JPH08329081A - Method and system for machine translation - Google Patents

Abstract

PURPOSE: To accurately translate a pair of a word and an affix appended to the word when the pair is found in the same document. CONSTITUTION: A translation processing part 3 translates a document in an input buffer 2 by referring to a dictionary 4 for translation. At this time, an affix representation processing part 5 checks whether or not there is a word A' generated by adding an affix to a word A in the document, checks whether or hot there is an equivalent B1', generated by adding an affix to an equivalent B to the word A, in equivalents B and B' to the word A' when the word A' is found, and decides words B1 and B1' as equivalents to the words A and A'. Consequently, the translation processing part 3 stores a translation obtained by translating the words A and A' in the document into the words B1 and B1', in an output buffer 7 and then displays it on a display 8.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a machine translation method for translating a sentence in a first language into a sentence in a second language and a machine translation device using this method.

[0002]

2. Description of the Related Art When translating a sentence in a first language (eg English) into a sentence in a second language (eg Japanese) using a conventional machine translation device, the word A in the sentence in the first language and this word When there is a word A ′ formed by adding an affix to A, when the word A and the word A ′ are independently translated into the words B and B ′ in the second language, B in the second language , B ′ does not necessarily inherit the relationship between A and A ′ in the first language.

For example, in the same sentence in the first language, accur
It is assumed that acy and inaccuracy are mixed, and that the translation dictionary "accuracy" stores the translated words "accuracy, accuracy, accuracy, precision" and the inaccuracy stores the translated words "inaccurate, errors".

[0004] Here, inaccuracy is accu
Since the word "racy" is added with the negative prefix "in", in a sentence in which these two words are mixed, it is likely that the correct answer is to be interpreted as an antonym of accuracy.

Nevertheless, in the conventional machine translation device, the accuracy and the inaccuracy were translated independently, so the accuracy was translated as "precision",
Inaccuracy may be translated as "error", and there is a possibility that the above-mentioned antonym may not be translated, and thus there is a problem that the translation accuracy is reduced.

Therefore, in order to improve the translation accuracy in the above case, if the translation is performed by using the co-occurrence information of each word, the word A and the word A'are translated independently and
Even when the words B and B'in the language are obtained, the relationship between B and B'in the second language is A in the first language due to the co-occurrence information described above.
It is possible to increase the possibility of translating as if the relationship between A and A'is inherited. However, in order to perform translation using such co-occurrence information, a large memory area is used to store the co-occurrence information, which causes a problem that the memory capacity becomes large.

[0007]

When a sentence in the first language is translated into a sentence in the second language without using cooccurrence information by using a conventional machine translation device, the word A in the sentence in the first language is used. And if there is a word A'which can be affixed to this word A, the words B and B in the second language obtained by translating the words A and A '
Also, although there are many cases where the correct answer is obtained when translating so that they have the same relationship with the first language words A and A ', the words A and A'
When the words B and B'in the second language become the words B and B'in the second language, the relationship between the words B and B'in the second language changes the relationship between the words A and A'in the first language. It is not always inherited, and there was a problem that the translation accuracy declined.

Therefore, the present invention has been made in view of the above circumstances. When a word and a pair of words with an affix to the word are in the same sentence, these words are accurately translated and the translation accuracy is improved. It is an object of the present invention to provide a machine translation method and a machine translation device using this method.

[0009]

The invention according to claim 1 is the first
The translation target sentence written in the language is referred to the second translation dictionary.
In a machine translation method for a machine translation device for translating a sentence in a language, whether or not there is a word A ′ that can be given to the word A in the sentence to be translated by adding an affix in the first language to this word A And if there is, a word B1 that can be obtained by adding an affix in the second language to the translated word B1 in the translated word B of the word A
It is checked whether ′ is in the translated word B ′ of the word A ′, and if there is, a method of translating the word A into the word B1 and the word A ′ into the word B1 ′ is adopted.

According to a second aspect of the present invention, an affix table in which affix information in the first language and affix information in the second language are collected as a list is provided, and a word formed by adding an affix in the first language to the word A Whether there is A'or the word B1 '
The method is used to check whether or not is in the translated word B ′ of the word A ′ by referring to the affix table.

According to a third aspect of the present invention, there is provided a dedicated user affix table for registering arbitrary affix information by a user, and there is a word A'composed of the word A and an affix in the first language. Whether or not the word B1 'is in the translation B'of the word A'is checked by referring to the affix table and the user affix table.

