JP3394856B2 - Translation method and machine translation device - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a translation method for translating an input document and a machine translation apparatus adopting this translation method, and particularly to a translation method and a machine translation capable of avoiding meaningless translation as much as possible. Regarding the device.

[0002]

2. Description of the Related Art In a machine translation apparatus that automatically translates a document written in one language into a document written in another language according to an operator's instruction, each sentence constituting a document to be translated is taken in order from the beginning. , Translate each sentence into a sentence in the other language. As described above, conventionally, the entire document is divided into sentence units and each sentence is individually translated.

[0003]

However, the translation target document contains various words and phrases. Then, among these, there are also words and phrases such as proper nouns and document names that do not make sense even when translated. For example, when the document is a paper, it is meaningless to translate a reference list inserted in the end of the document and referred to or cited in this paper, or a person's name, a book title, or a document name included in the paper. In particular, in many cases, it is easier to identify the actual document if the original name of the original document is not translated as the journal name of the document.

Further, when a name is long, an abbreviation may be used in a paper. In such a case, if the abbreviation is used first in the paper, it is generally written together with the official name,
After that, only abbreviations are used. For example, an abbreviated name and a formal name [MT (Machine Translat
It does not make sense to translate the phrase [ion]] into [machine translation (machine translation)].

[0005] In addition, without translating the inside of the parentheses (formal name), [MachineTranslati
on)] is considered, but the relationship between the abbreviation and the official name becomes unclear.

In order to prevent such an unnatural translated document, for example, the documents to be translated are sequentially displayed on the display screen of the display device so that the operator does not need to translate them. A method is conceivable in which the specified sentence or phrase is not translated by designating with a mouse or cursor, or by giving a list of phrases that are not translated.

However, in the method in which the operator designates the translation-unnecessary portion as described above, as a result, the operator needs to look over the entire document to be translated, so that the burden on the operator is significantly increased. I have a problem to do.

Further, the operator needs to have a language ability enough to understand the original document to be translated to some extent. The present invention has been made in view of such circumstances, author name included in the references listed in the input document, thus to detect the proper names of title, etc., translated <br/> unnecessary It is an object of the present invention to provide a translation method and a machine translation device that can suppress the translation of a part as much as possible and obtain an easily readable translation sentence.

[0009]

[0010]

[0011]

[0012]

[Means for Solving the Problems ] To solve the above problems
In the translation method of the present invention , each headline portion included in the input document described in the first language or located between the sentences is sequentially detected using the headline portion detecting unit , and this is detected. Judgment is made using the reference start part judgment section to determine whether the heading part that indicates the start of the description of the reference
However, as a result of this judgment, it was judged that the description of the reference was started.
When the next heading is detected , proper nouns such as author names, book titles, and literature names included in the documents or sentences that are input are detected and stored in the proper noun storage unit.
After being stored in the word storage unit, the input document or sentence written in the first language is sequentially translated into a document or sentence written in the second language using the translation unit, and the input document or sentence the inclusion of proper names stored in the proper noun storage unit sentences
Then, the translation noun is not translated by the translation unit using the translation unnecessary phrase setting unit, and the proper noun is described as it is at a corresponding position in the document described in the second language.

Further, in the machine translation apparatus of another invention, a headline portion detecting means for detecting each headline portion included in the document described in the input first language or located between the sentences. When the heading part detected by the heading part detection means is a heading part indicating whether the description of the reference is started or not, and when the reference start determination means determines that the reference is started , A proper noun detecting means for detecting proper nouns such as an author name , a book title, and a document name included in a document or a sentence input until the heading portion detecting means detects the next heading portion, and a proper noun detecting means. a proper noun storage unit for storing the detected proper names, each
After the proper noun is stored in the proper noun storage unit, a translation unit that sequentially translates the input document or sentence described in the first language into each sentence described in the second language is input. When the document or sentence includes a proper noun stored in the proper noun storage unit, the proper noun is directly translated in the document described in the second language without translation in the translation unit. And a translation unit stop control unit described in the position.

[0014]

[0015]

[0016]

[0017]

[0018]

In the translation method and machine translation device configured as described above, the column (text) in which the list of references is described
The proper nouns such as the author name, the book title, and the document name included in each sentence are detected and stored.

When a proper noun stored in the storage unit appears in the process of sequentially translating each sentence of the input document, this proper noun can be regarded as a proper noun related to the reference document. , It is not translated and remains as the original text.

