JP3011273B2 - Translation result management method and machine translation system - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a translation result management method and a machine translation system.

2. Description of the Related Art Recently, a machine translation system has been developed which can automatically obtain a translation by simply inputting an original text. Here, in such a system, the translation solution by the translation unit can be displayed and edited by a translation editor.

Problems to be Solved by the Invention However, the conventional translation editor has a main function of displaying a translation result, and does not pay much attention to its editing. In particular, when modifying the translated sentence, the output solution itself is corrected, and therefore cannot be returned to the original translated sentence itself. Also, when correcting a solution whose translation result is immature and unusable, it may be more convenient to completely replace the translated document.However, the conventional method is based on the original translation, so I can't take it. That is, according to the conventional method, it can be said that there is practically no user solution.

Means for Solving the Problems In a machine translation system having a plurality of buffers each holding a plurality of translation solutions based on one original sentence, a plurality of translation solutions based on one original sentence output from the machine translation system are provided. Is held in each of the plurality of buffers, any one of the plurality of held translation solutions is displayed, and the displayed translation solution is provided in a buffer provided separately from the plurality of buffers. Then, on this buffer, a user solution as an arbitrary editing result by the user is transferred and copied to a buffer capable of creating and holding the user solution, and the edited user solution is displayed.

Further, the machine translation system has a plurality of buffers for respectively holding a plurality of translation solutions based on one original sentence, and a buffer provided separately from the plurality of buffers, and the plurality of buffers are held in the plurality of buffers. Displaying any one of the plurality of translation solutions based on the one original sentence, transferring the displayed translation solution to the separately provided buffer, and arbitrarily setting the translated translation solution by the user. The user solution can be created and retained on the separately provided buffer, and the edited user solution is displayed.

Since the user solution is managed as a separate translation from the translation solution, it is possible to freely create and edit the user solution while maintaining the original form of the translation solution, and to return the translation solution to its original form. It's easy.

Embodiment An embodiment of the present invention will be described with reference to the drawings. First,
The concept or procedure of the machine translation system used in the present invention will be described with reference to the algorithm of FIG. 3 using a general syntax transfer method. This is, for example, an English-Japanese machine translation system. After inputting an original sentence, a morphological analysis for the spell check is selectively performed. And
Perform general morphological analysis. This is an analysis in units of words, and is an analysis process that focuses on, for example, special characters such as hyphens. Next, syntax analysis is performed. This is a syntactic analysis to know the structure of a sentence. After performing an English structure identification process such as eliminating semantically incorrect analysis from the result of the syntax analysis to create an English intermediate structure, a syntax conversion process is performed. That is, the intermediate structure is converted from an intermediate structure based on English expressions into an intermediate structure that can easily express Japanese expressions.
Then, Japanese language generation is performed. That is, in order to generate a natural Japanese sentence from the intermediate structure of Japanese, first, syntax generation is performed, and then morpheme generation for adjusting the arrangement order of particles and auxiliary verbs is performed. Thereafter, a translated sentence, for example, Japanese is finally output.

Here, a morphological analysis is performed on the input original text, and the result is fed back to the user for pre-editing, or a translation result is displayed bilingually to be edited after editing to the user's desired one. -Has a post-edit function.

The translation result management method according to the present embodiment relates to the processing in the post-editing unit (translation editor) for displaying and editing the translation result output from the translating unit. It does not depend on any translation method.

The block diagram is constructed as shown in FIG. That is, an input unit 1 for inputting commands, texts and the like, a translation unit 2 for executing a translation process on the input original text, a post-editing unit 3 for displaying and editing translation results and the like, an original text correction, etc. An editing unit 4 for editing the
It comprises a memory 6 having a plurality of buffers 5 for holding solutions, and a control unit 7 for controlling the whole.

In such a configuration, a translation result management method according to the present embodiment will be described. First, the original text input from the input unit 1 and edited by the editing unit 4 as necessary is added to the translation unit 2.
Outputs N translation solutions, the sequel middle part 3 prepares (N + 1) solution holding buffers 5 in the memory 6. Then, after the N translation solutions are individually held in the buffers 5 from 1 to N, the (N + 1) th buffer 5 holds the first translation solution, and this is temporarily set as the first solution. Is displayed. Correction and editing of the following solutions are performed only on the (N + 1) -th buffer 5, and the originals of the translation solutions are individually retained in the first to N-th buffers 5. Then, according to a command from the user, an arbitrary M-th solution such as the second, third or the like is calculated as (N +
1) Transfer and copy to the 5th buffer 5, arbitrarily edit the solution, and create a user solution. Conversely, this (N
The solution in the (+1) -th buffer 5 is copied to the first to N-th original positions and stored. That is,
In the present embodiment, due to the presence of the (N + 1) th buffer 5, the user solution is managed as a separate translation from the translation solution by the 1st to Nth buffers 5, and is managed equally. As a result, it is possible to freely edit the translated sentence as desired by the user, and to maintain the original form of the machine translated sentence.

