JP2583898B2 - Post-editing machine for machine translation system - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a post-editing device for a machine translation system.

2. Description of the Related Art In general, a translated sentence after machine translation (post-editing in a broad sense) is usually performed by using an editing function provided in a translation device. In general, the method is performed in two stages, that is, a stage in which the correspondence between the original sentence and the translated sentence is maintained (post-editing in a narrow sense) and a stage in which correction is completely free (free edit). The former mainly focuses on the work of selecting the most appropriate one from a plurality of translation sentences and translation word candidates resulting from the fact that the interpretation of the original sentence is not uniquely determined. At this stage, since it is necessary to clearly indicate which translated sentence or translated word is generated from which original sentence, editing work that breaks the correspondence such as insertion or deletion of characters is prohibited. Also, it is clearly distinguished between the free editing mode in which the free editing operation is normally permitted and the subsequent editing mode, and the text once subjected to the free editing is configured so that the post editing can no longer be performed. It is configured so that post-editing can be performed after returning to the previous state.

Although it is inevitable that the post-editing function is restricted in maintaining the original-translation correspondence, it is too restrictive and inconvenient.

SUMMARY OF THE INVENTION The present invention has been made in view of the above circumstances, and has been developed after a machine translation system capable of expanding a range of editing operations allowable in post-editing while ensuring a correspondence between an original sentence and a translated sentence. The purpose is to obtain an editing device.

In order to achieve the above object, the present invention provides an input unit for inputting characters and the like, an original sentence storage unit for storing an original sentence, a translated sentence storage unit for storing a translated sentence obtained by translating the original sentence, and a phrase for identifying a phrase of the translated sentence. The identification unit and the phrase position of the translated sentence, the phrase length of the translated sentence and the phrase position of the corresponding original sentence using the phrase record including the phrase length and the position of the 1 to m words constituting the phrase of the original sentence corresponding to the translated sentence. Correlation relationship storage means for holding the correspondence relationship, a display screen, a cursor control means for moving a cursor on the display screen in units of phrases, and an original sentence for displaying the original sentence stored in the original sentence storage means on the display screen. Display means, translation display means for displaying the translation stored in the translation storage means on the display screen, and highlight table for highlighting an arbitrary word group in the original text displayed on the display screen Indicating means.

Hereinafter, an embodiment of the present invention will be described with reference to the drawings.
This embodiment is applied to, for example, a post-editing device for an English-Japanese machine translation system, and the configuration of this embodiment is conceptually shown in FIG. First, a keyboard 1 is provided as input means for inputting characters including symbols and the like. or,
A CRT 2 is provided which constitutes a display screen for displaying information about the original sentence, the translated sentence, and the like. Further, an original sentence storage means 3 for storing the original sentence and a translated sentence storage means 4 for storing a translated sentence corresponding to the original sentence, which is machine-translated, are provided. Here, the phrase storage means 4 is connected to a phrase identification means 5 for identifying a phrase of the stored translation.
Further, to the storage means 3 and 4, a correspondence storage means 6 for holding a correspondence between a phrase of a translated sentence and an original sentence word group corresponding to the phrase is connected.

Further, between the original text storage means 3 and the CRT 2, an original text display means 7 for displaying the original text on the CRT 2 is provided. Similarly, a CR is provided between the translation storage means 4 and CRT2.
Translation display means 8 for displaying a translation at T2 is provided, and between the correspondence storage means 6 and CRT2, highlighting means for highlighting an arbitrary word group in the original text displayed on CRT2. 9 are provided. Further, between the translation storage means 4 and the phrase identification means 5, there is provided a phrase content transformation means 10 for changing the content of the phrase of the stored translation. Further, there is provided a cursor control means 11 for operating the keyboard 1 to control the movement of the cursor on the CRT 2 on a phrase basis.

Here, the translation storage unit 4 and the correspondence storage unit 6
FIG. 2 shows a conceptual configuration of the phrase record in the above.
As shown in the figure, first, items of a sentence number, a clause number, a previous clause number, a next clause number, a head position, a clause length, and a cursor possession flag are set. The contents of the cursor possession flag are updated by the cursor control means 11, and the phrase identification means 5 is used for identifying the phrase where the cursor is located. In addition, a character code constituting a phrase, specifically, a main translation character string (first translation) is used. At this time, the phrase record has not only the main translation character string but also the translation information of the second translation word or less indicated by the second translation character string or the nth translation character string. Further, for the correspondence storage means 6, information relating to the original sentence word group, specifically, information relating to the original sentence word positions 1, 2 to m is also stored in the phrase record.

That is, in this embodiment, the processing unit at the time of document editing is not a character unit but a phrase unit. First, the translated word, Japanese, is represented as a set of phrase records. The phrase records include a phrase record connected to the phrase chain and a separated phrase record separated from the phrase chain. Accordingly, the phrase record is information indicating the position of the phrase, a character string of Japanese (translated word) constituting the phrase, and a group of English words (original text) corresponding to the phrase as shown in FIG. The information indicating the position in the English sentence is stored. Such a phrase record is output from the translation program and passed to the post-editing program of the post-editing device of the present invention.

