JP2001344236A - Method and device for learning translated word, storage medium and translation system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To automatically learn translated words by using documents of 1st and 2nd languages in the same field. SOLUTION: After previously securing the documents of the 1st and 2nd languages in the same field, the learning of translated words is executed by a used translated word judging processing part 3. The processing part 3 regards each word of the document of the 1st language as an entry word, makes the entry word and its part of speech a pair and extracts at least one translated word information. Respective words of the document of the 2nd language in the same field as that of the document of the 1st language are regarded as entry words, each entry word and its part of speech are made a pair and at least one or more translated word information is extracted. Whether the same translated word information as that of a word of the 1st language exists in the entry words of words of the 2nd language or not is retrieved, and when the translated word information coincides with the entry word, the entry word is learned as a used translated word.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a translation word learning method for learning a translation of each word in a first language using a document in a first language and a document in a second language, and reflecting the translation in a translation process. The present invention relates to a learning device, a storage medium, and a translation system.

[0002]

2. Description of the Related Art In general, in a translation system, a dictionary for each field is prepared in advance, a field of a document of a first language to be translated is specified, and a dictionary for a second language in the same field as the field is used for the first translation. Translation of language documents has been done.

[0003] However, aside from a general dictionary,
Preparing more specialized disciplines is very expensive in terms of price, and new words appear rapidly with the advancement of science and technology. There are issues that must be addressed.

Therefore, in the conventional translation system, when learning a translated word in a certain field and reflecting it in a translation result, the following several learning methods are used.

[0005] One of them is that after translating an original sentence, a user who has knowledge that can determine a correct translated word in the field of the original sentence displays and compares the original sentence and the translated sentence one by one, and changes the translated word interactively. This is a method of finding one suitable translation and saving the translation.

Another method is to, when a certain word appears in the original sentence, determine the field of the original sentence, find a translated word in the field, and save the translated word.

[0007]

[0005] Among the above two methods, the former method of learning a translated word is such that the learning of a translated word must be performed by a user who always has expert knowledge of the original sentence and the field of the original sentence. This is a very laborious and time-consuming task because it is difficult and it is necessary to find the appropriate translation while changing the translation interactively.

In the latter translation word learning method, since a translation word is found based on a specific word, the translation word reflected in the translation result may suddenly change depending on the sentence before and after the specific word appears in the original text. Alternatively, when a plurality of specific words in different fields appear in the same document, there is a problem that translation results are not uniform.

SUMMARY OF THE INVENTION The present invention has been made in view of the above circumstances, and it is possible to use a first language document and a second language document to automatically learn a translation in the field of a first language document. It is an object of the present invention to provide a translated translation word learning method, a translated translation word learning device, and a storage medium.

It is another object of the present invention to provide a translation system that effectively utilizes automatically translated words for translation.

[0011]

In order to solve the above-mentioned problems, a translated translation learning method according to the present invention comprises a first language extracting method for extracting one or more translations of words constituting a document in a first language.
And a second step of sequentially comparing words constituting a document in a second language in the same field as the field of the first language with the one or more translated word information. In the second step, the second step A third step of learning the translated word information as a translated word when there is translated word information that matches the bilingual word, thereby automatically learning the translated word in the field from a document in a different language in the same field. It is possible.

A translation translated word learning apparatus according to the present invention comprises a document input means for inputting documents in first and second languages in the same field, and one of words constituting the document in the first language.
A used translation word determination processing unit that compares one or more pieces of the first translated word information with the words constituting the document in the second language, and learns translated word information that matches this word as a used translated word of the word in the first language; It is a configuration provided with.

According to the present invention, with the above-described configuration, the document input means secures a document in the first language and a document in the second language in the same field in advance, and generates a state in which the translated word can be learned. In this state, the translated word determination processing unit compares one or more pieces of the first translated word information of the word constituting the document of the first language with the word constituting the document of the second language, and matches the word. It is possible to automatically learn the translated word information to be used as the translated word of the word in the first language, and accumulate a large number of translated words based on reference materials in various fields.

[0014] The translation system according to the present invention further comprises a document input means for inputting documents in the same field in first and second languages, and one or more first and second words constituting words in the first language document. A translated word determination processing unit that compares translated word information with words forming a document of the second language, and learns translated word information that matches the word as a translated word of the first language word; And a translation processing unit that translates the document in the first language using the learned translated word when translating the document.

According to the present invention having the above configuration, the automatically learned translated word can be effectively used for translation.

[0016]

Embodiments of the present invention will be described below with reference to the drawings.

First, an embodiment of a translated word learning method according to the present invention will be described.