According to a fourth aspect of the present invention, there is provided a machine translation method in a machine translation device for translating a translation target sentence written in a first language into a sentence in a second language by referring to a translation dictionary. It is checked whether or not there is a word A ′ satisfying the similarity condition for being similar to the word A in the inside, and if there is, a similarity condition for being similar to each other in the translated word B of the word A and the translated word B ′ of the word A ′. It is examined whether or not there are translation words B1 and B1 ′ that satisfy the above condition. If there are translation words B1 and B1 ′, the word A is translated into the word B1 and the word A ′ is translated into the word B1 ′.

According to a fifth aspect of the present invention, the similarity condition between the words in the first and second languages is set such that the same character continuously exists between two words at a predetermined ratio or more. It is adopted.

According to a sixth aspect of the present invention, the similarity condition between the words in the first and second languages is set such that the same character continuously exists between two words in a predetermined number or more. It is adopted.

According to a seventh aspect of the present invention, in a machine translation device for translating a translation target sentence written in a first language into a sentence in a second language by referring to a translation dictionary, a translated word of the word in the first language is referred to. Translated word extraction means for extracting from a translation dictionary;
An affix storage unit that stores affix information of a language and affix information of the second language as a list, and word A in the sentence to be translated.
In response to the affix information in the affix storage means, the first retrieving means for retrieving whether the word A has an affix in the first language is added to the first retrieving means. When the word A ′ is searched by the search means of No. 1, the translated word B1 in the translated word B of the word A extracted by the translated word extracting means
It is searched by referring to the affix information in the affix storage means whether or not a word B1 'formed by adding an affix in the second language is in the translated word B'of the word A'extracted by the translation word extraction means. When the word B1 ′ is searched by the second searching means that performs the above, the word A is replaced by the word B.
1 has a configuration including a translation means for translating the word A ′ into the word B1 ′.

The invention according to claim 8 has a construction in which a user is provided with registration means for registering arbitrary affix information in the affix storage means. In the invention of claim 9, a dedicated user affix storage means for registering arbitrary affix information by the registration means is provided, and the first and second retrieval means search for the word A ′ and the translated word B ′. At this time, the affix information in the user affix storage means is referred to together with the affix storage means.

According to a tenth aspect of the present invention, in a machine translation device for translating a translation target sentence written in a first language into a sentence in a second language by referring to a translation dictionary, the translated word of the word in the first language is referred to as the translated word. A translated word extraction means for extracting from the translation dictionary, a first comparison means for determining whether or not there is a word A ′ satisfying a similarity condition for being similar to the word A in the translation target sentence, and the first comparison means. When it is determined that the word A'is present, the word A extracted by the translation word extraction means
Second comparing means for determining whether or not there are translation words B1 and B1 ′ satisfying similar conditions for the translation word B and the translation word B ′ of the word A ′. When it is determined that the word B1 'is present, the word A is translated into the word B1 and the word A'is translated into the word B1'.

According to an eleventh aspect of the present invention, the same character is a predetermined ratio between two words as the similarity condition that is used as a reference when the first and second comparing means compare the similarities between words. It has a configuration that adopts the condition that it continuously exists.

According to a twelfth aspect of the present invention, as the similarity condition, which is used as a reference when the first and second comparing means compare the similarities between words, a predetermined number of identical characters are provided between two words. It has a configuration that adopts the condition that it continuously exists.

[0020]

According to the machine translation method of the present invention, the word A in the sentence to be translated is first assigned to the word A.
It is checked whether or not there is a word A ′ that can be added with a language affix, and if there is, a word B1 ′ that can be added with a second language affix to the translated word B1 in the translated word B of the word A is the word A ′. It is checked whether or not it is included in the translated word B'of the above, and if there is, the word A is translated into the word B1 and the word A'is translated into the word B1 '.

In the machine translation method according to the second aspect of the present invention, an affix table that collects affix information in the first language and affix information in the second language as a list is provided, and the word A is used.
It is checked by referring to the affix table whether or not there is a word A'which can be given an affix in the first language, or whether the word B1 'is in the translation B'of the word A'.

In the machine translation method of the third aspect of the present invention, a user affix table dedicated for registering arbitrary affix information by the user is separately provided, and the word A is given an affix in the first language. Whether or not there is a word A ′ to be processed or whether or not the word B1 ′ is in the translation B ′ of the word A ′ is checked with reference to the affix table and the user affix table.

