original language / translation word correspondence information buffer 3f, division position buffer 3g, translation word position buffer 3h for each phrase, translation phrase buffer 3i,
The output format buffer 3j, auxiliary verb buffer 3k, phrase translation flag buffer 3l, and tree structure pointer buffer 3m are set. In addition to the dictionary memory 4a, the storage unit 4 stores a grammar rule memory 4b, a postfix word table memory 4c, an auxiliary translation table memory 4d, which stores grammatical rules used by the syntax analysis unit 2b during syntax analysis. The other translation rule memory 4e, phrase division rule memory 4f, and auxiliary verb code table memory 4g are set.

Further, the translation module 2A according to the present embodiment, as shown in FIG. 2, has a target language generation unit 2d, in order to generate a translated phrase whose content is accurate and easy to understand when translating in phrase units. Tree structure tracing means 11, node information acquisition means 12, auxiliary verb sentence translation method determination means 13, auxiliary verb detection means 14, parallel node detection means 1
5, it has a dictionary description translation word generation means 16, an adjunct word generation means 17, and an auxiliary verb translation generation means 18. Further, as shown in FIG. 3, the translated phrase construction unit 2f includes a blank phrase detection unit 21, a relative clause detection unit 22, a generation position correction unit 23, an auxiliary symbol generation unit 24, and an auxiliary modifier generation unit 2.
5, it has an antecedent acquisition means 26 and a postpositional word generation means 27.

FIG. 4 is a flow chart of the phrase unit translation processing operation executed by the input unit 1, the translation control unit 2 and the output unit 3 under the control of the control unit 6. The phrase-based translation processing operation in this embodiment will be described below with reference to FIG.

In step S1, the original text is input from the input unit 1. Then, the input unit 1 recognizes the character string forming the original sentence and divides the input character string into partial character strings based on the blank character (space). Then, each divided partial character string (morpheme candidate) is stored in the original text buffer 3a with position information added thereto, as shown in FIG. Further, a translation process in phrase units (hereinafter referred to as phrase translation) is instructed from the input unit 1 and recognized by the input unit 1. Then, the phrase translation flag is set in the phrase translation flag buffer 3l of the main memory 3 according to the recognition result "phrase translation". In step S2, the dictionary lookup / morpheme analysis unit 2a of the translation module 2A executes the dictionary lookup / morpheme analysis process on the original sentence stored in the original sentence buffer 3a as described above, and the obtained analysis is performed. The result is stored in the dictionary lookup buffer 3b of the main memory 3. In step S3, the syntactic analysis unit 2b of the translation module 2A executes the syntactic analysis process to determine the structure analysis tree of the original sentence. Then, the obtained structure analysis tree is stored in the pre-conversion syntax buffer 3c in the main memory 3. In step S4, the structure conversion tree 2c of the translation module 2A converts the original structure analysis tree stored in the pre-conversion syntax buffer 3c into a Japanese structure analysis tree, and the converted structure analysis tree in the main memory 3 is obtained. It is stored in the syntax buffer 3d.

In step S5, the translation module 2A
The target language generation unit 2d executes the target language generation subroutine, which will be described in detail later, to determine the translation of each word of the original sentence. Then, each determined translation word is stored in the translation word buffer 3e of the main memory 3. Further, the source language / translation word correspondence information is stored in the source language / translation word correspondence information buffer 3f of the main memory 3. In step S6, the division position determination unit 2e of the translation module 2A
Thus, the division position of the original sentence into phrases is determined and stored in the division position buffer 3g of the main memory 3. Note that the phrase division position is determined by applying the phrase division rule stored in the storage unit 4 as shown in FIG. 8 to the structure analysis tree stored in the syntax buffer 3c before conversion. Done by. The actual operation of the division position determination at that time is to search the nodes of the above structural analysis tree in a predetermined order, check whether or not each phrase division rule is applicable at each node, and if applicable, Is determined as the phrase division position.

In step S7, the translation module 2A
By the translated phrase construction unit 2f of the original phrase / translated word correspondence information (see FIG. 15) created in step S5 and the division position information created in step S6 (see FIG. 16).
Based on the reference), the words of the translated words (hereinafter referred to as translated words) that form each phrase are obtained. Then, the word position (translated word position) of each translated word in the translated sentence is obtained by referring to the translated word buffer 3e and set in the translated word position buffer 3h for each phrase in the main memory 3. The format of the translated word position buffer 3h for each phrase is shown in FIG. In step S8, the translated word position of the translated word determined as a result of the execution of the target language generation subroutine in step S5 by the translated phrase construction unit 2f is not yet set in the translated word position buffer 3h for each phrase ( (For example, the phrase to which the phrase belongs is not determined.) For the translated word, the phrase to which the translated word belongs is determined and the translated word of the translated word is added to the corresponding phrase in the translated word position buffer 3h for each phrase, as described later in detail. The word position is set.

In step S9, the translated phrase construction unit 2f determines whether or not each translated word position set in the translated word position buffer 3h for each phrase needs to be corrected. If the translated word position exists, a correction operation is performed. The above-mentioned correction operation is performed as described in detail later according to the correction rule stored in the other translation rule memory 4e of the storage unit 4.

In step S10, the translated word phrase construction unit 2f uses the translated word buffer for each phrase stored in the translated word position buffer 3h for each phrase created in steps S7 to S9, based on the translated word position for each phrase. The translated word string for each phrase is obtained by referring to 3e, and the obtained translated word string for each phrase is stored in the translated phrase buffer 3i. In step S11, the output format shaping unit 5B adjusts the format of the translated word string for each phrase stored in the translated phrase buffer 3i to the format for output and stores it in the output format buffer 3j. The output format is, for example, as shown in FIG. 19, that the start positions of the original sentence and the translated sentence of each phrase are aligned and the correspondence between the original sentence phrase and the translated sentence that is the translated sentence is clear. This is the format used. In step S12, the output unit 5
As shown in FIG. 19, the output format buffer 3j
The original phrase and the translated phrase stored in are output.

FIG. 5 is a flowchart of a target language generation subroutine executed by the target language generation unit 2d in step S5 in the flowchart of the phrase unit translation processing operation shown in FIG. The target language generation subroutine will be described below with reference to FIG.

When the structure analysis tree of the original sentence is converted into the Japanese structure analysis tree in step S4 in the flowchart of the phrase-based translation processing operation shown in FIG. 4, the target language generation subroutine starts. In step S21, the parallel clause detecting means 15 of the target language generator 2d determines whether or not there are parallel phrases or clauses in the Japanese structural analysis tree obtained by the syntax converter 2c. It As a result, if they exist, the node corresponding to the parallel phrase or parallel node in the structural analysis tree is pointed, and the process proceeds to step S23. On the other hand, if it does not exist, the process proceeds to step S22. The above-mentioned "pointing to a node" means that the tree structure pointer buffer 3m of the main memory 3 mentioned above.
Is to store the node number in the structural analysis tree. In this way, the information on the node to be focused on at present is held. In step S22, the dictionary description translation word generation means 16 and the adjunct word generation means 17 determine the translation word of each word of the original sentence based on the translation word stored in the dictionary lookup buffer 3b and store it in the translation word buffer 3e. Further, the source language / translation word correspondence information is stored in the source language / translation word correspondence information buffer 3f. After that, the process returns to step S6 in the flowchart of the phrase-based translation processing operation shown in FIG.