[0020]

[0021]

Embodiments of the present invention will be described below with reference to the drawings. (First Embodiment) FIG. 1 is a block diagram showing a schematic configuration of a machine translation apparatus to which the translation method of the first embodiment is applied.

The machine translation device of this embodiment is composed of an information processing device such as a computer. Then, the control unit 1 configured as a software on the application program stores a document to be translated written from the outside via an input unit, for example, a storage unit 2 formed in an HDD, a CRT. A display unit, a bitmap display, goggles, a display unit 3 for displaying a document before translation such as projection on a wall or a translated sentence, a document to be translated is input, and an operator inputs various operation commands. An input unit 4 including a keyboard, a mouse, a light pen, a digitizer, and the like for operating is connected.

Further, the control unit 1 is connected to a translation unit 5, a headline unit detection unit 6, and a reference document start determination unit 7, which are also software-configured on the application program. A knowledge dictionary 8 formed on a database is connected to the translation unit 5 .

The translation unit 5 translates the input sentence described in the first language into a sentence in the second language while referring to the knowledge dictionary 8. The headline portion detection unit 6 detects a headline portion from each word / phrase included in each input sentence. Specifically, each piece of information shown below to identify the heading included in the document (for example, in English) (a) All capital letters or all capital letters of capital letters (b) Chapter number, section number, etc. There is no (C) period, etc. at the beginning of the line, and there is a line break, and the next line and no connection are detected, and the part containing this information is detected as the heading part.

The reference reference start determination unit 7 next determines whether or not the description content of the headline portions sequentially detected by the headline portion detection unit 6 indicates the description start of the reference literature.

The control unit 1 has a function of controlling the operation of the entire machine translation apparatus. Then, the translation process for the document stored in the storage unit 2 is executed according to the flowchart shown in FIG.

The flow chart is started, and one of the documents stored in the storage unit 2 at P (program step) 1 is started.
Read one sentence and call it document A. At P2, the sentence number i of one sentence of the document A is initialized to 1, and the document sentence flag F is set.
Is initialized to 0.

At P3, the i-th sentence of the document A is read to be a sentence B. Then, the sentence B is sent to the heading section detecting section 6 (P4). The headline portion detection unit 6 determines whether or not this sentence is a headline portion by the method described above. In the case of the headline portion (P5), the sentence B is sent to the reference start determination unit 7 (P6).

The reference start determination unit 7 determines whether or not the relevant headline portion is the start of reference according to the flowchart shown in FIG. That is, sentence B is [reference], [reference], [ Re
In the case of a phrase such as [ ference ], this heading section indicates the start of the reference , and in other cases, it is determined that the heading section is a normal heading section.

In P7 of FIG. 3, when the corresponding heading section indicates the start of the reference document, the document sentence flag F is set to 1 (P8). Then, the sentence B is sent to the translation unit 5 (P9), the translation result is obtained from the translation unit 5, and is output to, for example, the display unit 3 (P10). When the translation process for one sentence is completed and it is confirmed that the sentence number i has not reached the final sentence number i _E (P11), the sentence number i is updated (P11).
12), return to P3.

Further, in the case of the normal heading part other than the start of the document in P7, the document sentence flag F is reset to 0 (P13).
If the sentence B is not a heading portion but a normal sentence at P5, the value of the document sentence flag F is checked (P14), and 1
If it is set to, the sentence B is determined to be a sentence in the document sentence, and the untranslated sentence B is directly transferred to P10 as a translation result. If it is set to 0 in P14, the corresponding sentence B is a sentence included in a normal sentence, so P
Go to 9 and translate the sentence B.

In P11, when the sentence number i reaches the final sentence number i _E , the translation process for this document A ends. In the machine translation device configured as described above, the headline portion detection unit 6 sequentially searches for each headline portion included in the document A to be translated stored in the word storage unit 2. When the heading portion is detected is determined to be a header portion indicating a description starts references in reference start determination unit 7 included from the header portion before the next header portion
That each sentence is not translated.

That is, a line is newly added to the final position of a document such as a paper, followed by a heading portion such as [references], and a column (sentence) in which a list of documents cited or referred to in this paper is described. ), This list of documents is not translated.