(1) A word of each sentence of a document in the first language (for example, English) to be translated is used as a headword, the part of speech of the word is determined, and the headword and the part of speech are paired from the translation dictionary. At least one or more translation information is extracted and stored in the first language translation table. Here, the document in the first language is, for example, English reference data in a specific field recorded in a file or the like, or information input by a user from a keyboard or the like based on the English reference data in a specific field. Homepage written in HTML language and Web site linked to it
Page.

(2) Next, a word of each sentence of a document in a second language (for example, Japanese) in the same field as the field of the first language (for example, sports) is used as a headword, and the part of speech of the word is determined. Then, as a pair of the headword and the part of speech, at least one or more translated word information of the word is extracted from the translation dictionary and stored in the second language translated word table. Here, the second language document is, for example, Japanese reference material of a specific field recorded in a file or the like, or information input by a user from a keyboard or the like based on the Japanese reference material of a specific field. Or a Japanese homepage described in the HTML language and a Web page linked thereto.

(3) In the relation between the headword, part of speech, and translation information of the first language document stored in the translation word table and the headword, part of speech, and translation information of the second language document, the first language document If there is a matching headword among the headwords of the second language document stored in the translation word table based on the corresponding word information, the corresponding headword is extracted. , Are sequentially stored as learning candidate translations corresponding to the headwords of the document in the first language, and finally are automatically learned as one use translation.

(4) By automatically repeating the processing steps (1) to (3) described above, the used translation words are sequentially accumulated.

Therefore, according to the above-described method for learning a translated word, it is possible to use reference materials and the like in the same field having different languages.
It is possible to automatically learn the translation of a specific field of the second language, and to sequentially accumulate the translation corresponding to the field of the document without requiring specialized knowledge in that field.

Next, an embodiment of a translation system including a translated word learning device according to the present invention will be described with reference to FIG.

Referring to FIG. 1, reference numeral 1 denotes a document input unit for inputting a document described in a different language in the same field.
Reading a document such as a reference material in a different language in the same field, which is recorded in a file 1a such as a ROM or a floppy (registered trademark) disk, and stores it in the document storage means 2; Note that the file 1a itself may be considered as the document storage unit 2. The document input means 1 includes appropriate reference materials described by the user in different languages in the same field, for example, documents,
By inputting from the keyboard 1b or the like while referring to sports newspapers, magazines, etc., the information is stored in the document storage means 2, or necessary information is taken from a homepage or the like described in a different language of the same field and stored in the document storage means 2. By doing so, documents written in different languages in the same field can be obtained. In addition, by the access by the user,
When the translation processing is performed by displaying homepages described in the same field in different languages on the same screen, these homepages may be included in the document storage unit 2.

Reference numeral 3 denotes a file 1a, a different language of the same field stored in the document storage means 2 including homepage information,
For example, a translated word determination processing unit configured with a CPU that finds a translated word to be used for translation processing from a first language, for example, an English document to be translated, and a Japanese document, a second language, and automatically learns, A translation dictionary storage unit 4, a translation word learning storage medium 5, a determination processing data storage unit 6, and a learning dictionary storage unit 7 are connected to the determination processing unit 3.

The translation dictionary storage unit 4 stores the first language (original language).
Dictionary that has information necessary to analyze a sentence in English, a dictionary that replaces words in the first language with words in the second language, a dictionary that has information necessary to generate sentences in the second language, and other necessary dictionaries In addition, grammar rules for the first language and the second language are recorded.

The storage medium 5 stores a later-described translated word learning program, and includes, for example, a magnetic disk, a magnetic tape, a CD-ROM, a DVD-ROM, a floppy disk, an MO, a CD-R, a memory card, and the like. Is used.

The judgment processing data storage section 6 is a part for sequentially storing the data in the middle of performing the use translation word judgment processing in accordance with the translation word learning program stored in the storage medium 5, and as shown in FIG. 2, for example. It is composed of a first language translation word table 6a, a second language translation word table 6b, a used translation word candidate table 6c, and the like.

The first language translation word table 6a is divided into headword, part of speech, and translated word information areas. In the headword area, words constituting each sentence in the first language document are registered. The part of speech of the word is registered in the area, and at least one or more pieces of translated word information for the word are registered in the translated word information area. The second language translation table 6b is divided into headwords, parts of speech, and translation information areas in the same manner as the first language translation table 6a. Are registered, and the used word candidate table 6c is divided into headword, part of speech, translated word information, and heading name areas, and the headword area includes words for the first language. In the candidate translation word usable and heading name area, a headword in the first language related to the candidate translation is registered.

