JPH0844748A - Machine translation system - Google Patents

Abstract

PURPOSE:To reduce operation and the efficiency of equivalent specifying operation by providing a cursor-associative dictionary retrieval part which retrieves a dictionary for the word at the position in an original language to which a cursor moves. CONSTITUTION:An original sentence is inputted through a character input part, 1 and held in an original sentence holding part 3. The original sentence is displayed at a sentence display part 4. When the cursor moves on the original sentence as indicated from a cursor movement part 2, the cursor-associative dictionary retrieval part 10 actuates a morpheme analysis part 5 as to the part where the cursor moves. The morpheme analysis part 5 divides the part where the cursor moves into morphemes by using information in the dictionary part 6 to obtain equivalent candidates for the respective words. The morpheme division results of the original sentence are held as part of original sentence information in an original sentence holding part 3. The pairs of the respective words and equivalent candidates are held in an original word equivalent holding part 7. An equivalent selection part 8 displays the equivalent candidates for the respective words, which are held in the original word equivalent holding part 7, at the sentence display part 4.

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.

[0002]

2. Description of the Related Art A conventional machine translation system also has a function of designating a translated word for each word of an original sentence.
The methods were roughly classified into two methods. One is a method of changing the translated word in the post-editing stage after once translating the original sentence. The other is a method of designating a translated word for each word in the pre-editing stage before translating the original sentence.

The former method can be called a post-edited translated word changing method. The typical operation of this method is as follows.

In this method, since the translation has already been executed, the word recognition in the original sentence and the parallel translation dictionary search for each of these words have already been completed. If there is a target word to be corrected in the translated text, the user returns to the original word in the original text leading to the target word and specifies it. The system displays to the user a list of translations that are different from the translations that are being output for the specified source language. The user selects an appropriate translation from the translation candidates. The system sends the original word and the designated translation to the translation core, and the translation core executes the translation again according to the instruction.

The latter method can be called a pre-edited translated word designation method. The typical operation of this method is as follows.

[0006] In this method, since the translation has not been executed yet, in the case of a language such as Japanese in which a word break is not clearly indicated, word recognition in the original text is still performed. No bilingual dictionary search has been performed for each word in the original text. In this case, the range of words in the original text needs to be explicitly indicated to the system by the user himself. Specifically, the word is identified by designating the start point and the end point of the word. Since the operation of designating the start and end points of a word is the same as a general range designating operation, the user must also explicitly instruct the system to perform a dictionary search for words in that range. The system displays the translation for the word specified by the user. The user selects an appropriate word from the translation candidates. The system sends the original word and the designated translated word to the translation core, and the translation core executes the translation again according to the instruction.

[0007]

The two conventional methods for specifying a translated word have poor operation efficiency in the following points.

The first post-edited translated word changing method has a drawback in that the operation of associating the original word with the translated word resulting from the translation is complicated. There are actually methods to display the original sentence and its translated text side by side, such as horizontally arranging the corresponding sentences, vertically arranging the original sentence one sentence at a time, and vertically arranging the translated sentence in order. In the case of a language pair in which the word order in a sentence is significantly different, such as English, it is not immediately apparent to the user which word in the source sentence corresponds to which word in the translated sentence. Furthermore, it is difficult to deal with the situation where the translated word which the user does not intend is output and the translated word needs to be changed. Therefore, the conventional system has a function that, when a word in the original sentence or the translated sentence is designated with a mouse cursor or the like, the corresponding word in the translated sentence or the original sentence is highlighted so that the user can easily display it. It was However, in such an operation, an explicit operation for associating the word in the original sentence with the translated word in the translated sentence is required as a pre-step of specifying the translated word to be originally performed, and accordingly, the procedure is complicated and inefficient. .

In the second pre-edited translated word designating method, the user must designate the beginning and end of a word, and after the range designating operation, the dictionary search must be explicitly designated. There is The operation efficiency is poor in that the number of operations that the user has to do is large.

An object of the present invention is to provide a method for reducing the number of operations that the user has to do in designating a translated word as much as possible and improving the efficiency of the translated word designating operation.