In the machine translation method of the invention of claim 4, the machine translation method in a machine translation device for translating a translation target sentence written in a first language into a sentence in a second language by referring to a translation dictionary. Then, it is checked whether or not there is a word A ′ that satisfies the similarity condition to be similar to the word A in the translation target sentence, and if there is, a translation B of the word A and a translation B ′ of the word A ′ are included. It is checked whether or not there are translation words B1 and B1 ′ that satisfy the similarity condition for being similar to each other. If there are translation words, the word A is translated into the word B1, and the word A ′ is translated into the word B1 ′.

In the machine translation method of the fifth aspect of the present invention, a similar condition is used in which the same characters are continuously present at a predetermined ratio or more between the words in the first and second languages. In the machine translation method according to the sixth aspect of the present invention, the similar condition is used in which the same character is continuously present in a predetermined number or more between the words in the first and second languages.

In the machine translation apparatus of the seventh aspect of the present invention, the translated word extraction means extracts the translated word of the word in the first language from the translation dictionary. The affix storage means stores affix information in the first language and affix information in the second language as a list. The first search means refers to the affix information in the affix storage means to determine whether or not there is a word A ′ formed by giving the affix of the first language to the word A in the sentence to be translated. And search. When the word A ′ is searched by the first searching means, the second searching means gives a suffix of the second language to the translated word B1 in the translated word B of the word A extracted by the translated word extracting means. Whether or not the word B1 'that can be created is in the translated word B'of the word A'extracted by the translated word extracting means is searched by referring to the affix information in the affix storage means. When the word B1 ′ is searched by the second searching means, the translation means translates the word A into the word B1 and the word A ′ into the word B1 ′.

In the machine translation apparatus of the eighth aspect, the registration means registers the affix information arbitrarily input by the user in the affix storage means. In the machine translation apparatus according to the invention of claim 9, the first and second search means register any affix information by the registration means together with the affix storage means when searching the word A ′ and the translated word B ′. The affix information stored in the user affix storage means is referred to.

In the machine translation apparatus of the tenth aspect of the present invention, the translated word extraction means extracts the translated word of the word in the first language from the translation dictionary. The first comparing means is a word A satisfying a similarity condition for being similar to the word A in the sentence to be translated.
It is determined whether there is ´. When the second comparing means determines that the word A ′ exists by the first comparing means, the translated word B of the word A extracted by the translated word extracting means.
Then, it is determined whether or not the translated words B1 and B1 'satisfy the similar conditions for being similar to each other in the translated word B'of the word A'. The translation means uses the word B1 by the second comparison means.
If it is determined that there is ', the word A is translated into the word B1 and the word A'is translated into the word B1'.

According to another aspect of the machine translation apparatus of the present invention, the first and second comparing means are based on a similar condition that the same character continuously exists between two words at a predetermined ratio or more. To determine the similarity between words.

In the machine translation apparatus of the twelfth aspect of the present invention, the first and second comparing means are based on the similar condition that the same character continuously exists between two words in a predetermined number or more. To determine the similarity between words.

[0030]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of a machine translation device of the present invention. Reference numeral 1 is an input device composed of a keyboard or the like for inputting characters and various function commands by keys, and has a translation conversion key. 2 is an input buffer that stores the sentence to be translated,
3 is a translation processing unit that translates the first language stored in the input buffer 2 into a second language using the translation dictionary 4, 4 is a translation dictionary accommodating a second language corresponding to the first language, 5
Is an affix expression processing unit that selects a translation candidate to be preferentially output by referring to the affix expression table 6 from the translation candidates of the second language obtained by translation, and 6 is affix expression information of the first language, An affix expression table containing affix expression information of the second language corresponding to the affix expression of the first language, 7 is an output buffer for storing the translation result of the translation processing unit 3, and 8 is a translation result in the output buffer 7. The display and 9 for displaying are an external storage device that stores the translation target sentence, the translated sentence, and the like in a file.

Next, the operation of this embodiment will be described. The first language shall be English and this shall be translated into the second language, Japanese. First, an English sentence is keyed in from the input device 1 and stored in the input buffer 2, or an English sentence to be translated which is stored in advance in the external storage device 9 is stored in the input buffer 2 (step 201).