As a modified example, when N translation solutions are generated, N buffers 5 are prepared without preparing the (N + 1) -th buffer from the beginning, and each of the N buffers 5 is individually translated. Keep the solution. And according to the user's request,
Further, an empty buffer may be prepared, and this empty buffer may be used in the same manner as the (N + 1) -th buffer to freely edit the empty buffer. At this time, if the already-existing alternative solution is displayed, free translation (creation of a user solution) that can sufficiently refer to the result of the machine translation becomes possible. At this time, if the empty buffer is provided with a function capable of copying any of the N solutions, it is more effective to improve the workability. The number of such empty buffers need not be limited to one, but may be additionally registered as the (N + 1) -th solution, or may be copied to buffers 1 to N. That is, all of the plurality of user solutions created for one original sentence may be managed as separate translated sentences. Then, as the final solution, any one of the translation solution from the translation unit and the user solution may be examined.

In addition, the system of the embodiment and the system of the modified example are not completely independent, and various systems can be considered as an intermediate system between them.

 Now, this embodiment will be described with a specific example. For example, disagree with the original text “Two of a trade seldom agree.”

 2 Mostly disagree with trade.

 Two in the industry do not agree.

When three translated sentences are obtained, the translated sentences are respectively stored in the first to third buffers 5, and 4 (=
The first solution is stored in the (N + 1) th buffer. By showing the fourth buffer 5 to the user, the user can freely edit the actual translation solution while keeping it as it was.

In any case, according to the user solution management method of the present invention, it is possible to keep the translation solution before editing and correction as it is, and it is easy to return to the state before correction. . Further, it is possible to easily create a completely independent user solution of a user who has left the translation solution.

As described above, according to the present invention, in a machine translation system having a plurality of buffers each holding a plurality of translation solutions based on one source sentence, a plurality of buffers based on one source sentence output from the machine translation system are provided. Are respectively stored in the plurality of buffers, any one of the plurality of stored translation solutions is displayed, and the displayed translation solution is provided separately from the plurality of buffers. A buffer in which a user solution, which is an arbitrary editing result by the user, can be created and retained on the buffer, and copied and displayed, and the edited user solution is displayed;
Further, the machine translation system has a plurality of buffers for respectively holding a plurality of translation solutions based on one original sentence, and a buffer provided separately from the plurality of buffers, and the plurality of buffers are held in the plurality of buffers. Displaying any one of the plurality of translation solutions based on the one original sentence, transferring the displayed translation solution to the separately provided buffer, and arbitrarily setting the translated translation solution by the user. The user solution can be created and retained on the buffer provided separately as described above, and the edited user solution is displayed, so that the original form of the translation solution is retained and the translation solution is returned to the translation solution. The user solution can be freely edited and edited while the user's solution is made easy, and the user can create the best translation by reworking the translation once created without discarding it. One in which it is possible to go up.

[Brief description of the drawings]

1 is a flow chart, FIG. 2 is a block diagram, and FIG. 3 is a flow chart showing a syntax transfer method.

 ────────────────────────────────────────────────── ─── Continuation of the front page (56) References JP-A-62-171055 (JP, A) JP-A-61-150068 (JP, A)

Claims (2)

(57) [Claims] 1. A machine translation system having a plurality of buffers each holding a plurality of translation solutions based on one original sentence, wherein the plurality of translation solutions based on one original sentence output from the machine translation system are stored in the plurality of translation solutions. A plurality of buffers, each one of the held plurality of translation solutions is displayed, and the displayed translation solution is provided separately from the plurality of buffers, A translation result management method, wherein a user solution, which is an arbitrary editing result by a user, is transferred and copied to a buffer capable of creating and holding the edited user solution, and the edited user solution is displayed. 2. A machine translation system, comprising: a plurality of buffers each holding a plurality of translation solutions based on one original sentence; and a buffer provided separately from the plurality of buffers. One of a plurality of translation solutions based on one original sentence held in the buffer is displayed, and the displayed translation solution is transferred and copied to the separately provided buffer. A machine translation system wherein a user solution, which is an arbitrary editing result by a user, can be created and held on the separately provided buffer, and the edited user solution is displayed.