In step S23, the contents of the auxiliary verb buffer 3k in the main memory 3 are cleared. Step S24
Then, the tree structure tracing means 11 determines the positional relationship between the respective lower nodes by referring to the parts of speech and grammatical elements of the respective lower nodes acquired by the node information acquiring means 12, The node corresponding to the longest child of the currently pointed node is pointed.

In step S25, the auxiliary verb detecting means 14 is used.
Determines whether the currently pointed node has an auxiliary verb. As a result, if there is an auxiliary verb, the process proceeds to step S26, and if not, the process proceeds to step S27. The auxiliary verb is detected as follows. That is, since the auxiliary verb here is an auxiliary verb on the English structural analysis tree, it does not exist on the converted (Japanese) structural analysis tree. Therefore, the tree structure tracing means 11 and the node information acquisition means 12 are controlled so that on the structure analysis tree before conversion (English) corresponding to the pointed point on the structure analysis tree after conversion (Japanese). This is done by tracing the nodes. In step S26, the auxiliary verb code of the auxiliary verb detected as described above by the auxiliary verb detecting means 14 is referred to the auxiliary verb code table stored in the auxiliary verb code table memory 4g of the storage unit 4 in a format as shown in FIG. Obtained. Then, the obtained auxiliary verb code is stored in the auxiliary verb buffer 3k.

In step S27, the auxiliary verb sentence translation method determining means 13 refers to the contents of the phrase translation flag buffer 3l to determine whether or not the phrase translation is instructed. As a result, if the phrase translation is instructed, the process proceeds to step S28, and if not, the process proceeds to step S34. In step S28, the dictionary description translation word generation means 16 and the adjunct word generation means 17 generate a translated word string suitable for the original word string belonging to the currently pointed node based on the translated words stored in the dictionary lookup buffer 3b. It is determined and stored in the translated word buffer 3e. At this time, the auxiliary verb translation generating means 18 transforms the end into a final form corresponding to the auxiliary verb code stored in the auxiliary verb buffer 3k. However, when the auxiliary verb code is not stored in the auxiliary verb buffer 3k, the auxiliary verb code is simply transformed into the final form.

Here, the adjunct word generating means 17 generates a translated word that does not correspond to the original word. As an example of the generation method, there is a method of using information on case structure. In this method, when generating a translated word corresponding to an original word, the case information of the original word stored in the dictionary memory 4a is referred to, such as a case particle in the target language (Japanese). It determines the adjunct.

In step S29, the tree structure tracing means 11 refers to the part-of-speech, the grammatical element and the like acquired by the node information acquiring means 12 to determine whether or not the currently pointed node has a younger brother node. To be determined. As a result, if there is a younger brother node, the process proceeds to step S31, and if not, the process proceeds to step S30. In step S30, the main memory 3
The contents of the auxiliary verb buffer 3k of are cleared. After that, the process returns to step S6 in the flowchart of the phrase-based translation processing operation shown in FIG.

In step S31, the tree structure tracing means 11 points the younger brother node determined to exist in step S29. In step S32, the dictionary description translation word generation means 16 generates the translation word of the conjunction at the younger brother node and stores it in the translation word buffer 3e. In step S33, the tree structure tracing means 11 points the next younger brother node (that is, the node parallel to the firstborn node in the conjunction). After that, the process returns to the step S25 and shifts to the process for the younger brother node.

If it is determined in step S34 that the phrase translation is not instructed in step S27, the tree structure tracing means 11 and the node information acquisition means 1
By 2, it is determined whether or not the currently pointed node has a younger brother node. As a result, if there is a younger brother node, the process proceeds to step S35, and if not, the above step S28.
Returning to, the translation of the current node is generated. Step S35
Then, the dictionary description translation word generation means 16 and the adjunct word generation means 17 determine a translation word word string suitable for the original language word string belonging to the currently pointed node and store it in the translation word buffer 3e. However, in this case, whether or not the auxiliary verb code exists in the auxiliary verb buffer 3k, the end of the section is transformed into a continuous discontinuation form in which the effect of the auxiliary verb does not reach. After that, the above step S
Returning to step 31, the point is set to the younger brother node determined to exist in step S34, and the processing shifts to the younger brother node. If it is determined in step S29 that there is no younger brother node anymore, the process returns to step S6 in the flowchart of the phrase-based translation processing operation shown in FIG. 4 via step S30.

Hereinafter, in the input section 1, the original sentence "It would push them to the end of" in which a parallel clause including an auxiliary verb exists.
The phrase-based translation processing operation shown in FIG. 4 and FIG. 5 will be specifically described by taking the case where "the creditors' lineand make it unlikely." is input as an example.

First, the original sentence read from the input section 1 is
It is stored in the original text buffer 3a as shown in FIG. Then, a dictionary lookup / morphological analysis process and a syntactic analysis process are performed to obtain a structure analysis tree of the original sentence as shown in FIG. 9 and stored in the syntax buffer 3c before conversion. Further, a syntax conversion process is performed to obtain a Japanese structural parse tree as shown in FIG. 10, and it is stored in the converted syntax buffer 3d.
... Steps S1 to S4

Further, since a parallel clause exists in the Japanese structural analysis tree, the node number "1" (see FIG. 10) of the "parallel clause" is stored in the tree structure pointer buffer 3m. Then, three nodes of the node “main node (node number = 2)”, the node “connective node (node number = 18)”, and the node “subject lacking main node (node number = 19)” belonging to the node “parallel node” Of the child nodes, the node “main node” is determined to be the first child, and the node number “2” of the node “main node” is determined.
Is stored in the tree structure pointer buffer 3m. ... Steps S21, S23, S24

Here, the node in the structural analysis tree (FIG. 9) before conversion corresponding to the node "main node (node number = 2)" which is the above-mentioned child is the node "main node (node number = 2)". , And the node “auxiliary verb (node number = 6)” exists below it. Then, the auxiliary verb code "9" of the auxiliary verb "would" corresponding to the node "auxiliary verb (6)" is obtained by referring to the auxiliary verb code table shown in FIG. 11, and is set in the auxiliary verb buffer 3k as shown in FIG. To be done. ... Steps S25 and S26

