JPH03282965A - Machine translating system - Google Patents

Abstract

PURPOSE:To reduce translation work by utilizing past translated results and replacing only a part whose contents are different from an original by correctly translated contents. CONSTITUTION:A translation control part 9a controls a translation part 5, etc., and translates the 1st language to the 2nd language. A sentence comparing part 9b compares the data of the 1st language inputted from an input part with the data of the 1st language recorded in the 1st language recording part 7a of a recording part 7. At the time of deciding that only a part of phrases appearing in the sentence is different between the inputted 1st language data and the recorded 1st language data as the result of comparison based upon the comparing part 9b an attribute comparing part 9 compares the attributes of respectively different phrases with each other. At the time of deciding that the attributes of the phrases concerned are equal by the comparison of the attributes, a phrase replacing part 9d replaces the translated data of the different part in the 2nd language data by proper translated data and then outputs a correct translated result.

Description

DETAILED DESCRIPTION OF THE INVENTION [Objective of the Invention (Field of Industrial Application) The present invention relates to a machine translation system that efficiently translates data in a first language into data in a second language.

(Prior Art) In recent years, machine translation systems that use computers to automatically translate data in a first language into data in a second language have been attracting attention.

In this machine translation system, after translating an input document in a first language into a second language, the original text in the first language and the translation result in the second language are associated and recorded sentence by sentence. It is now possible to modify the translation results while maintaining the state of this association.

Furthermore, current machine translation systems are often used as translation support devices. That is, a human first converts the original text into a simple sentence, and then inputs it into a machine translation system.The resulting translation result is further modified by a human and used as the final translation result.

On the other hand, when translating a large number of documents, these documents often contain sentences that differ only in some words and phrases and are otherwise identical. However, in conventional machine translation, each sentence is translated independently, so if the already translated sentence and the sentence to be translated differ only in some words that appear in the sentence. , there was no relationship between the two, and therefore previous translation results could not be used. Therefore, even when translating a text that has already been translated and a text that differs only in some of the words that appear in the text, in order to obtain the final translation result, machine translation is performed completely independently. The same corrections as those made must be made again, which is extremely troublesome.

(Problems to be Solved by the Invention) As mentioned above, in conventional machine translation systems, only some of the words that appear in the sentence are translated when the translation has been completed and the translation result has been obtained. When translating a different sentence anew, it is not possible to use past translation results even in cases where the translation result can be obtained simply by changing the translation of the different word or phrase. Therefore, when creating the translated text, the same corrections made to the translation result by machine translation must be made again in the new translation, resulting in a loss of time and effort. There was a problem.

Therefore, in a machine translation system, when a document to be newly translated contains a sentence that differs from the already translated sentence in only some of the words in the sentence, the translation result of the sentence is By using the translation result of a text that has already been translated, with the different translations replaced, it is possible to avoid making the same corrections as those made to the text that have already been translated. The purpose is to provide a machine translation system that makes it possible to reduce translation work.

[Configuration of the Invention (Means for Solving the Problem) In order to achieve the above object, the present invention provides a translation means for translating data in a first language into data in a second language;
a recording means for recording data in a first language and a data in a second language in association with each other; an input means for inputting the data in the language; and a data in the first language input from the input means. , a comparison means for comparing the difference with the first language data recorded in the recording section, and a comparison means for determining the difference in a part of the two first data and the attribute of the different part. When it is determined that they are equal,
This corresponds to the recorded first language data. The present invention is characterized by comprising a control means for replacing the different part of the data in the second language with a result of translating the different part of the input data in the first language into the second language.

(Operation) In the machine translation system of the present invention, when performing translation,
Compare the document to be translated with a document that has already been translated, and if there is a sentence in which only some of the words in the sentence are different, and the attributes of the different words are the same, the differences in the already translated results By replacing the translated words of a part, it is possible to obtain the final translation result without making the same corrections as the previous corrections, which reduces the translation work and is effective when updating documents. can.

(Example) Hereinafter, one embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing a schematic configuration of an English-Japanese machine translation system to which the present invention is applied. In FIG. 1, an input unit 1 is used to input various data such as a keyboard, a mouse, or a voice input device.