(Second Embodiment) FIG. 4 is a block diagram showing the schematic arrangement of a machine translation apparatus according to the second embodiment of the present invention. The same parts as those in the first embodiment of FIG. 1 are designated by the same reference numerals. Therefore, detailed description of the overlapping portions will be omitted.

In the machine translation device of this embodiment, in addition to the translation unit 5, the headline detection unit 6, and the reference start determination unit 7, the proper noun extraction unit 9 and the translation unnecessary phrase setting unit 1 are provided.
0 and a proper noun storage unit 11 are provided.

The proper noun extracting unit 9 extracts proper nouns such as an author name, a book title, a magazine name, a publishing company name, etc., which are included in each input sentence. Specifically, the part delimited by "", "", and "," is extracted as one proper noun block. With this, for example, "E.ITHO, New Machine Translation
System, T-journal, 1994 ”from a series of phrases" E.ITHO "
Proper nouns such as "New Machine Translation System" and "T-journal" can be extracted.

The proper noun storage unit 11 stores the proper nouns extracted by the proper noun extraction unit 9. In addition, the translation-unnecessary phrase setting unit 10 does not need to translate the proper noun when a proper noun stored in the proper noun storage unit 11 is found in each sentence B of the document A sequentially read from the storage unit 2. It has a function of giving a flag.

Then, the control unit 1 executes the translation process for the document stored in the storage unit 2 according to the flowcharts shown in FIGS. The processing operation from Q1 to Q8 in the flowchart of FIG. 5 is the same as the processing operation from P1 to P8 in FIG.

In Q8, the document sentence flag F is set to 1 when the detected headline portion indicates the start of the reference document,
When the sentence number i has not reached the final sentence number i _E , the process proceeds to Q9, the sentence number i is updated (Q10), and the process returns to Q3.

Further, in the case of a normal heading part other than the start of the document in Q7, the document sentence flag F is reset to 0 ( Q12 ).
If the sentence B is not a headline portion but a normal sentence in Q5, the value of the document sentence flag F is checked ( Q13 ),
If it is set to 1, it is determined that the sentence B is a sentence in the document sentence, and the sentence B is changed to "" or "," as described above.
Separate the parts separated by "" and "," (Q14),
The separated words and phrases are regarded as proper nouns and registered in the proper noun storage unit 11 (Q15). Then, proceed to Q9. Q1
If it is set to 0 in 3, the relevant sentence B is a sentence included in a normal sentence, so the process proceeds to Q9.

When the sentence number i reaches the final sentence number i _E in Q9, a proper noun storage unit for proper nouns such as author names, book titles, journal names, publisher names, etc. appearing in the document sentence for this document A. The registration process for 11 is completed.

Next, in Q16, the sentence number i is returned to the initial value of 1, and the actual translation process for the document A is started. Q17
Then, the i-th sentence of the document A stored in the storage unit 2 is read and set as the sentence C. In Q18, this sentence C is sent to the translation unnecessary phrase setting unit 10.

In Q19, the translation-unnecessary phrase setting unit 10 determines whether the phrase of the proper noun stored in the proper noun storage unit 11 is included in the input sentence C, and if it is included, Set the corresponding phrase as a phrase that does not require translation. Then, the sentence C including this translation unnecessary phrase is sent to the translation unit 5. If the translation-unnecessary phrase is not included, this sentence C is directly used in the translation unit 5
Send to.

The translation unit 5 translates the sentence C input from the translation unnecessary phrase setting unit 10 (Q21). In this case, if the sentence C has a phrase set as a translation unnecessary phrase, this phrase is not translated.

When the translation process for one sentence C is completed, it is confirmed that the sentence number i has not reached the final sentence number i _E (Q22), and the sentence number i is updated (Q23), Q1.
Return to 7.

When the sentence number i reaches the final sentence number i _E in Q22, the translation process for this document A is completed. In the machine translation device configured as described above, the storage unit 2
The proper noun storage unit 11 detects the proper nouns such as the author name, the book title, and the document name included in each sentence C in the sentence in which the list of the reference documents is written at the end of the document A to be translated stored in Stored in.

Then, each sentence C of the input document A is input.
When a proper noun stored in the proper noun storage unit 11 appears in the process of translating in sequence, this proper noun can be regarded as a proper noun related to a reference document, and thus is not translated and remains as the original sentence. Be done.