Since the "phrase translation" is commanded, the original word string "It them creditors'" belonging to the node "main clause (2)" currently pointed in the converted structural analysis tree shown in FIG. From "line of end to push", the translated word string of the node "main clause (2)" is determined based on the translated word stored in the dictionary lookup buffer 3b. Then, in this case, the end of the clause is transformed into the final form “to darou” according to the auxiliary verb code “9”. As a result, the translated words for each original word belonging to the node “main clause (2)” in the original sentence are “that”, “these”, “creditors”, “no”, “line”, “no”, “end”, and “insert”. It is determined. In addition, translated words “,” “wa” “wo” that do not correspond to the original words are generated and set in the translated word buffer 3e as shown in FIG. Here, as described above, the translated word that does not correspond to the original word is generated by referring to the case information stored in the dictionary memory 4a of the storage unit 4 by the attached word generation means 17, 34 pays attention to the contents of the "case name" stored in the dictionary memory 4a as shown in FIG. 34, and attaches the attached word corresponding to the contents of the "case name" to the storage unit 4 as shown in FIG. It is generated from the word table 4c. … Steps S27 and S28

When the above "phrase translation" is not commanded, the node "connective (18)" and the node "subject lacking main node (19)" are added to the node "main node (2)" currently pointed to. ) ”Exists as a younger brother node, so the node“ main clause (2) ”whose end is transformed into a continuous discontinuation form that does not have the effect of the auxiliary verb“ would ”
The translation word of the source word string belonging to is determined and the translation word buffer 3
It will be stored in e. Therefore, as shown in FIG. 14, the translated word of the original word “push” in the original sentence is “put”. ... Steps S27, S34, S35

After that, the younger brother node "conjunction (18)" exists at the node "main node (2)" currently pointed. Therefore, the younger brother node "conjunction (18)" is pointed to, and the translation "and" of the conjunction "and" is determined and set in the translation buffer 3e as shown in FIG. And furthermore
The node “main subject missing main clause (19)”, which is the next younger brother node regarding the node “main clause (2)”, is pointed. ... Steps S29 and S31
~ S33

Hereinafter, the node in the structural analysis tree before conversion corresponding to the node "subject missing main clause (19)" is the node "subject missing main clause (23)", and the node "auxiliary verb" exists below it. do not do. Therefore, the translated word string of the node "subject missing main clause (19)" is determined from the original word string "it unlikely make" belonging to the node "subject missing main clause (19)". However, in this case, the node "auxiliary verb" is added to the lower node.
, The contents of the auxiliary verb buffer 3k have not been updated. Therefore, the end of the section is "Main Section (2)"
Similar to the end of, the verb is transformed into the final form "to be" according to the auxiliary verb code "9" stored in the auxiliary verb buffer 3k. As a result, the node “subject missing main clause” in the original text is
The translated word for each original word belonging to (19) ”is determined to be“ that ”,“ impossible ”,“ will do ”,“. ”. In addition, a translated word "a" and "in a state" that do not correspond to the original word is generated and set in the translated word buffer 3e as shown in FIG. Here, the relevant node “subject lacking subject (19)”
Since there is no younger brother node in, the target language generation subroutine ends. ... Steps S25, S27 to S30

As described above, the translated word string obtained for each clause is set in the translated word buffer 3e.
At that time, word position information in the translated sentence is added to each translated word. The word position information in this case is not the phrase unit but the word position information in the unit of one whole original sentence. The reason for doing so is that it is possible to make many of the configurations common to the conventional machine translation device that outputs a translated sentence in units of one sentence. At the same time, the original word / translated word correspondence information indicating the correspondence between the position of each original word in the original sentence and the position of the corresponding translated word in the translated sentence is shown in FIG.
As shown in 5, it is set in the source language / translation word correspondence information buffer 3f. The original word position information necessary at that time is stored in the original buffer 3a in a format as shown in FIG.

Next, the phrase division rule shown in FIG. 8 is applied to the English structural analysis tree shown in FIG. 9 stored in the pre-conversion syntax buffer 3c as follows. The position is determined. First, the following applies to the node "main node" of node number "2":-Separate the parts grouped by the "main node" of phrase division rule 2. Here, the part that is summarized by the node "main clause" of the node number "2" is "It would push them to the end of the creditors' l
Since it is "ine", the original sentence is divided between the word "line" and the word "and". Next, to the node "subject lacking main clause" of the node number "23":-Separate the parts gathered at the nodes of the "main clause lacking case elements" of the phrase division rule 2 are applied. Here, since the part of the node number “23” that is summarized by the node “subject main missing clause” is “make it nulikely”, it is divided between the word “and” and the word “make”.

As a result of applying the phrase division rule as described above, the original sentence finally has the word "line" and the word "and".
"It would push them to the end of the creators'li" in two places, between the word "and" and the word "the".
It is divided into three phrases, "ne", "and", "make it unlikely". The word positions “1”, “12” and “1” in the original sentence of the first words “It”, “and” and “make” of each phrase thus divided
3 "is obtained by referring to the original text buffer 3a of FIG. 7. Then, each obtained word position is stored in the division position buffer 3g of the main memory 3 as shown in FIG.

Then, the word position of the translated word forming each obtained phrase is obtained as follows.
For example, the first phrase "It would push them to the end of the creditors' l
In the case of "ine", by referring to the "word position" column of "serial number 1" in the division position buffer 3g as shown in FIG. 16, the starting word position "1" of the first phrase in the original sentence is can get. Also, by referring to the "word position" column of "serial number 2", the start word position "12" of the second phrase and the end word position "11" of the first phrase
Is obtained. From the above, in the first phrase of the original sentence, each original word word position is "1, 2, 3, 4, 5, 6, 7, 8,
It can be seen that each of the original words is "9, 10, 11". Next, referring to the source word / translated word correspondence information buffer 3f as shown in FIG. 15, the translated word position "1, 12, 4, 11, 10 of the translated word corresponding to the original word forming the first phrase. , 7, 6, 8 ”is obtained. Furthermore, if the words are sorted in ascending order except for duplicates, the translated word position “1,4,6,7,8,10,11,12” of the translated word that constitutes the translation of the first phrase in the original sentence. Is obtained. The translation word position of the translation word that constitutes the translation of the first phrase thus obtained is stored in the "translation word position" column of phrase number 1 in the translation word position buffer 3h for each phrase.

Similarly, the translated word position "14" of the translated word that constitutes the translation of the second phrase of the original sentence and the translated word position of the translated word that constitutes the translation of the third phrase "16, 18,
20 "is obtained and stored in the" translated word position "column of the corresponding phrase number. … Step S7

Next, among the translated words, translated word positions whose translated word positions are not stored in the translated word position buffer 3h for each phrase, that is, translated word "ha", "wo", " The phrases to which "to" and "to the state" belong are determined to obtain respective translated word positions "2", "5", "17" and "19", and the corresponding phrases in the translated word position buffer 3h for each phrase are obtained. It is stored in the column of “translated word position”. As a result, the translated word position for each phrase is, as shown in FIG. 17, the first phrase 1,2,4,5,6,7,8,9,10,11,12 Phrase 14 The third phrase is 16,17,18,19,20.