The display section 3 may be any device capable of displaying characters, for example, a CRT display, CCD, etc., as appropriate.

The translation unit 5 refers to a dictionary 5a that stores rules related to translation and translations and attributes related to words and phrases, and translates English, which is a first language, into Japanese, which is a second language, with reference to the dictionary 5a. The search unit 5C includes a translation processing unit 5b, which searches the dictionary 5a for translations and attributes related to a phrase during translation.

The recording unit 7 independently records data in English, which is the first language, and data in Japanese, which is the second language.
English data and Japanese data as a result of its translation are recorded in correspondence with each other.

The control section 9 controls the entire machine translation system, and includes a translation control section 9a, a sentence comparison section 9b, an attribute comparison section 9c, and a word/phrase replacement section.

The translation control unit 5 9a is configured in the same manner as the conventional one, and controls the translation unit 5 and the like to translate the first language into the second language. At this time, the sentence comparison section 9b compares the first language data input from the input section 1 with the first language data recorded in the first language recording section 7a of the recording section 7.

Further, the attribute comparison unit 9C determines that the input first language data and the recorded first language data are different only in a part of the words that appear in the text through comparison by the text comparison unit 9b. Compare the attributes of different words when If it is determined by this attribute comparison that the attributes of the words are the same, the word/phrase replacement unit 9d replaces the translated word data of the different part in the recorded second language data and outputs it as a translation result. .

Next, the operation of the control section 9 will be explained using the flowchart shown in FIG.

The control unit 9 starts its operation in response to a translation start request from the input unit 1. The control section 9 that has started its operation investigates whether already translated data is recorded in the recording section 7 via the translation control section 9a (step Sl). Unless already translated data is recorded in the recording unit 7, translation using past translation results cannot be performed, so step S3
Proceed to step 1 to translate.

data in the first language (hereinafter, this data will be simply referred to as data A)
) is input and the entire data is transferred to the translation unit 5 and translated into the second language (step 833). The translation result obtained in step 333 is recorded in the second language recording section 7b of the recording section 7 in correspondence with the first language data (step 529).

On the other hand, if translated data has already been recorded in the recording unit 7 in step S1, the process advances to step S3 and inquires of the operator about the use of previously translated and recorded translation results in the current translation.

When the operator decides not to use this translation result, the process proceeds to step S31, in which data in the first language to be translated is manually generated in the same way as in conventional machine translation systems, and the entire data is transferred to the translation unit 5 and translated into the second language. (Step 833) The translation result obtained here is stored in the second language recording section 7 of the recording section 7 in correspondence with the first language data.
b (step 529).

When the operator decides to use the previous translation result in step S3, the operator advances to step S5 and inputs conditions regarding the extent to which the input data and recorded data are to be used. Examples of conditions to be input here include the following.

(1) To what extent should words in a sentence be usable even if they are different?

(2) Which attributes should be made available when they are equal?

After inputting the conditions, data in the first language to be translated, ie, data A, is input in step S7.

Next, in step S9, first language data (hereinafter, this data will simply be referred to as data B) is input from the recording unit 7 as translation unit data for use in translation processing. Select the first sentence of the data A (step 5ll),
It is checked whether data (data D) that differs from this first sentence in only some words in the sentence exists in data B (step 813). In addition to using the judgment criteria given in advance to the device, the usage conditions of the translation result specified in step S5 may be used as the judgment criteria at this time.

If there is no data that differs from the first sentence in data A in only part of the words, that is, data D, the process advances to step S35, where the first sentence of data A is transferred to the translation unit 5, and this sentence is translated. Get results. If data D exists in step 813, words and phrases in the first sentence that are different from data D are
Then, the word/phrase in the data D that is different from the first sentence is set as F (step 517). Step S19
Then, this E is transferred to the translation section 5, and the search section 5c of the translation section 5 searches the dictionary 5a for the attribute and translated word of this E.

Then, the attribute comparison unit 9C compares the attribute of E and the attribute of F, and checks whether the condition specified in step S5 is met or not (step 521). If the condition specified in step S5 is not met, the process proceeds to step S35, where the first sentence is transferred to the translation processing section 5b of the translation section 5, and a translation result of this sentence is obtained.