After that, when the user presses the translation conversion key of the input device 1, the translation processing unit 3 starts the translation of the English sentence stored in the input buffer 2. That is, the translation processing unit 3 extracts the English sentence from the input buffer 2 sentence by sentence and converts the retrieved English sentence into a Japanese sentence by referring to the translation dictionary 4 as shown in FIG. 3 (step 202).

At this time, the translation processing unit 3 calls the affix expression processing unit 5 before confirming the translation of the one sentence (step 2).
03). In the affix expression processing unit 5, the word A ′ formed by adding the affix a to the word A constituting the English sentence extracted by the translation processing unit 3 is not only the English sentence but also the same English sentence to be translated in the input buffer 2. Whether or not it is found by referring to the affix expression table 6 as shown in FIG. 4 (step 20
4) It is determined whether or not the word A ′ in which the affix a is added to the word A is in the same English sentence (step 20).
5).

As a result, when the word having the affix a is added to the same English sentence, the affix expression processing section 5 translates the English words A and A'into B (collectively a plurality of translated words). , B ′ (collectively a plurality of translations) are extracted from the translation dictionary 4 (step 206), and in the translations of the English word A ′, the Japanese affix b corresponding to the English affix a is the translation B1. It is determined whether or not there is a translated word B1 ′ that is created by adding (step 207).

As a result, when there is a translated word B1 'in which the Japanese affix b corresponding to the English affix a is added to the translated word B1, the affix expression processing unit 5 selects one of the English words A and A'. , And the candidate priority of the translated words B1 and B1 ′ is set to the first priority (step 208).

The word A formed by adding the affix a to the word A
If ′ is not in the same English sentence, or after performing the processing of step 208, the translation processing unit 3 stores the translation result in the output buffer 7 (step 209). The display 8 displays the translation result in the output buffer 7 on the screen (step 210).

After that, the translation processing unit 3 determines whether or not the above-mentioned translation has been performed for all English sentences in the input buffer 2 (step 211). If not, step 20.
Returning to step 2, the translation process is continued, and if done, the process ends.

Next, a specific example of the above translation process will be described. When the affix expression processing unit 5 finds inaccuracy in the same English sentence in the input buffer 2 by referring to the affix expression table 6 shown in FIG. 4 for the accuracy in the English sentence extracted by the translation processing unit 3. , A
ccuracy translation "accuracy, precision, accuracy, precision"
And the translated word “inaccurate, incorrect” of inaccuracy is extracted from the translation dictionary 4.

After that, the affix expression processing unit 5 determines the accura.
It is found by referring to the affix expression table 6 shown in FIG. 4 that the translation "inexact" in which the Japanese affix "in" is added to the translation "accurate" of cy is found in the translation of inaccuracy. Accuracy and ina in the processing unit 3
Among the translated word candidates of "ccuracy", "correctness / inaccuracy" is set as the first priority. Therefore, in this example, a
ccuracy is translated as "accurate", and inaccuracy is translated as its antonym "inaccurate."

The translated text stored in the output buffer 7 can be stored in the external storage device 9 by operating the input device 1 to issue a storage instruction. According to the present embodiment,
When the word A'in the English sentence is translated into Japanese and the word A'formed by adding the affix a to the word A is present in the English sentence, the translated word B1 in the translated words B, B'of each word A, A ' If there is a translated word B1 'formed by adding the affix b, the words A, A'
Since the translated words can be B1 and B1 ', the translation of the words A and A'when the word A' with the affix a attached to the word A as described above is in the same sentence can be accurately performed. You can improve the translation accuracy.

FIG. 5 is a block diagram showing another embodiment of the present invention. In this example, in addition to the configuration shown in FIG. 1, a user affix expression table in which a user registers information such as an affix expression in the first language and an affix expression in a second language corresponding to the affix expression in the first language. Have ten.

The user affix expression table 10 has a structure as shown in FIG. 8, and the structure is the same as the affix expression table 6 shown in FIG. 4, but the accommodation information can be registered by the user. However, it is different from the affix expression table 6.

Next, the operation of this embodiment will be described with reference to the flowchart of FIG. Steps 601-603
The process up to is the same as the process up to steps 201 to 203 in FIG. After the processing of step 603, the affix expression processing unit 5 determines that the word A ′ having the affix a added to the word A forming the English sentence extracted by the translation processing unit 3 at step 604 is the input buffer as well as the English sentence. It is searched for by referring to the affix expression table 6 as shown in FIG. 4 and the user affix expression table 10 as shown in FIG.

