JP3134400B2 - Machine translation system - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a machine translation system which analyzes a sentence in a first language and translates it into a sentence in a second language based on the sentence. Articles that cannot be uniquely determined from the language
Also, the present invention relates to a machine translation system that can translate efficiently.

[0002]

2. Description of the Related Art In recent years, a machine translation system for automatically translating a first language into a second language by using a computer has been receiving attention.
In addition, it can be performed efficiently and is considered to become increasingly important in the future.

For example, when a Japanese sentence is translated into an English sentence by such a machine translation system, the Japanese sentence has no information and cannot be uniquely determined from the sentence. However, there are some items that are indispensable in composing English sentences. For example, this is the article. The articles include "a", "the", and "no article", but in order to create an accurate English sentence, it is necessary to use different ones in consideration of various conditions.

In such a case, the conventional machine translation system generates articles according to the set conditions when conditions are set in the system itself, and when the conditions are not set in the system itself, the operator himself / herself prepares the articles in advance. Generated articles set in the system.

[0005] For example, when translating Japanese from "students study" into English, if no condition is set in the system itself, the article (for example, "the") set by the operator in the system Based on "The stu
ent studios. "Was translated uniformly.

[0006]

However, there is no problem because the article itself set in the system by the operator as described above is set based on the intention of the operator. Articles that are generated based on the conditions that are included include the type of sentence to be translated,
It may not always be appropriate based on the degree of the operator's knowledge of English as a language.

In such a case, the operator generally tries to give priority to the article set by the operator, but in the above-mentioned conventional translation system, the system operates the article generated according to the conditions set for the article. Therefore, the article set by the operator has priority over the article set by the operator, and the article set by the operator cannot be generated. For this reason, there is a problem that the operator needs to make corrections at the time of editing after translation is completed.

SUMMARY OF THE INVENTION The present invention has been made to solve this problem, and considers the translation conditions for articles set by the operator in the machine translation system in advance by taking account of the article translation conditions set in the system itself. And the degree of reflection in the translation result sentence can be selected.
It is another object of the present invention to provide a machine translation system that can efficiently translate text into a second language.

[0009]

In order to achieve this object, a machine translation system according to the present invention comprises an input unit for inputting a sentence in a first language, and a first language sentence input from the input unit. In a machine translation system having a translation unit for translating into two-language sentences and a display unit for displaying a second-language sentence translated by the translation unit, the machine translation system is used for translation processing by the translation unit, and A translation dictionary that stores article information determined for each word constituting the translated text of the second language or corresponding to a set of two or more words, and a translation dictionary of the second language translated through the translation unit. An article setting means for setting a desired article in advance in a word constituting a sentence, a first translation condition using article information stored in the translation dictionary, and a desired article set via the article setting means The second translation condition using Translation condition setting means for setting, a display control means for displaying the first and second translation conditions set by the translation condition setting means on the display unit, and a first translation condition from each translation condition displayed by the display unit. A selection unit for selecting a second translation condition; and an article generation unit for generating an article for a second language sentence translated via the translation unit based on the translation condition selected by the selection unit. It is characterized by having. Here, it is desirable that the article information stored in the translation dictionary include article information that is conventionally determined according to a syntax or the like.

[0010]

In the machine translation system according to the present invention having the above-described structure, first, a sentence in the first language is input by the input unit, and the sentence in the first language is displayed on the display unit. Next, when translating the first language sentence into the second language sentence via the translator, the translation is stored in the translation dictionary and determined corresponding to each word constituting the second language sentence, or two or more. The first translation condition using article information conventionally determined according to a syntax or the like which is a set of words and the second translation condition using a desired article set via the article setting means are translated condition setting means. Is set by After the first and second translation conditions are displayed on the display unit, the first or second translation condition is selected from the respective translation conditions via the selection unit. Thereafter, an article is generated for the text in the second language by the article generating means based on the selected translation condition.

[0011]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below in detail with reference to the drawings. Here, for convenience of understanding and explanation of the invention, the first language, which is the language to be translated, is Japanese, and the second language, which is the translation language, is English.

FIG. 1 is a block diagram schematically showing the configuration of a machine translation system according to the present embodiment. The system includes an input unit 1 composed of an input device such as a keyboard, a display unit 2 composed of a display device such as a CRT, and the like. It comprises a printing unit 3 composed of a printing device such as a scanning printer and a data processing unit 4 for performing data processing for translating Japanese into English.