If the condition specified in step S5 is met in step S21, the process advances to step 323 and the translation result of data D is read from the recording unit 7.

Subsequently, in step S25, the word/phrase replacement unit 9d replaces the original translated word in the translation result with the translated word of E, and converts the data A
Let this be the translation result of the first sentence of

This completes the translation of the first sentence.

Thereafter, the above steps (steps 313, 525) are similarly repeated until there are no more sentences to translate (step S27.
537).

When the translation of the entire data A is completed in this way, the process advances to step S29, and the data A and its translation results are transferred to the recording section 7.
The process ends.

Next, translation processing in the apparatus of this embodiment will be explained using example sentences.

In step S5, "1/3 or more of the words in the sentences match" and "different words have the same parts of speech" are input as usage conditions for the previous translation results.

Next, in step S7, rHe is used as the sentence to be translated.
enjoyed his mother's
Enter confidence, J. Subsequently, steps 89 to Sll are performed, and in step S13, r
He enjoyed former 5friend's
Suppose that confidence, J is obtained. This 2
The different words in the two sentences are r mother's
J and r friend'sJ, both of which are nouns, match the condition of step S21.

Therefore, rHe enjoyed his fr
iend 5conf 1dnce, read J's translation result "He was trusted by his friends" from the recording section,
The translation of r f'riend'sJ in this is "friend".
By replacing the translation word "mother" with "motherJ", the input sentence "rHe enjoyed his
The translation result of mother seonf'1dence, J is obtained, and the operator does not need to operate during this time.

In this way, in the machine translation system of this embodiment, since the previous translation results are used by replacing the translated words of the different words in the original text, the results of the previous translations of similar sentences are modified. Therefore, it is possible to avoid making the same corrections as those previously made, and it is possible to reduce the effort and time required for translation.

Note that the present invention is not limited to the embodiments described above. For example, in the above-mentioned embodiment, the condition for determining the degree of difference in words that can be used is specified as the ``proportion of the same word in the sentence'', but this can be changed by specifying it for each part of speech, or `` "The central verbs are equal", "The final words of the sentences are equal"
There are also methods such as ``extracting and comparing keywords'' and ``comparing syntax analysis results.''

It is also possible to combine these.

Furthermore, regarding the conditions regarding attributes for use, in the above example, the condition was that "the parts of speech are the same", but there are also conditions such as semantic features of words such as "person" and "food", cases such as possessive case, etc. Conditions other than the part of speech may be used. Furthermore, in addition to equality, relationships such as inclusion based on superior/lower concepts may also be used.

Furthermore, the above relationship can be applied not only to the word dodoma but also to phrases and clauses.

Furthermore, in the above embodiment, the data to be compared with the input data is a single piece of data, but it is also possible to use it to compare a plurality of pieces of data. Furthermore, the translation results used for comparison are not necessarily limited to those translated by the same device.

Furthermore, in the above embodiment, information is not given to the operator when the translated words have been replaced, but information on which sentence was used is added to the sentence for which the translated word has been replaced, and the search can be performed using this information. It is also possible to display the replacement location to the operator by highlighting the translated word or phrase.

[Effects of the Invention] As described above, according to the present invention, in a machine translation system, by using the translated words of the different parts of the original text among the past translation results, Editing work can be reduced, and efficient translation work such as shortening of translation results can be realized.

[Brief explanation of drawings]

FIG. 1 is an overall block diagram of an embodiment of the present invention, and FIG. 2 is a flowchart showing an example of the processing flow of a translation control section used in an embodiment of the present invention. 1... Input section 3... Display section 5... Translation section 7... Recording section 9... Translation control section

Claims (1)

[Scope of Claims] Translation means for translating data in a first language into data in a second language; recording means for recording data in the first language and data in the second language in association with each other; an input means for inputting language data; a comparison means for comparing the first language data input from the input means with the first language data recorded in the recording section; , When it is determined by this comparing means that there is a difference in a part of the two first data and that the attributes of this different part are the same, the second language corresponding to the recorded first language data is A machine translation system comprising: a control means for replacing the different part of the input data in a first language with a result of translating the different part of the input data in a first language into a second language.