The learning dictionary storage unit 7 learns and stores a used translation word which is data of a used translation word determination result performed in accordance with the translation word learning program.

If the fields are stored in the tables 6a to 6c in addition to the headwords, parts of speech, translated word information, etc., the used dictionary can be registered in the learning dictionary storage unit 7 for each field. It can be used effectively in later translation processing.

Although each of the storage units 2, 4, 6, and 7 is individually provided, an area corresponding to an arbitrary plurality of storage units is formed, and one storage unit stores necessary data. It may be a configuration.

Further, as shown in FIG. 3, the used translation word determination processing section 3 reads a document in a first language of a specific field from the document storage means 2 and stores this document in the translation dictionary storage section 4. After parsing based on the dictionary and the grammatical rules, the first language word part-of-speech deciding means 31 determines each word as a headword, and determines the part of speech of the word being the headword. After the part of speech is determined, at least one or more translation information is extracted based on the dictionary of the translation dictionary storage unit 4 as a pair of the headword and the part of speech, and the first translation is registered in the first language translation table 6a. A second language document in the same field as the field of the first language document is read from the information extracting means 32 and the document storage means 2, and this document is structured based on the dictionary and grammar rules of the translation dictionary storage unit 4. After the analysis, each of the words is used as a headword, a second language word part of speech determination means 33 for determining the part of speech of the word being the headword, and the part of speech by the part of speech determination means 33 are determined. A second translated word information extracting unit 34 that extracts at least one or more translated word information based on the dictionary of the translation dictionary storage unit 4 as a pair with the part of speech and registers the translated word information in the second language translated word table 6b;
For the translation information of each headword registered in the first language translation table 6a, the second language translation table 6b
And a translation word learning means 35 for storing the translation word information in the learning dictionary storage unit 7 as the translation word used when the translation word information matches the headword.

8 is based on a translation processing program, uses necessary dictionaries and grammatical rules in the translation dictionary storage unit 4, and
The translation processing unit 9 translates a document in the first language into a document in the second language using the dictionary translation word stored in the learning dictionary storage unit 7. Is a translation processing reflection unit that reads out a dictionary stored in the learning dictionary storage unit 7 when the first language word is received under the first language and reflects it in the translation process.

The document translated by the translation processor 8 is output from the translation result output means 10. The translation result output means 10 is stored in the file 10a, displayed on the display unit 10b, or printed out to the printer 10c.

Next, the operation of the translation system including the above-described translation word learning apparatus or an example of the function realization of the CPU by the translation word learning program stored in the recording medium 5 will be described with reference to FIGS. .

When the operation of the system starts, the recording medium 5
After reading the translation word learning program from and storing it in an appropriate memory, a predetermined process is executed according to the program.

That is, the translated word determination processing unit 3 is configured as shown in FIG.
As shown, after reading a document in the first language of a specific field from among documents such as reference materials in different languages of the same field stored in the document storage means 2 in advance, including the file 1a and the homepage information 1c, Each sentence of this document is individually taken out, a syntax analysis is performed based on the grammar rules stored in the translation dictionary storage unit 4, and the part of speech of each word is determined (S1).
S4: first language word part of speech determination function). By repeating the processing of steps S1 to S4, the part of speech of each word constituting a certain sentence is determined. For example, a sentence in an English document that is the first language is “Heis ar
The part of speech is determined for each word for "eliver" as shown in FIG.

Subsequently, j ← 0 is set, the j-th word is extracted, and the pair of the headword and the part of speech of the extracted word is checked. Translated word information is extracted based on the headword, part of speech,
The translated word information is registered in the translated word table for first language 6a (see FIG. 2A) in the order of the translated word information (S5 to S9: first translated word information extraction function). In this case, for example, the English word "reliever"
In response to this, “rescuer”, “relief pitcher”,... Are extracted as translated word information.

With respect to the document in the first language, after extracting the translated word information for all the words of the i-th sentence (S10), the above-mentioned processing is repeatedly executed from the next sentence to the last sentence (S11).

Next, as shown in FIG. 6, the used translation word determination processing unit 3 stores in advance the second language in the same field as the document in the first language from the document storage means 2 including the file 1a and the homepage information 1c. After reading a document such as a reference material of a word, each sentence of this document is individually taken out and subjected to syntax analysis based on the grammatical rules stored in the translation dictionary storage unit 4.
The part of speech of each word is determined (S11 to S15: second language word part of speech determination function). These steps S11 to S15
Is repeated to determine the part of speech for each word constituting a certain sentence.