The data processing section 4 further comprises an editing section 5, a translation section 6, and a dictionary section 7, as shown in FIG. The editing unit 5 is for pre-editing a sentence suitable for translating Japanese into English, and for performing post-editing for correcting the translated English sentence to appropriate English. As shown in FIG. A translation condition setting unit 5A for setting a plurality of translation conditions described below for articles that cannot be uniquely translated from the word information into English, a display control unit 5B for displaying the set translation conditions on the display unit 2, a display unit 2 A translation condition selection unit 5C for selecting, as a system, a translation condition selected via the input unit 1 from among the translation conditions displayed in the translation unit 6, and a translation unit 6 and a dictionary unit 7 based on the translation conditions selected in this manner. And a translation control unit 5D for executing translation via the control unit.

A translator 6 executes a translation process from Japanese to English, and morpheme analyzer 8 that divides Japanese into words, and dictionary information is added to each word divided based on an analysis dictionary 13 described later. , A syntax analysis unit 9 that analyzes words and syntax from Japanese into English based on a conversion dictionary 14 described below, and a syntax conversion based on a generation dictionary 15 described below. Syntax generation unit 11 for generating correct English syntax from English converted by unit 10
Generation unit 12 that generates a translation result sentence in English and English
It is composed of

Further, the dictionary unit 7 stores an analysis dictionary 13 storing information necessary for analyzing a Japanese structure, and information necessary for converting Japanese words and syntax into English. A conversion dictionary 14 and a generation dictionary 15 storing information necessary for generating an English syntax from the English converted in this way are provided.

Next, the operation of the machine translation system of this embodiment having the above configuration will be described. The Japanese sentence input from the input unit 1 is pre-edited by the editing unit 5 into a form suitable for translation into English, and then converted into morphological analysis information stored in the analysis dictionary 13 by the morphological analysis unit 8. Is divided into words based on After that, the syntactic analysis unit 9 adds dictionary information to each word divided based on the syntactic analysis information in the analysis dictionary 13 and analyzes the Japanese syntax.

Further, in the syntax conversion unit 10, the Japanese words and syntax are converted into English based on the conversion information stored in the conversion dictionary 14. As a result, the syntax generation unit 11 generates the correct English syntax based on the dictionary information stored in the generation dictionary 15 in the words and syntax converted into English as described above.

Here, taking an article as an example of an English item that cannot be uniquely determined from Japanese information, control for generating such an article will be described with reference to FIG.

In FIG. 3, the operator corrects a Japanese sentence input from the input unit 1 to the data processing unit 4 into Japanese suitable for translation into English by the editing unit 5, and first sets an article in S1. I do. The reason for setting the articles in advance in this way is to enable the generation of articles according to the operator's intention even for words that do not have article information in the conversion dictionary 14 and the generation dictionary 15 of the data processing unit 4. is there. Here, it is assumed that the article “a” has been set by the operator.

Next, in S2, it is determined how much the article set by the operator in S1 is reflected in the translation result sentence. At this time, an option indicating the degree of reflection is displayed on the display unit 2 for use by the operator.

Here, the article is translated into the syntax generator 11 of the translator 6.
, The following three generation criteria can be considered.

(1) First, for example, as shown in FIG. 4, article information stored in the generation dictionary 15 (for example, article information in which the definite article "the" is added immediately before the adjective "same") When generating an article, or as shown in FIG. 5, article information stored in the conversion dictionary 14 (for example, when bilingual English is used in the meaning of a Japanese heading, the article used is always determined. In some cases, an article may be generated based on the determined article information. In these cases, the generated articles are all highly accurate.

(2) Next, article information for generating an article based on the conventional usage of English stored in the generation dictionary 15 (for example, immediately after a preposition such as "... in the box", or "... the cake w
This is a case where an article is generated based on article information that a definite article "the" is often added immediately before the antecedent of a relative clause, such as "hich I mode". in this case,
Due to exceptions, the accuracy of the generated article is lower than in (1).

(3) Finally, since there is no such article information on the article, the article is generated based on the article information preset by the operator (for example, the article "a" set in S1). This is the case. In this case, the accuracy of the article generated as a whole depends on (2)
Lower than the case.

In consideration of the above three article generation criteria, the following options can be considered as options to be displayed on the display unit 2 in S2. A. Articles set by the operator are used as all articles. B. Articles generated based on article information stored in the conversion dictionary 14 or the generation dictionary 15 preferentially use them,
For other articles, the articles set by the operator are used. C. Articles generated based on article information stored in the conversion dictionary 14 or the generation dictionary 15 and article information according to the conventional usage of English stored in the generation dictionary 15 use the same, and other articles are used. For, the article set by the operator is used.

Option A is selected when the operator's English ability is extremely high or when translating special Japanese. Option B is used when the operator's English ability is relatively high or when the translation result sentence is The option C is selected when accuracy is required up to a relatively small portion, and the option C is selected when the operator's English ability is low or the translation result sentence does not require much accuracy.

The above three options are displayed on the display unit 2 in S2. When the operator selects option A in S2, the article "a" set by the operator in S3 is reflected in all articles, and "a" is generated for all articles from this.