Since the translated word for each phrase thus obtained does not have a translated word that presents a translated word position that needs to be corrected, a translation for each phrase is constructed as follows. First, the translated word positions of the translated words that make up the first phrase are 1, 2, 4, 5, and 5 according to the "translated word position" column of phrase number 1 in the translated word position buffer 3h for each phrase shown in FIG. It is obtained as 6, 7, 8, 9, 10, 11, and 12. Next, the translated word "it" at the translated word position 1 and the translated word "ha" at the translated word position 2 are obtained from the translated word buffer 3e shown in FIG. Below, the translated words "there", "o", "creditor", "no", "line", "no", "end", "ni" and "will enter" are obtained in order. Then, as shown in FIG. 18, the translation phrase buffer is obtained by connecting the translation words thus obtained in the ascending order of the translation word positions, and obtaining the character string "it will put them at the end of the creditor's line". Stored in the "translated phrase" column of phrase number 1 of 3i. Similarly, the translated word "and" that composes the second phrase,
The translation words “it”, “o”, “impossible”, “in the state” and “will” that form the third phrase are obtained, and the obtained translation word is set to the translation word position. The character strings obtained by connecting in ascending order are stored in the "translated phrase" column of the corresponding phrase numbers in the translated phrase buffer 3i. As a result, translations for each phrase are constructed as shown in FIG. … Steps S9 and S10

When the translated phrase for each phrase of the original sentence is obtained as described above, the output format shaping unit 2B
Thus, as shown in FIG. 19, the head positions of the original sentence phrase and the translated sentence phrase are aligned and displayed on the output unit 5.
... Steps S11 and S12

As described above, in the present embodiment, the target language generator 2d of the translation module 2A is
Auxiliary verb sentence translation determining means 13, auxiliary verb detecting means 14, parallel clause detecting means 15 and auxiliary verb translation generating means 18 are provided, and further, auxiliary verb codes of auxiliary verbs detected by the auxiliary verb detecting means 14 in the main memory 3 for each clause. An auxiliary verb buffer 3k for storing is provided. When an original sentence in which a parallel clause including an auxiliary verb exists is input to the input unit 1, the target language generation unit 2d causes the end of the translation of each clause forming the parallel clause to be an end form corresponding to the auxiliary verb. The modified target language is generated. Therefore, the original sentence "It would have a parallel clause containing an auxiliary verb."
push them to the end of the creditors' lineand ma
The result of the phrase translation for "ke it unlikely." It would push them to the end of the creditors' line and that would put them at the end of the creditor's line and make it unlikely. The effect of the auxiliary verb "would" appears on the translation of the first phrase as "I will put it in ", and the effect of the auxiliary verb "would" appears as the translation of the first phrase. The unnaturalness of not entering is eliminated.

Next, the original sentence "Keating is the subje" has a structure in which one clause is separated from the other clause.
ct of a class-action lawsuit allegingthey misled i
Investors about the safety of junk bonds. "is input from the input unit 1 as an example, the phrase-based translation processing operation shown in FIGS. 4 and 5 will be specifically described.

Similar to the case of the above-mentioned original sentence, the dictionary lookup / morphological analysis process, the syntactic analysis process and the syntactic conversion process are performed. As a result, a structural analysis tree in English is obtained as shown in FIG. Steps S1 to S4 Next, since there are no parallel clauses in the original sentence, the translated words are arranged in the order of arrangement of the original words in the converted Japanese structural analysis tree, and the translated words of each word of the original sentence are determined. It Then, the determined translated word is stored in the translated word buffer 3e as shown in FIG. Further, the original language / translated word correspondence information is stored in the original language / translated word correspondence information buffer 3f as shown in FIG. … Step S5

Next, for the node “main clause (16)” in the structure analysis tree before conversion as shown in FIG. 20, among the phrase division rules shown in FIG. The parts that are grouped at the nodes of are separated. Here, the node “main node” with node number “16”
The part summarized in is `` they misled investors about the safety of junk bo
Since it is "nds", the original sentence is divided between the word "alleging" and the word "they". Then, the word positions "1" and "10" in the original sentence of the first word of each of the phrases thus divided are shown in FIG. 23 together with the numbers of the phrase division rules applied when determining the phrase division points. It is stored in the division position buffer 3g. … Step S6

Next, the "word position" column in the division position buffer 3g as shown in FIG. 23 and the original word / translation word correspondence information buffer 3f as shown in FIG. 22 are referred to to compose each phrase in the original sentence. The translated word position of the translated word is obtained. Further, the translated word position of the translated word whose translated word position is not stored in the translated word position buffer 3h for each phrase is determined. In this way, the translated word position for each phrase is determined as shown in FIG. 24 and stored in the translated word position buffer 3h for each phrase. … Steps S7 and S8

Next, the blank phrase detection means 21 and the generation position correction means 23 in the translated phrase construction section 2f.
According to the correction rule stored in the other translation rule memory 4e of the storage unit 4, the auxiliary symbol generation means 24 determines whether or not each translated word position needs to be corrected and performs the correction operation as follows. Be seen. In this case, the above correction rule is: (I) If all of the following determination rules (1), (2), and (3) are satisfied, the symbol "~" and the translated word at the end of the phrase b are converted into the phrase a. It is generated at the discontinuous part of. (1) There is a discontinuity in the translated word position of the translated word that forms a certain phrase a, and all the translated word positions of the translated word that form another phrase b are one non-constant in the phrase a. Enter a continuous area. (2) The phrase division rule applied when dividing phrase b is rule number 2. (3) There is no source word corresponding to the translated word at the end of phrase b. It is.

First, in the first phrase stored in the translated word correspondence information buffer 3h for each phrase shown in FIG. 24 by the above-mentioned blank phrase detection means 21, each translated word position is a translated word position after sorting. Nevertheless, since the translated word position "15" is located next to the translated word position "2", this part is detected as a discontinuous part. In addition, it is detected that all the translated words (translated word positions 4 to 14) belonging to the second phrase enter this discontinuous portion. Therefore, in the determination rule (1), there is a discontinuous part in the translated word position of the translated word forming one phrase a, and all the translated word positions of the translated word forming another phrase b are the phrase a. “Enter within one discontinuous point in” is satisfied. Further, the content of the division position buffer 3g shown in FIG. 23 is referred to, and the applied rule number at the time of phrase division regarding the second phrase is "2". Therefore, it is determined that the determination rule (2) "The phrase division rule applied when dividing phrase b is rule number 2" is satisfied. The translated word position of the translated word at the end of the second phrase is "14". However, referring to the source word / translated word correspondence information buffer 3f shown in FIG. It can be seen that there is no source word that exists. Therefore, the decision rule
(3) It is determined that "the original word corresponding to the translated word at the end of phrase b does not exist" is satisfied. That is, since the original sentence satisfies all the determination rules (1), (2) and (3), it is determined that the content of the translated word position buffer 3h for each phrase needs to be corrected.