Subsequently, j ← 0 is set, the j-th word is extracted, and the pair of the headword and the part of speech of the extracted word is confirmed. Translated word information is extracted based on the headword, part of speech,
The translated word table 6b for the second language (FIG. 2B)
(S16 to S20: second translated word information extraction function). In this case, if the document in the first language is a sports field, and if the Japanese word "relief pitcher" in the same sports field is the "relief pitcher", as shown in FIG. .. Are extracted as translated word information.

With respect to the second language document, after extracting the translated word information for all the words of the i-th sentence (S21), the above-mentioned processing is repeatedly executed from the next sentence to the last sentence (S22).

Further, as shown in FIG. 7, the used translation word judgment processing unit 3 reads the contents of the first language translation word table 6a of the judgment processing data storage unit 6 (S23), and i in it is used.
The second headword “for example, reliever” is extracted (S24, S25), and j ← 0 and k ← 0 are set (S2).
6) The j-th translated word information “rescuer” is extracted. Thereafter, the second language translation table 6b is searched based on the j-th translated word information "rescuer" (S27, S28), and there is a headword that matches "rescuer" among the headwords. It is determined whether or not it is (S29). In this case, there is a mismatch, so if the search for the next word information “relief pitcher” in the second language translation word table 6b is similarly performed through step S32, the headword “relief pitcher” is present. Therefore, as shown in FIG. 2C, in addition to the headword and the part of speech that are words in the first language, the matched “relief pitcher” is registered as a candidate for a learning translation word in the translation candidate word table 6c as shown in FIG. S23 to S30: learning candidate translation word extraction function). Then, the same processing as described above is executed as to whether or not there is a heading name for the learning candidate translation word "relief pitcher" other than "reliever" (S31), and if it matches at step S29, it is additionally registered as a heading name.

Subsequently, the process proceeds to step S33, where it is determined whether or not there is one translation candidate word. If there is one, at least the headword and the learning candidate translation word of the used translation word candidate table 6c are extracted and the learning dictionary storage unit 7 is retrieved. And, if necessary, the part of speech, the heading name, and the field of the document are registered.

In step S33, the learning candidate translation is 1
If not, the process proceeds to step S35 to determine whether there are two or more learning candidate translations (S35). If there are two or more, only learning candidate translations that match the headword are learned. Register it in the dictionary storage unit 7 (S35 to S40).

Thereafter, it is determined whether or not the processing has been performed up to the last headword stored in the first language translation word table 6a. If the processing has been completed, the processing is terminated, or as shown in FIG. Transition.

When the translation processing section 8 receives the end of the translation word learning processing from the used translation word determination processing section 3, it executes the translation processing according to a translation processing program (not shown). That is, the translation processing unit 8 sets i ← 0, extracts the i-th sentence of the document in the first language, and executes the translation processing (S
51 to S53). By the way, in the translation process, whether a learning dictionary has already been created for documents in the first language,
Alternatively, when the learning dictionary is given a higher priority, a request is made to the translation processing reflecting unit 9 for each word of the document in the first language, and a translated sentence is generated using the learning dictionary registered in the learning dictionary storage unit 7. (S54, S55). Then, the translation is performed up to the last sentence of the first language, and the processing is completed (S56).

Therefore, according to the above-described embodiment, in the first language and the second language which are different from each other in the same field, each word of the document in the first language is used as a headword and the part of speech is determined. Word and part of speech as a pair 1
Extract one or more translated word information. Similarly, for the second language, the headword and the part of speech are paired, and one or more translated word information is extracted. Then, based on the translated word information of the document in the first language, search for the headword determined in the document in the second language,
If there is the same headword as the translated word information, the translated word information of the relevant field is learned as a used translated word, so that a suitable dictionary suitable for the field can be selected from documents such as various reference documents, and a learning dictionary can be created. In addition, dictionaries of various specialized fields can be sequentially created without providing expensive specialized field dictionaries, and can be used as those having high practical value.

Also, by taking in various electronic information,
It is possible to automatically learn the translations used in a specific field, and to easily create a learning dictionary without requiring any manpower and skill.

Furthermore, since the translations used are learned from documents in the same field and different languages, even if the first language document in the same field is translated using the learning dictionary, the translations do not suddenly change, and the unified You can perform translations with words.

The present invention is not limited to the above-described embodiment, but can be implemented in various modifications without departing from the scope of the invention. Further, the embodiments can be implemented in combination as much as possible, and in that case, the effect of the combination can be obtained. Furthermore, the above embodiments include various upper and lower stage inventions.
Various inventions can be extracted by appropriately combining a plurality of disclosed components. For example, if an invention is extracted because some constituent elements can be omitted from all the constituent elements described in Means for Solving the Problems, the omitted part may be omitted when implementing the extracted invention. Is appropriately supplemented by well-known conventional techniques.