That is, in this example, the word A ′ in which the affix a is added to the word A is searched for not only by referring to the affix expression table 6 but also the user affix expression table 10, which is different from the previous embodiment. Subsequent steps 605 to 611
Processing is the same as the processing of steps 205 to 211 of the previous embodiment.

Next, the process of registering information in the user affix expression table 10 by the user will be described with reference to the flowchart of FIG. When the user presses the user affix expression registration key of the input device 1, the affix expression processing unit 5 enters the registration mode to the user affix expression table 10 and the screen of the display 8 is in a state as shown in FIG. 9 (A). From Figure 9
The registration mode screen is displayed as shown in (B) (step 7).
01).

Next, when the user operates the input device 1 to input the affix expression information to be registered (step 702), the affix expression processing unit 5 registers the input affix expression information in the user affix expression table 10 ( Along with step 703), affix expression information is displayed in the registration screen of the display 8 as shown in FIG. 9B.

As a result of such registration processing, affix expression information is registered in the user affix expression table 10 as shown in FIG. The user is such a user affix expression table 1
The affix expression information to 0 is properly registered.

However, in the registration in the user affix expression table 10, the attribute information can be omitted, but the remaining affix expression information must be registered.

Next, a specific example of the above translation process will be described. The affix expression processing unit 5 uses calcul in the same sentence.
Even if ate is found, the affix expression table 6 shown in FIG.
If you refer to only the
unable to find the calculate, calcula
The translation of te is translated as "expectation".

However, in this example, the affix expression processing unit 5 is shown in FIG.
Since the user affix expression table 10 as shown in FIG. 4 can be referred to, this table can be used to calculate the calc in the same sentence.
It is possible to find the ‘ulate’ and the ‘recalculate’, and set the respective translated words as “calculation” and “recalculation” to cause the translation processing unit 3 to perform accurate translation.

According to this embodiment, the affix expression information required by the user can be registered in the user affix expression table 10.
In addition, the word A ′ to which the affix a is attached to the word A of the translation target sentence is searched for not only by referring to the affix expression table 6 but also the user affix expression table 10, so that the affix a is attached to the word A. The type of the existing word A ′ can be changed in accordance with the content of the sentence to be searched, and as a result, this type of word can be translated accurately.

Note that the user affix expression table 10 can be omitted as a configuration in which the user directly registers the affix expression information in the affix expression table 6. FIG. 10 is a block diagram showing still another embodiment of the present invention. In the configuration of this example, a similar word priority processing unit 11 is provided instead of the affix expression processing unit 5 and the affix expression table 6 of the configuration shown in FIG.

This similar word priority processing unit 11 determines whether or not there are words A and A'similar in spelling to the first language to be translated, and if they exist, also in the second language which is a translation of these words. , B, B ′, and the like, it is determined whether or not there is a translation word with similar spelling, and if there is a translation word, B, B ′ is applied as the translation word for A, A ′.

Next, the operation of this embodiment will be described with reference to the flow chart shown in FIG. Step 111
The processing up to 112 is the same as the processing up to steps 201 to 202 in FIG. The translation processing unit 3 calls the similar word priority processing unit 11 before confirming the translation of one sentence (step 113). After the processing of step 113, the similar word priority processing unit 11 has a word A ′ having a similar spelling to the word A that composes the English sentence extracted by the translation processing unit 3 in step 114.
Checks whether or not the above-mentioned English sentence is in the same English sentence as the translation target in the input buffer 2.

That is, in this example, a spelling word having a spelling almost the same as the spelling of the word A is regarded as a similar word A '. For example, if the spelling of both words is 60%, both words are regarded as similar. You Next, the similar word priority processing unit 11 determines in step 115 whether a word with similar spelling is found in the same sentence. However, that the spelling between words is the same means that the characters (alphabet in this case) forming both words are the same in succession.

As a result, when similar words A and A'are present in the same English sentence, the similar word priority processing unit 11 translates the English words A and A'into B (collectively a plurality of translation words) and B. ´
(Collectively a plurality of translated words) is extracted from the translation dictionary 4 (step 116), and it is determined whether or not translated words B1 and translated words B1 ′ having similar translated words are present among these translated words (step 117). .

As a result, when there are similar translated words B1 and B1 ', the similar word priority processing unit 11 selects the translated words B1 and B1' among the translated words of the English words A and A'in the translation processing unit 3.
Processing for setting the candidate priority order of 1st to 1st (step 1
18). However, the fact that the translated words are similar means that the characters that make up both translated words are consecutively the same.