Then, according to the correction rule (I), the translated word "to" at the translated word position 14 located at the end of the second phrase is detected by the generation position correcting means 23 at a discontinuous portion in the first phrase. Certain translated word position 1
Moved in front of the 5 translation word “claim”. That is, in the present embodiment, the adjunct word generating means in claim 4 is realized by the adjunct word generating means 17 of the target language generating section 2d and the generation position correcting means 23 of the translated sentence phrase constructing section 2f. Further, the auxiliary symbol generation means 24 of the translated phrase construction unit 2f generates the symbol "~" before the translated word "to" in the discontinuous portion of the first phrase. As a result, the content of the translated word position buffer 3h for each phrase changes as shown in FIG. here,
The translated word position “−100” means the translated word position of the translated word “˜”. When the value of the translated word position is a negative number, the translated word stored in the translated word table memory 4d of the storage unit 4 is not used instead of using the translated word stored in the translated word buffer 3e. Is meant to be used. In the case of this example, the translated word (symbol) "..." With the code "-100" is obtained by referring to the auxiliary translated word table of the auxiliary translated word table memory 4d as shown in FIG. … Step S9

Then, based on the contents of the translated word position buffer 3h for each phrase, a translation for each phrase is constructed by referring to the translated word buffer 3e and the auxiliary translation table, and as shown in FIG. It is stored in the buffer 3i. Then, the head positions of the original sentence phrase and the translated sentence phrase are aligned and displayed as shown in FIG. ... Steps S10 to S12

As described above, in the present embodiment, the translation rule word position correction rule (I) is stored in the other translation rule memory 4e of the storage unit 4. Then, all of the judgment rules (1), (2) and (3) are satisfied by the hollow phrase detecting means 21 in the translated phrase construction unit 2f, so that the contents of the translated word position buffer 3h for each phrase can be corrected. When it is determined that the translation word is necessary, the translation position, which is the translation word located at the end of the second phrase, is corrected by the generation position correction means 23 and the auxiliary symbol generation means 24 of the translation phrase construction unit 2f according to the correction rule (I). The translation word "to" at position 14 is preceded by the symbol "~" and moved to the translation word "assertion" at translation word position 15 which is a discontinuity in the first phrase.

Therefore, according to the present embodiment, the original sentence “Keating is the subject of a class-action l” having a structure in which a certain clause is separated from the other clause so as to be completely omitted.
awsuit allegingthey misled investors about the saf
The result of the phrase translation of `` ety of junk bonds. '' is Keating is the subject of a class-action lawsuit alleging they are misled investors about the safety of junk bonds. The structure of the sentence, which was obtained as if the investor was mistaken about the bond safety, and the second phrase, which is a case element of the sentence, was completely separated from the place of "to" in the first phrase. It is easy to understand that each of the translated phrases has a word and each translated phrase is easy to read.

Next, the original sentence "The plan was opposed by some o" having a relational clause in which case elements are missing when separated.
f top executiveswho invested in the company's secu
"Rities." is input from the input unit 1 as an example, and the phrase-based translation processing operation shown in FIGS. 4 and 5 will be specifically described.

Similar to the above case, the dictionary lookup / morpheme analysis processing, the syntax analysis processing and the syntax conversion processing are performed. As a result, a structural analysis tree in English is obtained as shown in FIG. Steps S1 to S4 Next, since there are no parallel clauses in the original sentence, the translated words are arranged in the order of arrangement of the original words in the converted Japanese structural analysis tree, and the translated words of each word of the original sentence are determined. It Then, the determined translated word is stored in the translated word buffer 3e as shown in FIG. Further, the original language / translated word correspondence information buffer 3f stores the original language / translated word correspondence information as shown in FIG. … Step S5

Next, among the phrase division rules shown in FIG. 8, for the node “relative pronoun (15)” and the node “subject missing main clause (16)” in the structure analysis tree before conversion shown in FIG. If there is a "relative" contact point to the immediate left of the node of "main clause lacking case element" in Phrase Division Rule 5, then the contact point of "relative clause" and "main clause lacking case element" It is applied that the contact points of and are separated as one phrase. Here, the part in which the node “relative pronoun” with the node number “15” and the node “subject lacking main node” with the node number “16” are summarized is “who invested in the company'securi”.
Since it is "ties", the original sentence is divided between the word "executives" and the word "who". Then, the word positions "1" and "10" in the original sentence of the first word of each of the phrases divided in this way, together with the number of the phrase division rule applied at the time of determining the phrase division portion, are shown in FIG.
As shown in FIG. …
Step S6

Next, the "word position" column in the division position buffer 3g as shown in FIG. 32 and the original / translated word correspondence information buffer 3f as shown in FIG. 31 are referred to and each phrase in the original sentence is constructed. The translated word position of the translated word is obtained. Further, the translated word position of the translated word whose translated word position is not stored in the translated word position buffer 3h for each phrase is determined. Thus, the translated word position for each phrase is determined as shown in FIG. 33 and stored in the translated word position buffer 3h for each phrase. … Steps S7 and S8

Next, the relative clause detecting means 22, the generation position correcting means 23, the auxiliary modifier generating means 25, the antecedent obtaining means 26 and the postposition generating means 27 in the translated phrase constructing section 2f are used to store the data in the storage section 4. According to the correction rule stored in the other translation rule memory 4e, it is determined whether each translated word position needs to be corrected and the correction operation is performed as follows. The correction rule in that case is (II) if all the following judgment rules (4) and (5) are satisfied, the word corresponding to the antecedent of the relative phrase in phrase a is acquired from the immediately preceding phrase, and An appropriate modifier is added to the word, and the word is generated in the position of the missing case of the phrase a. (4) The rule division rule applied when dividing a phrase a is rule number 5. (5) There is a “main clause” in which the case element is missing in phrase a. It is.

First, by the relational clause detecting means 22,
The content of the division position buffer 3g shown in FIG. 32 is referred to, and the application rule number for phrase division regarding the second phrase is "5". Therefore, the determination rule (4) "Apply when division of certain phrase a is performed." The rule division rule is rule number 5, ”is determined to be satisfied. Also, referring to the structure analysis tree before conversion shown in FIG. 29, it can be seen that the second phrase has the node “subject-missing main clause (16)” which is the main clause lacking case elements. Therefore, it is determined that the determination rule (5) "there is a" main clause "in which the case element is missing in phrase a""is satisfied. That is, since the original sentence satisfies all the determination rules (4) and (5), it is determined that the content of the translated word position buffer 3h for each phrase needs to be corrected.