[0053]

As described above, according to the present invention, it is possible to automatically learn a translation in the field of the first language document from the first language document and the second language document. It is possible to provide a translated word learning method, a translated word learning device, and a storage medium that can easily create dictionaries of various specialized fields without having a dictionary of the field.

Further, the present invention can provide a translation system that can generate a translated sentence using an appropriate translation suitable for the field by using the translation automatically learned.

[Brief description of the drawings]

FIG. 1 is a configuration diagram showing an embodiment of a translation system including a translated word learning device according to the present invention.

FIG. 2 is a diagram illustrating an example of an array of data in a determination processing data storage unit illustrated in FIG. 1;

FIG. 3 is a block diagram illustrating one function of a translation word determination processing unit 3 shown in FIG. 1;

FIG. 4 is a diagram illustrating translated word information in a sentence of a document in a first language.

FIG. 5 is a view for explaining the operation of the translated word learning device according to the present invention and a translated word learning program in a storage medium.

FIG. 6 is a diagram for explaining the operation of the translated word learning device according to the present invention and the translated word learning program of the storage medium, which are continued from FIG. 5;

FIG. 7 is a diagram for explaining the operation of the translated word learning device according to the present invention and the translated word learning program in the storage medium, which are continued from FIG. 6;

FIG. 8 is a flowchart illustrating the operation of a translation process.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 ... Document input means 3 ... Used translation word determination processing part 4 ... Translation dictionary storage part 5 ... Storage medium 6 ... Determination processing data storage part 7 ... Learning dictionary storage part 8 ... Translation processing part 9 ... Translation processing reflection part 10 ... Translation result Output means

 ────────────────────────────────────────────────── ─── Continuing from the front page (72) Inventor Etsuo Ito 1 Toshiba-cho, Fuchu-shi, Tokyo F-term in the Fuchu factory of Toshiba Corporation (reference) 5B091 AA06 BA02 CC02 CC15 CC16 EA01 EA17

Claims (5)

[Claims] 1. A first step of extracting one or more translation information of a word constituting a first language document, and a word constituting a second language document in the same field as the field of the first language. A second step of sequentially comparing the one or more pieces of translated word information, and a third step of learning the translated word information as a translated word when there is translated word information matching the word in the second language in the second step. And a step of learning a translated word. 2. A document input means for inputting a document in a first language and a second language in the same field, one or more first translation information of a word constituting the document in the first language, and the second language A translated word determination unit for comparing translated words that match the word with the words constituting the document, and learning the translated word information as the translated words of the word in the first language. 3. The translated translated word learning device according to claim 2, wherein the used translated word determination processing unit sets each word of the document in the first language stored in the document storage unit as a headword, and First language word part-of-speech determining means for determining part-of-speech; first part-of-speech information extraction for extracting and registering at least one or more pieces of translated word information by pairing the headword and part of speech after the part-of-speech determination by the part-of-speech determining means Means and a second language word that determines each word of a second language document in the same field as the field of the first language document stored in the document storage means as a headword and determines the part of speech of each word A part-of-speech determining means, a second part-of-speech information extracting means for extracting and registering at least one or more pieces of translated word information by pairing the headword and the part of speech after the part-of-speech determination means, and the first translated word Information extraction A translation word learning means for sequentially searching for headwords registered by the second translation word information extracting means with respect to the translation word information registered by the column, and learning as a used translation word when the translation word information matches the headword. A translation translation word learning device comprising: 4. The translation apparatus according to claim 2, further comprising: performing a translation process of the first language document by using the learned used translation word when translating the first language document. A translation system comprising a processing unit. 5. A storage medium storing a translated word learning program for learning a translated word using a first language document and a second language document in the same field, wherein a computer reads the translated word learning program, With each word of the first language document as a headword, the first language word part of speech determination function to determine the part of speech of each word, and after the part of speech determination by this part of speech determination function, the headword and part of speech are paired, A first translated word information extraction function for extracting and registering at least one or more translated word information, and using each word of a second language document in the same field as the field of the first language document as a headword, A second language word part-of-speech determination function for determining the part-of-speech of a word, and a part-of-speech determination by the part-of-speech determination function. Based on the translated word information extraction function and the translated word information registered by the first translated word information extraction function, the headwords registered by the second translated word information extraction function are sequentially searched, and the translated word information and the A computer-readable storage medium that realizes a translated word learning function of learning as a used translation when a headword matches.