JPH04318671A - Machine translation system - Google Patents

Abstract

PURPOSE:To utilize a result of past by storing a result of analysis. CONSTITUTION:After the output 9 of translation 8, an input statement is stored 10. If a new input statement is the same as the previously input statement, the previously analysed result is output 5. If part of the newly input statement is the same as the previously stored input statement, that part is replaced with the previously analysed result 7. In translation 8, an input statement of which the whole or part is already translated is translated. Accordingly, by translating the input statement, linking can be specified, thereby being able to avoid analysing the same input statement more than once.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a translation system for a translation device which inputs a character string in a source language and outputs a character string in a target language.

0002

2. Description of the Related Art Hitherto, parentheses have been commonly used as a means of specifying dependencies in input sentences. For example, see the document "Hirai et al., Ambiguity detection method for Japanese dependency structure for pre-editing for machine translation, Information Processing Society of Japan Transactions v.
ol. 31, No. 10, Oct. 1990, P. 14
There is one shown in ``30 Figure 7''. In the same figure, "■If the interpretation of the system is incorrect, please use the pre-edit symbol [ ] to clearly indicate the destination."

For example, for the input sentence ``a girl with black eyes and long hair'', it is ``a girl with black eyes and long hair''.
By specifying a dependency using parentheses, as in the following example, modification between words separated by parentheses is suppressed. For example, the dependency candidate "black haired" is not adopted.

0004

[Problems to be Solved by the Invention] Conventional means for specifying dependencies use parentheses, and the left and right parentheses must always correspond, and a means to confirm this is required. Further, when an inexperienced user specifies multiple dependencies, it is difficult to enter the parentheses for specifying the intended dependencies, and mistakes are likely to occur.

[0005] This invention was made to solve the above-mentioned problems, and it is possible to specify dependencies by performing translation in advance, and for words or sentences that have been translated in advance, it is possible to specify dependencies in input sentences. Even if it appears multiple times,
The purpose is to obtain a machine translation method that can avoid performing analysis processing each time.

[0006]

[Means for Solving the Problems] The machine translation method according to the present invention includes a means for storing analysis results, and when the analysis results of a part or all of an input sentence have already been stored, the machine translation method according to the present invention The translation is carried out using the following methods.

[0007]

[Operation] When the input sentence 1 in the source language is translated and storage of the analysis result is specified, the analysis result 1 is stored in the storage means. When input sentence 2 is the same as input sentence 1 or includes input sentence 1, regarding the analysis of input sentence 1 in input sentence 2, analysis result 1 stored in the storage means is used as is,
Translate input sentence 2.

[0008]

【Example】

Example 1. FIG. 1 is a flowchart showing one embodiment of the present invention, and FIG. 2 is a flowchart showing part 10 in FIG. 1 in more detail.

Next, an explanation of the operation will be given with an example.

Translate the input sentence "I have long hair." After outputting, specify to store the input sentence and analysis results. The storage means stores the input sentence "I have long hair" and its analysis results below.

[0011]

[Math 1]

[0012] The input sentence ``black eyes'' is translated. After outputting, specify to store the input sentence and analysis results. The input sentence ``black eyes'' and the following analysis results are stored in the storage means.

[0013]

[Math 2]

[0014] Next, the input sentence "Girl with long black eyes and long hair"
Translate. In Figures 1 to 6, the analysis results for parts of the input sentences ``black eyes'' and ``long hair'' have already been stored, so in Figures 1 to 7, those parts are replaced with the previous analysis results as shown below. Translate as follows.

[0015]

[Math 3]

Unintended partial analysis such as ``black hair'' or ``long girl'' is not attempted.

[0017]

As described above, according to the present invention, dependency can be specified by translation, so there is no need to edit using parentheses. Furthermore, even if a stored input expression appears repeatedly in an input sentence, it is not necessary to perform repeated analysis processing each time.

[Brief explanation of the drawing]

FIG. 1 is a flow chart showing one embodiment of the present invention.

FIG. 2 is a flowchart showing in detail the section 10 in FIG. 1;

Claims (1)

[Claims] [Claim 1] A translation device for translating a text in a source language into a text in a target language, comprising means for storing analysis results and means for specifying storage of analysis results, wherein the analysis results are already stored in the input sentence. If it contains a phrase or sentence,
A machine translation method that detects words or sentences having the analysis results and translates them based on the analysis results stored in the storage means.