On the other hand, when there are no similar words in the same sentence, or after the processing of step 118, the translation processing unit 3
Stores the translation result in the output buffer 7 (step 11
9). The display 8 displays the translation result in the output buffer 7 on the screen (step 120). After that, the translation processing unit 3 determines whether or not the above-described translation has been performed for all English sentences in the input buffer 2 (step 121). If not, the process returns to step 112 to continue the translation process, and if so, The process ends.

Next, a specific example of the above translation process will be described. If the similar word priority processing unit 11 finds a word inaccuracy similar to the accuracy in the English sentence extracted by the translation processing unit 3 (in this case, the spelling is 80% identical) in the same English sentence in the input buffer 2, acc
uracy's translation "accuracy, accuracy, precision, accuracy" and i
The translated word “inaccuracy, error” of naccuracy is extracted from the translation dictionary 4.

Thereafter, the similar word priority processing section 11 determines that the acc
The translation word "inaccurate" that is similar to the translation word "correct" of uracy (in this case, constituent characters are 67% identical in succession) is inacc
When it is found that it is in the translated word of uracy, the process of setting "correct or inaccurate" as the first priority among the translated word candidates of accuracy and inaccuracy in the translation processing unit 3 is performed. As a result, in this example, accuracy
Is translated as "accurate" and inaccuracy is translated as its antonym "inexact".

According to the present embodiment, similar words in the same sentence that is the object of translation are assigned similar translated words in the respective translated words, so that there is almost the same effect as in the above-described embodiment. . Further, since the affix expression table and the like are not required in this example, the structure of the device can be simplified and the device can be made inexpensive.

As a criterion (similarity condition) by which the similar word priority processing unit 11 judges the similarity of words, the number of the same characters may be used instead of the ratio of the same characters continuously contained between the words. . Further, the criterion (similarity condition) for determining the similarity of words may be different between the first language and the second language.

Further, in all the above-mentioned embodiments, the words A and A'are searched for in the translation target sentence in the input buffer 2. However, in the translation target sentence, the user searches the range for searching the words A and A '. It is also possible to set and perform detailed translation. The reason is that even in the case of the same sentence to be translated, there are cases where it is ineligible to translate the word A in the same way, because different things are discussed between the first chapter and the second chapter, for example. In such a case, it is better to limit the search range for words A and A'to each chapter.

Further, in all the above-mentioned embodiments, the affixes of the first language (English) stored in the affix expression table 6 and the user affix expression table 10 are prefixes, but if there is a similar relationship, the suffixes But of course it is okay.

[0065]

As described above, according to the inventions of claims 1, 2, and 7, when a word and a word pair formed by adding an affix to this word are in the same sentence, these words are accurately identified. The translation accuracy can be improved by translating into.

According to the inventions of claims 3, 18, and 19, the above-mentioned effect can be surely obtained according to the contents of the document to be translated. According to the inventions of claims 4, 5, 6, 10, 11, and 12, when a word and a pair of words with an affix to the word are in the same sentence, these words can be easily or simply structured. Can be translated accurately.

[Brief description of drawings]

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

FIG. 2 is a flowchart showing a translation processing procedure of the device shown in FIG.

FIG. 3 is a diagram showing an example of contents of the translation dictionary shown in FIG.

FIG. 4 is a diagram showing an example of contents of the affix expression table shown in FIG. 1.

FIG. 5 is a block diagram showing another embodiment of the present invention.

6 is a flowchart showing a translation processing procedure of the apparatus shown in FIG.

7 is a flowchart showing a registration processing procedure of affix expressions in the user affix expression table shown in FIG.

8 is a diagram showing an example of contents of a user affix expression table shown in FIG.

9 is a diagram showing an example of a screen of the display 8 shown in FIG.

FIG. 10 is a block diagram showing still another embodiment of the present invention.

11 is a flowchart showing a translation processing procedure of the apparatus shown in FIG.