Then, the antecedent acquisition unit 26 searches the antecedent of the relative pronoun "who" in the second phrase from the first phrase according to the correction rule (II), and the noun "executives" is searched. Is determined to be the target antecedent. Further, since this noun “executives” is a word whose original word position is “9” in the original sentence,
The source word / translation word correspondence information buffer 3f shown in FIG. 1 and the translation word buffer 3e shown in FIG.
As the translated word corresponding to "ecutives", the "executive" whose translated word position is "11" is obtained. Then, by the generation position correcting means 23 of the translated phrase construction unit 2f, the translated word "stem" of the translated word number 11 is included in the translated word belonging to the phrase number 2 in the translated word position buffer 3h for each phrase shown in FIG. Is added.

At this time, it is necessary to determine the generation position of the translated word "stem" in the second phrase. This is based on the case grammar and is the same as in the case of the normal translated word generation. It is determined by the language generator 2d.
Here, in the present embodiment, the generation position of the case element of the sentence is described in the verb dictionary. And
In the case of the original sentence, the central verb of the clause forming the second phrase is “invested”, and the syntax of the clause is a modification of the form of subject + verb + prefix verb (missing subject). is there. This has already been determined by the syntax analysis unit 2b as shown in FIG. In addition, the translated word of the word “invest” is also determined based on the result of the syntax analysis processing.

FIG. 34 is a conceptual diagram of data relating to the translated word "invest" selected in the original sentence of the original word "invest" stored in the dictionary data memory 4a in the storage unit 4. The “first case” in FIG. 34 is the “subject” of the case in the original sentence, and the “second case” in FIG. 34 is the “preposition phrase” of the case in the original sentence. Then, it is assumed that the first case standard generation position at the time of generating Japanese is designated as "first". Then, since the translated word “stem” acquired by the antecedent acquisition unit 26 is the “subject” with respect to the original word “invested”, the generation position correction unit 22 sets the translated word “stem” to the second phrase. It is generated as a translation of the relative pronoun "who" at the first position in. That is, in the present embodiment, the relative position translation generation means in claim 6 is realized by the generation position correction means 23.

Further, Japanese case elements include postpositional words such as case particles. Information for generating this postword is also stored in the dictionary data memory 4a. Therefore, the postpositional word generation means 27 obtains the first case name "action subject" of the verb "invest" from the dictionary data memory 4a shown in FIG. 34, and as shown in FIG. With reference to the postword generation information stored in the table memory 4c, since the case name is the "acting subject" and the section including the case (section forming the second phrase) is the main section, Translated word "Executive"
“Ha” is generated as a postfix. Further, by the auxiliary modifier generation means 25 of the translated phrase construction unit 2f,
The modifier of the translated word "executive" is determined as follows.
That is, the translated word “executive” is a noun because it is a noun. Therefore, the auxiliary translation table memory 4 of the storage unit 4
As shown in FIG. 26 in d, "the" which is defined as a modifier of the word by referring to the stored information
Is selected, and its generation position is set immediately before the translation word “executive”.

As a result, the contents of the translated word position buffer 3h for each phrase are as shown in FIG. still,
When the value of the translated word position is a negative number, the translated word stored in the translated word buffer 3e is not used, but the postword table memory 4c (FIG. 35) and the auxiliary translated word table memory 4d of the storage unit 4 are used. Referring to FIG. 26, the code "-1"
01 "means that the translated word" that "should be generated, and the code" -1 "means that the translated word" ha "should be generated. … Step S9

Then, based on the contents of the translated word position buffer 3h for each phrase, the translated word buffer 3e (FIG. 30).
With reference to the postword table memory 4c (FIG. 35) and the auxiliary translation table memory 4d (FIG. 26), the translation for each phrase is constructed, and as shown in FIG.
Stored in. Then, the positions of the heads of the original sentence phrase and the translated sentence phrase are aligned and displayed as shown in FIG. ... Steps S10 to S12

As described above, in the present embodiment, the translation rule position correction rule (II) is stored in the other translation rule memory 4e of the storage unit 4. Then, the relational clause detecting means 22 in the translated sentence phrase construction unit 2f determines that the contents of the translated word position buffer 3h for each phrase need to be corrected because all the determination rules (4) and (5) are satisfied. Then, the antecedent acquisition unit 26 of the translated phrase constructing unit 2f follows the correction rule (II), and the word "executives" which is the antecedent of the relative pronoun "who" in the second phrase and its translated word " Get executives. Then, the translated word “stem” is generated in the translated word belonging to the phrase number 2 by the generation position correcting means 23 of the translated sentence phrase constructing section 2f, and further, by the auxiliary modifier word generating means 25 and the postfix word generating means 27. Translated word "Executive" which is a relative translation by generating a modifier "no" and a postfix "ha"
It is added to.

Therefore, according to the present embodiment, the original sentence "Th
e plan was opposed by some of top executiveswho in
The result of the phrase "vested in the company's securities." was the plan by opposed of some of top executives who was invested in the company's securities. As described above, the missing case element "the cadre is" for the second phrase is supplemented, and it is very easy to understand even if the second phrase, which is a relative clause, is read alone.

The algorithm for the phrase-based translation processing operation in this embodiment is not limited to the flowcharts shown in FIGS. 4 and 5.

[0092]

As is apparent from the above, the target language generation unit in the machine translation device according to the first aspect of the present invention, when there is a parallel clause that shares only a specific case in the input sentence, For each of the clauses that make up the clause, a translated word string in which the last word is converted to the final form according to the auxiliary verb detected from the parallel clause is generated, so a parallel clause containing an auxiliary verb that shares only the above specific case is generated. It is possible to generate a translated word string that reflects the aspect based on the auxiliary verb for any of the clauses to be formed.

Therefore, the original sentence "It would push them to the end of the credi" in which a parallel clause containing an auxiliary verb exists.
The result of the phrase translation for'tors' lineand make it unlikely. 'is that it would push them to the end of the creditors' line and that would put them at the end of the creditor's line and make it unlikely. The effect of the auxiliary verb "would" appears on the translation of the first phrase as "I will put it in ". That is,
According to the present invention, it is possible to eliminate unnaturalness due to the effect of the auxiliary verb not appearing in the translated sentences of the parallel clauses including the auxiliary verb.

The target language generation unit in the machine translation device according to the second aspect of the present invention includes a parallel clause detecting means for detecting a parallel clause sharing only a specific case, and an auxiliary verb from each clause forming the parallel clause. Since it has auxiliary verb detection means for detecting and auxiliary verb translation generating means for generating a translated word string in which the last word is converted to the final form according to the detected auxiliary verb, only a specific case is shared in the input sentence. When there is a parallel clause including the auxiliary verb, the translation based on the auxiliary verb is performed in each of the clauses forming the parallel clause. That is, according to the present invention, similar to the invention according to claim 1, it is possible to eliminate the unnaturalness due to the effect of the auxiliary verb not appearing in the translated sentence of each clause forming the parallel clause including the auxiliary verb.

In the machine translation device according to the third aspect of the present invention, the translated phrase construction unit includes the received word when the related word and the received word in the input sentence are divided into different sections. Since a supplementary symbol is generated just before the word of the word in the translation word of the clause, which indicates that the clause including the word of the dependency originally exists, the word of the dependency is included in the translation string of the clause including the word of the phrase. It is possible to perform translation for each clause that clearly indicates that the containing clause originally exists.