Here, as shown in FIG. 3 corresponding to FIG. 2, all the phrase records in one sentence are connected to each other by a previous phrase number and a next phrase number to form one phrase chain. That is, FIG. 3 shows phrase records 12a, 12b, and 12c having phrase numbers 1, 2, 3,... In order from the left side, all having the same phrase number, and having the phrase record 12a at the beginning. If it is a record, the phrase record 12b indicates that the phrase record 12a is the preceding phrase and the phrase record 12c is the subsequent phrase.

Next, FIG. 4 shows a display screen (CRT) 2 for post-editing. The display screen is composed of a bilingual display screen divided into two right and left parts. Here, a plurality of sentences are shown in the drawing, but when post-editing is performed in units of one sentence, only one sentence is displayed in parallel. The cursor 13 for post-editing on such a display screen 2 is controlled by the cursor control means 11 so as to be positioned in the Japanese area as a translated word and to move the cursor in units of Japanese phrases. FIGS. 5 (a) to 5 (c) show the moving state of the cursor 13 in such a phrase unit. On the other hand, on the display screen 2, on the original sentence side, an original sentence word group corresponding to the translated phrase where the cursor 13 is located is highlighted by the highlighting control means 9, and the original sentence-translated sentence correspondence is clarified. FIG. 5 shows how the highlighted state corresponding to the position of the cursor 13 changes. Here, the emphasized display is indicated by the underline 14, but may be an emphasized display based on a difference in brightness of a display, a color display, or the like.

Further, on the display screen 2 for post-editing, a work area 15 for character input at the time of post-editing is secured on the lower side.
The contents of this area correspond to the contents of the character code area (translated character string) 16a in the separated phrase record 16 as shown in FIG.

Under such a configuration, in this embodiment, by operating the keyboard 1 while viewing the display screen 2 for post-editing,
With respect to the translated sentence, it is possible to delete, copy, move, insert or correct any phrase. Each of these editing processes is performed by removing a specified phrase record from a phrase chain or inserting a new phrase record into the phrase chain, as shown in the flowcharts of FIGS. 7 to 12, respectively.

As described above, according to the present embodiment, the post-editing work is performed in units of phrases, so that it is easy to always maintain the correspondence between the original sentence and the translated sentence even during the post-editing work. Free editing can be performed in post-editing without distinguishing between editing and editing. In other words, by introducing an editing function for each phrase, the number of editing operations that can be permitted in post-editing is increased, and restrictions are reduced, thereby making it easier for the operator to use. In particular, in this embodiment, the phrase number, the previous phrase number, the next phrase number, the head position,
The phrase record that has the phrase length, the cursor possession flag, and the character code that composes the phrase has information about the corresponding original word group, so you can delete, insert, move,
At the time of copying, it is easy to hold the correspondence between the phrase of the translated sentence and the word group of the original sentence. Further, since the phrase record includes not only the first translated word but also translated word information below the second translated word, the process of selecting the second translated word and below in post-editing is also simplified.

Effect The present invention is configured as described above, and performs post-editing on a phrase basis based on a phrase record including an original word position and the like, so that the correspondence between the original sentence and the translated sentence can be easily maintained. Since the phrase record also includes the requirement for the original sentence word position, it can be applied even when the word segments of the original sentence are not continuous, so that it can be freely edited in post-editing. The same degree of freedom as editing can be provided, and there is an effect that a post-editing function convenient for the operator can be obtained without being unnecessarily restricted.

[Brief description of the drawings]

1 is a block diagram, FIG. 2 is a conceptual diagram of a phrase record, FIG. 3 is a conceptual diagram of a phrase chain, FIG. 4 is a front view of a display screen, FIG. FIG. 5 is an explanatory diagram showing the movement of the cursor on the display screen, FIG. 6 is an explanatory diagram showing the correspondence between the display screen and the separated phrase record, and FIGS. 7 to 12 are flowcharts showing various processing routines. 1 ... keyboard (input means), 2 ... CRT (display screen), 3 ... original sentence storage means, 4 ... translated sentence storage means, 5 ...
... phrase identification means, 6 ... correspondence storage means, 7 ... original sentence display means, 8 ... translation sentence display means, 9 ... highlighting display means,
10 ... phrase content changing means, 11 ... Cursor control means

Claims (1)

(57) [Claims] An input means for inputting characters and the like, an original sentence storage means for storing an original sentence, a translated sentence storage means for storing a translated sentence obtained by translating the original sentence, a phrase identifying means for identifying a phrase of a translated sentence,
The correspondence between the phrase in the translated sentence and the corresponding phrase in the original sentence is determined by using a phrase record including the phrase position of the translated sentence, the length of the phrase, and the positions of l to m words forming the phrase in the original sentence corresponding to the translated sentence. Holding correspondence storing means, a display screen, cursor control means for moving a cursor on the display screen in units of phrases, original text display means for displaying the original text stored in the original text storage means on the display screen, A translation display means for displaying a translation stored in the translation storage means on the display screen, and highlighting means for highlighting an arbitrary word group in the original text displayed on the display screen. Post-editing device for machine translation system.