[Explanation of symbols]

1 ... Input device 2 ... Input buffer 3 ... Translation processing unit 4 ... Translation dictionary 5 ... Affix expression processing unit 6 ... Affix expression table 7 ... Output buffer 8 ... Display 9 ... External storage device 10 ... User affix expression table 11 ... Similar Word priority processing unit

Claims (12)

[Claims] 1. A machine translation method in a machine translation device for translating a sentence to be translated written in a first language into a sentence in a second language by referring to a translation dictionary, wherein the word A in the sentence to be translated is used. On the other hand, it is checked whether or not there is a word A ′ that can be formed by adding the affix in the first language to this word A.
It is checked whether or not there is a word B1 ′ formed by adding an affix in the second language to the translated word B1 in the translated word B of the word A, in the translated word B ′ of the word A ′. B1
A machine translation method characterized by translating the word A ′ into a word B1 ′. 2. An affix table that collects affix information in the first language and affix information in the second language as a list is provided.
A word A'that can be obtained by adding an affix in the first language to the word A
2. The machine translation method according to claim 1, further comprising checking whether or not the word B1 'is included in the translation word B'of the word A'by referring to the affix table. 3. A user affix table dedicated for registering arbitrary affix information by a user is separately provided, and whether or not there is a word A ′ configured by adding an affix in the first language to the word A, or 3. The machine translation method according to claim 2, further comprising checking whether or not the word B1 'is included in the translated word B'of the word A'by referring to the affix table and the user affix table. 4. A machine translation method in a machine translation device for translating a sentence to be translated written in a first language into a sentence in a second language by referring to a translation dictionary, wherein a word A in the sentence to be translated is used. It is checked whether or not there is a word A ′ that satisfies the similarity condition for similarity, and if there is, a translation word B1 of the word A and a translation word B1 that satisfies the similarity condition for being similar to each other in the translation word B ′ of the word A ′. It is checked whether or not there is B1 ', and if there is, B is the word A and B is the word A'.
A machine translation method characterized by translating into 1 '. 5. The similarity condition between the words in the first and second languages is that the same characters continuously exist between two words at a predetermined ratio or more. Item 5
The described machine translation method. 6. The similarity condition between the words in the first and second languages is that the same character continuously exists between two words in a predetermined number or more. Item 5. The machine translation method according to item 5. 7. A machine translation device for translating a translation target sentence written in a first language into a sentence in a second language by referring to a translation dictionary, and extracting a translation of a word in the first language from the translation dictionary. A translated word extraction unit, an affix storage unit that stores the affix information of the first language and the affix information of the second language as a list, and the word A in the first language of the word A in the translation target sentence. A first search means for searching for the word A ′ formed by adding an affix by referring to the affix information in the affix storage means, and the word A ′ is searched by the first search means. And a word B1 ′ formed by adding an affix in the second language to the translated word B1 in the translated word B of the word A extracted by the translated word extracting means is a translated word B ′ of the word A ′ extracted by the translated word extracting means. In the affix storage means Second searching by referring to the dictionary information
And the word B1 by the second searching means.
A machine translation device comprising: a translation unit that translates the word A into a word B1 and the word A ′ into a word B1 ′ when 'is searched. 8. The machine translation device according to claim 7, further comprising registration means for registering arbitrary affix information in the affix storage means by a user. 9. A dedicated user affix storage means for registering arbitrary affix information by the registration means is provided, and the first and second search means are used for searching the word A ′ and the translated word B ′. 14. The machine translation device according to claim 13, wherein the affix information in the user affix storage means is referred to together with the affix storage means. 10. A machine translation device that translates a translation target sentence written in a first language into a sentence in a second language by referring to a translation dictionary, and extracts a translation of a word in the first language from the translation dictionary. The translated word extraction means, the first comparison means for judging whether or not there is a word A ′ satisfying the similarity condition for being similar to the word A in the translation target sentence, and the word A ′ by the first comparison means. If it is determined that there is a translation word B, the translation word B and the word A ′ of the word A extracted by the translation word extraction means.
Second comparison means for determining whether or not there are translation words B1 and B1 'satisfying the similar conditions for being similar to each other in the translation word B', and it is determined by the second comparison means that the word B1 'exists. Then, the machine translation device comprising: a translation unit that translates the word A into the word B1 and the word A ′ into the word B1 ′. 11. The similarity condition, which is used as a reference when the first and second comparing units compare similarities between words, has the same character continuously between two words at a predetermined ratio or more. The machine translation device according to claim 10, wherein the machine translation device is present. 12. The similarity condition, which is used as a reference when the first and second comparing units compare the similarities between words, has the same character between two words continuously in a predetermined number or more. The machine translation device according to claim 10, wherein the machine translation device is present.