Therefore, the original sentence "Keating is the subject" has a structure in which one clause is completely omitted from another clause.
of a class-action lawsuit alleging they misled inv
The result of the phrase translation for `` estors about the safety of junk bonds. '' is `` Keating is the subject of a class-action lawsuit alleging they misled investors about the safety of junk bonds. They misleaded investors about the safety of junk bonds. " That is, according to the present invention, it is easy to understand that the clause on the engagement side, which is a case element of the clause, is completely omitted from the location of the supplementary symbol in the clause on the reception side, and a translated phrase that is easy to read can be obtained. .

Further, in the translation phrase construction unit in the machine translation device of the invention according to claim 4, the related word and the receiving word in the input sentence are divided into different clauses, and the clause including the related word is the receiving word. Hollow clause detecting means for detecting a hollow clause having a structure in which it is completely omitted from a clause containing, and auxiliary symbol generation for generating an auxiliary symbol immediately before the word of the word in the translated word string of the clause containing the word of the word. Since it has a means and an adjunct word generating means for generating an adjunct word immediately after the generated auxiliary symbol, when the related word and the receiving word in the input sentence are divided into different clauses, It is possible to perform translation for each clause by inserting an auxiliary symbol indicating that the clause on the word side of the relation has been omitted in the translated word string of the clause on the word side of the reception. That is, according to the present invention, as in the invention according to claim 3, it is easy to understand that the clause on the receiving side, which is a case element of the clause, is completely omitted from the clause on the receiving side, and a translation phrase that is easy to read is provided. Obtainable.

Further, the translation phrase construction unit in the machine translation device of the invention according to claim 5 is arranged such that the position of the missing case element in the translated word string of the separated relative case in which the case element is missing is related to the relational clause. Since the translated word of the antecedent of the phrase is generated, it is possible to obtain the translation result in which the translated word of the antecedent is inserted into the translated word string of the relative clause as the translated word of the relative phrase.

Therefore, the original sentence “The plan was opposed by” has a relational clause in which case elements are lost when separated.
some of top executiveswho invested in the company '
The result of the phrase translation for "s securities." is "The plan was opposed by some of top executives who invested in the company's securities. The executives invested in the company's securities." can get. That is, according to the present invention, the missing case element in the relational clause is compensated, and a very easy-to-understand translation result can be obtained even when read alone.

Further, the translated phrase construction unit in the machine translation device according to the sixth aspect of the present invention includes a relative clause detecting means for detecting a relative clause in which each element is separated and each element is missing, and an antecedent of the relative clause in the relative clause. An antecedent acquisition unit that obtains a translated word of the relative phrase, and a relative word translation generation unit that generates the translated word of the obtained antecedent at the position of the missing case element in the translation string of the relative clause as a translation of the relative phrase, Since the postposition word generating means for generating a postposition word is provided immediately after the generated translation word of the relative phrase, the translation word string of the relative clause separated by the division position determining unit and having the case element missing It is possible to obtain a translation result in which the translation of the antecedent is inserted as the translation of the relative. That is, according to the present invention, as in the invention according to claim 5, the missing case element in the relational clause is compensated, and the translation result of the relational clause which is very easy to understand even when read alone can be obtained.

Further, in the machine translation device of the invention according to claim 7, the translated phrase constructing unit generates an auxiliary modifier generating the modifier associated with the translation of the relative word generated by the relative translation generating means. Because it has the means
It is possible to further improve the readability of the relative clause in which the translated word of the relative word is generated.

[Brief description of the drawings]

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

FIG. 2 is a detailed configuration diagram of a target language generation unit in FIG.

FIG. 3 is a detailed configuration diagram of a translated phrase construction unit in FIG.

FIG. 4 is a flowchart of a phrase-based translation processing operation executed under the control of the control unit in FIG.

5 is a flowchart of a target language generation subroutine executed in the phrase unit translation processing operation shown in FIG.

FIG. 6 is an explanatory diagram of an analysis level by machine translation.

FIG. 7 is a conceptual diagram showing an example of the contents of the original text buffer in FIG.

FIG. 8 is a diagram showing an example of a phrase division rule.

FIG. 9 is a conceptual diagram showing an example of a syntax analysis tree before conversion.

FIG. 10 is a conceptual diagram showing an example of a syntax analysis tree after conversion.

11 is a conceptual diagram showing an example of the contents of an auxiliary verb code table memory in FIG.

FIG. 12 is a conceptual diagram showing an example of contents of an auxiliary verb buffer in FIG.

FIG. 13 is a conceptual diagram showing an example of contents of a translated word buffer in FIG. 1.

FIG. 14 is a diagram when the phrase translation is not commanded.
It is a conceptual diagram which shows the content of the translation word buffer corresponding to 3.

FIG. 15 is a conceptual diagram showing an example of contents of an original language / translation word correspondence information buffer in FIG. 1.

16 is a conceptual diagram showing an example of the contents of a division position buffer in FIG.

17 is a conceptual diagram showing an example of contents of a translated word position buffer for each phrase in FIG. 1. FIG.

FIG. 18 is a conceptual diagram showing an example of contents of a translated phrase buffer in FIG. 1.

FIG. 19 is a diagram showing a display example of a phrase translation result.

FIG. 20 is a conceptual diagram of a syntax analysis tree before conversion, which is different from FIG. 9.

FIG. 21 is a conceptual diagram showing contents of a translation word buffer different from FIG.

22 is a conceptual diagram showing the contents of an original language / translation word correspondence information buffer different from FIG.

23 is a conceptual diagram showing the contents of a division position buffer different from FIG.

FIG. 24 is a conceptual diagram showing the contents of a translated word position buffer for each phrase different from FIG.

FIG. 25 is a conceptual diagram showing the contents of a translated word position buffer for each phrase after correction is made to FIG. 24.

FIG. 26 is a conceptual diagram showing an example of the contents of an auxiliary translation table memory in FIG.

FIG. 27 is a conceptual diagram showing the contents of a translated phrase buffer different from FIG.

28 is a diagram showing a display example of a phrase translation result different from FIG.

29 is a conceptual diagram of a syntax analysis tree before conversion, which is different from FIGS. 9 and 20. FIG.

FIG. 30 is a conceptual diagram showing contents of a translation word buffer different from those in FIGS. 13 and 21.

FIG. 31 is a conceptual diagram showing the contents of an original language / translation word correspondence information buffer different from those in FIGS. 15 and 22.

FIG. 32 is a conceptual diagram showing contents of a division position buffer different from FIGS. 16 and 23.

FIG. 33 is a conceptual diagram showing the contents of a translated word position buffer for each phrase different from FIGS. 17 and 24.

34 is a conceptual diagram showing a part of the dictionary memory in FIG.