That is, since proper nouns relating to the references included in the text, not in the reference section of the document A, are not translated, unnecessary common translation is not performed, and a common sense and very readable document is obtained. it can.

(Third Embodiment) FIG. 7 is a block diagram showing the schematic arrangement of a machine translation apparatus according to the third embodiment of the present invention. The same parts as those in the second embodiment of FIG. 4 are designated by the same reference numerals. Therefore, detailed description of the overlapping portions will be omitted.

In the machine translation device of this embodiment, a continuous unified phrase detecting unit 12 is provided in addition to the translation unit 5 and the translation unnecessary phrase setting unit 10. The continuous unified phrase detecting unit 12 checks whether or not an abbreviated name and a formal name are written together in the input sentence. For example, if the terms of the “FAT (Fair Trade Commission)” exist, the parentheses are broken down into words such as “fair,” “trade,” and “committee,” and the initials are collected. JFTC ”is compared with the word immediately before the parentheses, and if they are the same, it is determined that the abbreviation and the official name are written together. In addition, "MAX (MultiAmeni
In the case of a continuous phrase of "ty eXpress", the upper case part in the parentheses is compared with the phrase immediately before the parenthesis, and if they are equal, it is determined that they are written together.

The translation-unnecessary phrase setting unit 10 adds a translation-unnecessary flag to both of the phrases when the sentences having the same meaning are consecutive in each sentence of the document sequentially read from the storage unit 2. Have the function to

Then, the control unit 1 executes the translation process for the document stored in the storage unit 2 according to the flowcharts shown in FIGS. The flowchart of FIG. 8 is started, and S
In step (1), one sentence of the documents stored in the storage unit 2 is read out to be a document A. In S2, the minute number i of one sentence of the document A is initialized to 1. In step S3, the i-th sentence of document A is read and set as sentence B. Then, this sentence B is sent to the continuous identical word detection unit 12 (S
4).

The continuous identical phrase detecting section 12 determines whether or not there is a continuous phrase (coincident phrase) having the same meaning in the sentence B according to the flowchart shown in FIG. When the sentence B is received in S11 of the flowchart of FIG. 9, if the sentence B has parentheses (S12), the character string in the parentheses is E, and the word (phrase) immediately before the parentheses is D (S13). The character string E is morphologically analyzed and divided into words (S14). Then, (1) if the first character of each word of the character string E can be spliced, and it matches the word (phrase) D immediately before the parenthesis (S15) (2) If the upper case part of the character string E can be spliced, immediately before the parenthesis (S16) (3) If the font in the character string E is different from the other characters, the word (phrase) D immediately before the parenthesis is matched.
If any of the conditions is satisfied (S17), the process proceeds to S21, and it is determined that the read sentence B has a word / phrase.

When none of the conditions (1), (2) and (3) is satisfied, the character string E is translated by the translation unit 5 in S18 and the translation result is set to F. Further, in S19, the word (phrase) D immediately before the parenthesis is similarly translated and the translation result is set to G. In S20, the translation results F and G are compared, and if they are equal,
It is determined that the read sentence B has a phrase written together.

If they are not equal in S20, or if the read sentence B does not have parentheses in S12, it is determined in S22 that there is no continuous phrase having the same meaning (coincident phrase) in this sentence B. To do.

When the determination process in the continuous identical word detecting unit 10 is completed, the process returns to S5 in FIG. 8, and if there is a part to be written together, the process proceeds to S6. In S6, the translation-unnecessary phrase setting unit 10 is activated to set two consecutive words having the same meaning (sentence phrase) in sentence B as a translation-unnecessary phrase, and the word set in this translation-unnecessary phrase is The existing sentence B is sent to the translation unit 5 (S7).

If it is determined in S5 that there is no additional writing part, the input sentence B is directly converted into the translation unit 5
(S7). The translation unit 5 translates the input sentence B and outputs the translation result to, for example, the display unit 3 (S
8). In this case, when the sentence B has a phrase set as a translation unnecessary phrase, this phrase is not translated.

When the translation process for one sentence B is completed, it is confirmed that the sentence number i has not reached the final sentence number i _E (S9), the sentence number i is updated (S10), and the process returns to S3. .

In S9, when the sentence number i reaches the final sentence number i _E , the translation process for this document A ends. In the machine translation apparatus configured as described above, in the document A to be translated stored in the storage unit 2, when words having the same meaning are consecutive, such as [abbreviation (formal name)], , Neither phrase is translated.