[0011]

The present invention holds a morphological analysis unit for morphologically analyzing an input original sentence, a dictionary unit used for morphological analysis and translation of the original sentence, and each word in the original sentence and a translated word for them. A machine translation having an original word translated word holding unit, a translated word selection unit that selects one translated word from translation candidate for each word, and a machine translation core unit that translates using a designated translated word for each word In the system, the machine translation system is characterized by having a cursor-linked dictionary search unit that searches a dictionary for a word existing at a position where the cursor has moved in the source language.

[0012]

In the present invention, the translation word designation by the user is performed in the following procedure.

First, when the user moves the cursor in the direction from the front to the back of the original sentence, the system automatically activates morphological analysis for the portion where the cursor has moved to perform word recognition, and the portion is moved. A bilingual dictionary search is performed for each word. This function is a cursor-linked dictionary search function. Then, the parallel translation display means displays the parallel translation of the search result near the original word. The user selects one appropriate translated word from the side-by-side translations displayed by the translated word selection means. Then, the original word translated word holding unit sends the original word and the translated word pair designated for the original word to the core of translation.
The translation core part translates the original sentence using the designated translation word.

Since the movement of the cursor in the above procedure has the role of instructing the display of parallel translations for each word,
It should be distinguished from the usual different cursor movements. In order to make this distinction clear to the system and to realize an operation method that is less separated from the usual cursor movement for the user, the cursor movement with this parallel translation display instruction function is
For example, it corresponds to the operation of pressing the normal cursor movement key while pressing the control key.

In the above operation procedure, the morphological analysis may be executed in advance before the cursor is moved.

[0016]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing one embodiment of the present invention. In FIG. 1, 1 is a character input unit such as a keyboard, 2 is a cursor movement instruction unit such as a keyboard or a mouse, 3 is an original sentence holding unit that holds information of an input sentence, 4 is a held sentence and its accompanying A sentence display unit such as a display device for displaying information to the user, 5 a morphological analysis unit for morphological analysis of the input original sentence, 6 a dictionary unit used for morphological analysis and translation of the original sentence, and 7 for each in the original sentence An original word translated word holding unit that holds a translated word obtained as a result of a dictionary search for a word, 8 is a translated word selection unit that presents a translated word candidate for each word to the user, and the user selects one translated word from them , 9 is a machine translation core having a translation designated translation function for translating each word using a translation designated. These functions are the functions usually provided in the conventional machine translation system.

The feature of the present invention is the cursor interlocking dictionary search unit 10 in FIG. 1, and this cursor interlocking dictionary search unit 10 is for a word existing at the position where the cursor moves in the source language. It has the function of searching the dictionary.

Next, the operation of the present invention will be described using a specific example.

The method for specifying a translated word in the present invention can be applied to a machine translation system for any language, but a case of a Japanese-English translation system for translating Japanese into English will be described below as an example.

An original sentence is input through the character input unit 1 and held in the original sentence holding unit 3. The original sentence is displayed on the sentence display unit 4. In that situation, when the cursor moves over the original sentence according to an instruction from the cursor moving unit 2, the cursor-linked dictionary search unit 10 activates the morphological analysis unit 5 for the portion where the cursor moves. The morpheme analysis unit 5 uses the information of the dictionary unit 6 to divide the portion where the cursor has moved into morphemes, and obtains a translation word candidate for each word. The movement of the cursor in this case has a different meaning from the usual simple cursor position change,
Since the purpose of the operation is to obtain translation candidates for the word at the position where the cursor has moved, the specific operation for moving the cursor in this case must be different from the normal operation.
For example, it can be realized by an operation of pressing a normal cursor movement key while pressing a special key such as a control key.

The result of morpheme division of the original sentence is held in the original sentence holding unit 3 as a part of the original sentence information. A pair of each word and a translation word candidate for the word is stored in the original word translation storage unit 7. The translated word selection unit 8 displays translated word candidates for each word held in the original word translated word holding unit 7 on the sentence display unit 4. On this occasion,
The translation candidate is displayed at a position adjacent to the original word.

In FIG. 2, after the cursor movement is instructed by the cursor moving unit 2, the translated word of the word included in the cursor moved portion is displayed on the screen, and the result of the translation of the original sentence using the translated word is displayed. It is the figure which showed the state until it used the example. There are several possible ways of displaying translated word candidates for each word. In FIG. 2, the highest-priority translated word candidate is displayed by replacing it with the original word, and the other candidates are displayed as a pull-down menu at a position adjacent to the original word.