If the operator selects the option B in S2, the process proceeds to S4. In S4, articles other than articles generated based on article information in the conversion dictionary 14 and the generation dictionary 15 are set. The article is reflected, and the article "a" set in S1 is generated for such an article.

Further, when the operator selects the option C in S2, the process proceeds to S5. In S5, the article set by the operator is converted to the article information of the conversion dictionary 14, the generation dictionary 15, and the English common words in the generation dictionary. The article other than the article generated based on the article information according to the usage is reflected, and for such an article, the article "a" set in S1 is generated.

Thereafter, after the translation result sentence is generated through the morpheme generation unit 12 and the translation is completed, the article generated in S3, S4, and S5 is checked in post-editing in S6, and other articles are checked in S7. Editing is performed after checking items.

Here, when the article is checked, the display unit 2 has a function of sequentially searching for only the article set by the operator and checking it while blinking.

As described in detail above, the machine translation system of the present embodiment can enable the operator to select the extent to which a preset article is reflected in the translation result sentence. As a result, Japanese can be accurately and efficiently translated into English, the work of checking articles during post-editing can be reduced, and a translated sentence according to the operator's English ability can be created.

Also, when checking the articles, the display unit 2 can check the articles set by the operator while sequentially blinking the articles, so that the articles can be checked very efficiently.

Therefore, for example, when the operator's English ability is high or when the translation result sentence is required to have a high accuracy, the option B can be selected by selecting the option B.
4. The article “a” set by the operator in S1 for all articles other than the article translated based on the article information in the generation dictionary 15
Is generated, the operator can sequentially search only the article on the display unit 2 at the time of post-editing to check the article while blinking it, and correct the article if it is incorrect. Therefore,
Articles can be checked very efficiently.

If the operator's English ability is low or if the translation result sentence does not need to be accurate to the finest part, selecting the above-mentioned option C allows the translation dictionary 14 and the article dictionary of the generation dictionary 15 to be selected. In addition to the information, only the articles set by the operator other than the articles generated based on the article information according to the English common usage of the generation dictionary 15 may be sequentially searched and checked while blinking.

[0036]

As described above, according to the present invention, the translation conditions related to articles set by the operator in the machine translation system are reflected in the translation result sentence in consideration of the article translation conditions set in the system itself. The degree of selection is selectable, so that sentences in the first language can be accurately and efficiently translated into sentences in the second language, the work of checking articles during post-editing can be reduced, and translation according to the language skills of the operator Can create sentences.

[Brief description of the drawings]

FIG. 1 is a block diagram schematically showing a configuration of a machine translation system.

FIG. 2 is a block diagram schematically showing a configuration of an editing unit.

FIG. 3 is a flowchart of an article generation control program.

FIG. 4 is an explanatory diagram schematically showing the contents of a generation dictionary.

FIG. 5 is an explanatory diagram schematically showing the contents of a conversion dictionary.

[Explanation of symbols]

 1 input unit 2 display unit 4 data processing unit 5 editing unit 5A translation condition setting unit 5B display control unit 5C translation condition selection unit 5D translation control unit 6 translation unit 7 dictionary unit 13 analysis dictionary 14 conversion dictionary 15 generation dictionary

Continuation of the front page (56) References JP-A-4-167067 (JP, A) JP-A-63-37472 (JP, A) JP-A-63-223962 (JP, A) Akira Kumano, Hiroyasu Nogami, Yumiko Yoshimura , Shinya Amano, "User Collaborative Japanese Generation System", Proc. Of the 42nd IPSJ Annual Conference (Early 1991), p. 3-35 to p. 3-36 (1991. 3) (58) Field surveyed (Int. Cl. ⁷ , DB name) G06F 17/20-17/28

Claims (2)

(57) [Claims] An input unit for inputting a sentence in a first language, a translation unit for translating a sentence in a first language input from the input unit into a sentence in a second language, and a second unit translated by the translation unit
A machine translation system having a display unit for displaying a sentence in a language, wherein each of the words constituting a second language sentence, which is used for translation processing by the translation unit and is translated via the translation unit, or two or more words A translation dictionary that stores article information determined corresponding to the set of : an article setting unit that previously sets a desired article to a word that constitutes a sentence of the second language translated through the translation unit; Translation condition setting means for setting a first translation condition using article information stored in a translation dictionary, and a second translation condition using a desired article set via the article setting means; Display control means for displaying the first and second translation conditions set by the setting means on the display unit; and selection means for selecting the first or second translation condition from the respective translation conditions displayed by the display unit The said Machine translation system characterized in that a article producing means for producing an article against sentence of the second language to be translated through the translation section based on the selected translation condition by-option means. 2. The machine translation system according to claim 1, wherein the article information stored in the translation dictionary includes article information that is conventionally determined according to a syntax or the like.