In the apparatus of the embodiment, "This is a MTS (Ma
chine Translation System). ”is not translated to“ This is MTS (Machine Translation System) ”, and“ This is MTS (Machine Translati
on System). "

In this way, when both the abbreviation and the official name are written together, both are not translated, so that it is possible to prevent unnatural sentences when both are translated.

Further, as described above, if only one of the parentheses (formal name) is not translated, the relationship between the abbreviation and the official name becomes unclear.
In the present invention, such unnaturalness is eliminated.

[0063]

As described above, in the translation method and machine translation apparatus of the present invention, a headline portion included in a document or a sentence is detected, and the detected headline portion indicates the start of writing the reference. In the case of, the author name, book title, and literature name included from this heading to the next heading
A proper noun, such as "," is detected and stored in the storage unit. So
Then, translate each sentence of the input document in order.
In the process, proper nouns stored in the memory appear.
And this proper noun is a proper noun related to a reference
Since it can be regarded as, it is not translated and remains as the original text. Shi
Therefore, there is no need for translation in references and text.
Since nouns are not translated, a natural translated document can be obtained.
It

[0064]

[0065]

[Brief description of drawings]

FIG. 1 is a block diagram showing a schematic configuration of a machine translation device to which a translation method according to a first embodiment of the present invention is applied.

FIG. 2 is a flow chart showing the operation of a reference start determination unit of the apparatus of the embodiment.

FIG. 3 is a flowchart showing an operation of a control unit of the apparatus of the embodiment.

FIG. 4 is a block diagram showing a schematic configuration of a machine translation device to which the translation method of the second embodiment of the present invention is applied.

FIG. 5 is a flowchart showing the operation of the control unit of the apparatus of the embodiment.

FIG. 6 is a flow chart showing the operation of the same control unit.

FIG. 7 is a block diagram showing a schematic configuration of a machine translation device to which the translation method of the third embodiment of the present invention is applied.

FIG. 8 is a flowchart showing the operation of the control unit of the apparatus of the embodiment.

FIG. 9 is a flowchart showing an operation of a continuous identical phrase determination unit of the apparatus of the embodiment.

[Explanation of symbols]

1 ... Control unit, 2 ... Storage unit, 3 ... Display unit, 4 ... Input unit, 5
... translation unit, 6 ... headline portion detection unit, 7 ... reference document start determination unit, 8 ... knowledge dictionary, 9 ... proper noun extraction unit, 10 ... translation unnecessary phrase setting, 11 ... proper noun storage unit, 12 ... consecutive identical phrases Detection unit

Claims (2)

(57) [Claims] 1. A finding each heading portion located between Included or sentence in a document described in a first language inputted
Sequential detection is performed using the partial detection unit, and it is determined using the reference document start unit determination unit whether or not the detected headline portion indicates the start of description of the reference document.
And, as a result of this determination, it is determined to describe the beginning of the reference and
When the next headline part is detected , proper nouns such as author names, book titles, and document names included in the document or sentence that are input are detected and stored in the proper noun storage unit, and the proper nouns are the proper nouns. After being stored in the noun memory,
Translates the input document or sentence described in the first language.
The translation unit is used to sequentially translate the document or sentence described in the second language, and the proper noun memory is stored in the input document or sentence.
When a proper noun stored in the section is included, no translation is required
The proper noun is not translated by the translation unit using the constant part , and the proper noun is directly described at the corresponding position in the document described in the second language. And how to translate. 2. Heading portion detecting means for detecting each heading portion included in a document described in the input first language or positioned between sentences, and heading portion detecting means for detecting the heading portion. Reference start determination means for determining whether or not the headline part is a headline part indicating the start of description of the reference, and when the reference start determination means determines that the reference starts, the headline part detection means Author name included in the document or sentence input until the heading of
Proper noun detecting means for detecting proper nouns such as book titles and literature names, proper noun storing section for storing proper nouns detected by the proper noun detecting means, and each proper noun stored in the proper noun storing section After being done
A translation unit that sequentially translates the input document or sentence written in the first language into each sentence written in the second language; and the proper noun memory for the input document or sentence. When a proper noun stored in a part is included, the proper noun is not translated in the translator, and the proper noun is directly written at a corresponding position in the document described in the second language. A machine translation device comprising stop control means.