In FIG. 2, each word in the original sentence is displayed by being replaced with the highest-priority translated word. At this stage, it is still not clear whether the displayed translated word is an appropriate translated word in the case of this original sentence. Since it is not guaranteed, the user selects an appropriate candidate from the translation candidates. Then, the selected one becomes a new top priority candidate and replaces the original word. In Fig. 2 (c), "the" vs. "the"
Although “United States” is displayed as the highest priority candidate, when the user selects “the US”, it is replaced and displayed as shown in FIG. In addition, in FIG. 2D, "col" is translated as "col".
2 is displayed as the highest priority candidate, but when the user selects “university”, FIG.
It is replaced and displayed as in (e).

The translated word pair designated by the user for each word of the original sentence is held in the original word translated word holding unit 7. The pair is transmitted to the machine translation core 9. 2 (e) shows a state immediately before data is sent to the machine translation core 9, and FIG.
(F) shows the translation result returned from the machine translation core 9. Since the machine translation core unit 9 has a function of translating the original sentence using the designated translation word, the translation result using the translation word as designated by the user is output.

Regarding the method of displaying the translation word candidates for each word, as shown in FIG. 3A, translation word candidates other than the highest priority translation word candidate need not be displayed unless there is an instruction to display them. Further, as shown in FIG. 3B, all the translated word candidates may be displayed at a position adjacent to the original word without replacing the original word with the translated word.

In the above description, the case where the cursor interlocking dictionary search unit 10 activates the morphological analysis at the position where the cursor is moved by the movement of the cursor has been described. May be certified. In that case, since the break of the word of the original sentence is already held in the original sentence holding unit 3, when the cursor moving unit 2 instructs the movement of the cursor,
The cursor-linked dictionary search unit 10 directly searches for a word dictionary. The operation thereafter is exactly the same as the above case.

Further, although the above is the case where the original sentence is input from the character input unit 1 such as the keyboard, the original sentence may be stored in another file in advance. FIG.
2 is a diagram in which an original text file reading unit 11 is added to FIG. It is also possible to import the original text file created in advance into the original text holding section 3 using the original text file reading section 11, and thereafter specify the translated word for each word of the original text by the cursor operation as described above. In this method, since the process of creating the original sentence and the process of specifying the translated word for each word for translating the sentence are separated, the thinking during the work is not confused and the work efficiency is high.

[0029]

According to the present invention, unlike the "post-editing translated word changing method" described in the section "Problems to be solved by the invention", it is not necessary to translate once. Compared with this method, the procedure to specify a translated word and translate it is reduced.

Further, unlike the "pretranslation word designation method" described in the section "Problems to be solved by the invention", the user does not need to explicitly indicate the start and end points of a word. Compared to this method, the procedure for specifying the translated word is reduced. Moreover, since it is not necessary to explicitly instruct the dictionary search, the procedure is further simplified.

According to the present invention, since the translated word of the word at the position is automatically displayed by moving the cursor,
You can perform translation word selection, which is the task you originally want to do, without interrupting your thinking.

[Brief description of drawings]

FIG. 1 is a block diagram showing one embodiment of the present invention.

FIG. 2 is a diagram showing an example of a state in which a translated word of a word included in a cursor moving portion is displayed on a screen after a cursor movement is instructed and a sentence is further translated.

FIG. 3 is a diagram showing an example of variations of a display method of translation word candidates for each word.

FIG. 4 is a diagram showing another embodiment of the present invention and is a block diagram including an original text file reading means for explaining a method of reading an original text from an original text file and designating a translated word.

[Explanation of symbols]

 1 Character input section 2 Cursor movement instruction section 3 Original sentence storage section 4 Sentence display section 5 Morphological analysis section 6 Dictionary section 7 Original word translated word storage section 8 Translated word selection section 9 Machine translation core section 10 Cursor-linked dictionary search section 11 Original text file reading section

Claims (1)

[Claims] 1. A morphological analysis unit for morphologically analyzing an input original sentence, a dictionary unit used for morphological analysis and translation of the original sentence, an original word translated word holding unit for holding each word in the original sentence and a translated word for each word, In a machine translation system that has a translation word selection unit that selects one translation word from translation word candidates for a word and a machine translation core unit that translates using a translation word specified for each word, a cursor in the source language A machine translation system characterized by having a cursor-linked dictionary search unit that searches a dictionary for a word existing at a position to which is moved.