35 is a conceptual diagram showing an example of the contents of the postword table memory in FIG. 1. FIG.

FIG. 36 is a conceptual diagram showing the contents of a translated word position buffer for each phrase different from FIGS. 17 and 24;

FIG. 37 is a conceptual diagram showing the contents of a translated phrase buffer different from those in FIGS. 18 and 27.

FIG. 38 is a diagram showing a display example of a phrase translation result different from FIGS. 19 and 28.

[Explanation of symbols]

1 ... Input unit, 2 ... Translation control unit,
2A ... translation module, 2a ... dictionary lookup / morphological analysis unit, 2b ... syntactic analysis unit, 2c
… Syntax converter, 2d… Target language generator, 2
e ... division position determination unit, 2f ... translated phrase construction unit,
3 ... main memory, 3c ... syntax buffer before conversion,
3d ... Syntax buffer after conversion, 3e ... Translated word buffer, 3f ... Original language / translated word correspondence information buffer, 3g ... Dividing position buffer, 3h ... Translated word position buffer for each phrase, 3i ... Translated phrase buffer,
3k ... auxiliary verb buffer, 4 ... storage unit,
4a ... dictionary memory, 4c ... postword table memory,
4d ... auxiliary translation table memory, 4g ... auxiliary verb code table memory, 5 ... output section, 6 ... control section,
11 ... Tree structure tracing means, 13 ... Auxiliary verb sentence translation method determining means, 14 ... Auxiliary verb detecting means, 1
5 ... Parallel clause detecting means, 18 ... auxiliary verb translation generating means, 21 ... hollow phrase detecting means, 22 ... relative clause detecting means, 23 ... generation position correcting means, 24
... auxiliary symbol generating means, 25 ... auxiliary modifier generating means,
26 ... Antecedent acquisition means, 27 ... Postfix generation means.

Claims (7)

[Claims] 1. A syntactic analysis unit for performing syntactic analysis of a sentence in a source language input from an input unit, a syntactic conversion unit for converting the obtained syntactic structure into a syntactic structure in a target language, and based on the conversion result. A target language generation unit that generates a translation string in the target language of the input sentence, a division position determination unit that determines the division position when dividing the input sentence into phrases or clauses according to the division rule, and the division result by the division position determination unit. A machine for translating an input sentence into phrases or clauses, having a translation phrase construction unit for obtaining a translation string in the target language for each phrase or clause based on the target language generated by the target language generation unit and the target language. In the translation device, when the input sentence includes a parallel clause sharing only a specific case, each of the parallel clauses forming the parallel clause. Auxiliary verbs are detected from each of the parallel clauses, and the words at the end of each of the parallel clauses are converted to the final form corresponding to the detected auxiliary verb to generate a target language translation string and share only the specific case. A machine translation device characterized by performing a translation for each clause by reflecting a feature based on the auxiliary verb in any of the clauses forming a parallel clause including the auxiliary verb. 2. The machine translation device according to claim 1, wherein the target language generation unit includes a parallel clause that shares only a specific case based on the syntactic structure in the target language obtained by the syntactic conversion unit. A parallel clause detecting means for detecting, an auxiliary verb detecting means for detecting an auxiliary verb from each clause detected by the parallel clause detecting means, and an auxiliary verb information storage means for storing information indicating the auxiliary verb detected by the auxiliary verb detecting means. , For each clause detected by the parallel clause detecting means, an auxiliary verb translation generation for generating a translated word string in a target language in which the last word is converted to an end form according to the information stored in the auxiliary verb information storage means A machine translation device comprising means. 3. A syntax analysis unit for performing syntax analysis of a sentence in a source language input from an input unit, a syntax conversion unit for converting the obtained syntax structure into a syntax structure in a target language, and based on the conversion result. A target language generation unit that generates a translation string in the target language of the input sentence, a division position determination unit that determines the division position when dividing the input sentence into phrases or clauses according to the division rule, and the division result by the division position determination unit. A machine for translating an input sentence into phrases or clauses, having a translation phrase construction unit for obtaining a translation string in the target language for each phrase or clause based on the target language generated by the target language generation unit and the target language. In the translation device, the translated phrase construction unit has a clause in which a dependent word and a dependent word differ depending on the division in the division position determining unit. When divided, an auxiliary symbol indicating that the clause including the related word originally exists is generated immediately before the corresponding word in the translated word string of the clause including the related word. Machine translation device. 4. The machine translation device according to claim 3, wherein the translated phrase construction unit is based on a translated string in the target language obtained for each clause, and has a dependency relation in an input sentence. And a word of the receiving word are divided into different clauses, and a clause containing the related word has a structure in which the clause containing the word of the receiving word is entirely omitted from the clause, and a hollow element detecting means, When the hollowing-out clause is detected by the detecting means, an auxiliary symbol is generated immediately before the word of the receiving word in the translated word string of the clause including the receiving word, which indicates that the clause including the related word is omitted. A machine translation device comprising: a symbol generating unit; and an adjunct word generating unit for generating an adjunct word representing the dependency relation immediately after the auxiliary symbol generated by the auxiliary symbol generating unit. 5. A syntax analysis unit for performing syntax analysis of a sentence in a source language input from an input unit, a syntax conversion unit for converting the obtained syntax structure into a syntax structure in a target language, and based on the conversion result. A target language generation unit that generates a translation string in the target language of the input sentence, a division position determination unit that determines the division position when dividing the input sentence into phrases or clauses according to the division rule, and the division result by the division position determination unit. A machine for translating an input sentence into phrases or clauses, having a translation phrase construction unit for obtaining a translation string in the target language for each phrase or clause based on the target language generated by the target language generation unit and the target language. In the translation device, the translated phrase construction unit is arranged in the translated word sequence of the related clause separated by the division position determination unit and lacking case elements. The position of the missing case elements, the machine translation apparatus and generating a translation of the antecedent relationship lyrics in the relative clause generated by the target language generator. 6. The machine translation device according to claim 5, wherein the translated phrase construction unit is based on the syntactic structure obtained by the syntactic analysis unit.
A relative clause detecting unit that detects a relative clause that is separated by the division position detecting unit and has a missing case element; and an input generated by the target language generating unit when the relative clause is detected by the relative clause detecting unit. An antecedent acquisition unit that obtains a translation of the antecedent of the relative in the relative clause based on the translation of the sentence, and a translation of the antecedent obtained by the antecedent acquisition unit in the translation of the relative clause. Immediately after the relative word translation generating means for generating the relative word at the position of the missing case element, and the relative word translation for the relative word generated by the relative word translation generating means, the case of the translated word of the relative word is displayed. A machine translation device comprising a postword generation unit that generates a postword to represent. 7. The machine translation device according to claim 6, wherein the translated phrase construction unit generates a modifier associated with a translated word of the relative word generated by the relative word translation generation unit, and the relative word. A machine translation device comprising auxiliary modifier generation means for improving the readability of a clause in which the translated word is generated.