JPH0378874A - Mechanical translating device for natural language - Google Patents

Abstract

PURPOSE:To simplify manual pre-editing and post-editing operations by displaying the precision outputted from a conversion table to the rule of a knowledge base part applied to the translation. CONSTITUTION:An input sentence to be translated is inputted via an original sentence input part 10 and stored in a storage circuit part 11. An analyzing engine part 14 performs the morpheme analyses including the analysis of parts of speech, the meaning analysis, etc., via a dictionary 12 for the input sentence stored in the part 11. Then the sentence structure is analyzed based on the rule of a knowledge base part 13. Thus the input sentence is translated. Under such conditions, the label and the rank of the applied rule are outputted to a conversion table 16 and at the same time the results of analyses. etc., are outputted to a storage circuit part 15. The table 16 outputs the precision in response to the received label and rank. Then the part 15 stores the results of analyzes given from the part 14 and the precision given from the table 16. An output display part 17 displays the precision together with the input sentence and/or the analysis results. Thus the working efficiency is improved for the mechanical translation.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a natural language machine translation device that mechanically performs translation between a plurality of languages such as Japanese and English.

[Conventional technology]

Until now, machine translation devices based on several methods have been realized, but they have not reached the level where they can correctly analyze all syntactic patterns. The current situation is that nothing comparable to the accuracy of manual translation has yet been achieved, as it is necessary to strongly limit the field of translation.

For this reason, in machine translation at the current technological level, it is necessary to manually pre-edit the original text to make it suitable for machine translation, and to post-edit to correct translation errors after translation. ing.

[Problem to be solved by the invention]

However, in the above-mentioned Briedet, it is not easy to judge where in the original text should be modified.
In the worst case scenario, you will have to consider every part of the original text, and you will not be able to bear the hassle. Furthermore, in post-editing, the translation result and the original text must be successively compared and proofread, which requires an enormous amount of work.

As described above, although the machine translation device was originally intended to free humans from translation work, it actually requires a large amount of manual work and has the disadvantage that there are not many benefits to machine translation.

The present invention has been proposed in view of the above points, and its purpose is to provide a natural language machine translation device that can simplify manual pre-editing and post-editing.

[Means to solve the problem]

In order to achieve the above object, the present invention includes a dictionary that stores information regarding double amounts and a knowledge base section that stores rules necessary for analyzing language events, and the information of the dictionary and knowledge base section is stored. In a natural language machine translation device that uses a machine to translate an input sentence, a conversion table that outputs a degree of accuracy indicating the accuracy of the translation with respect to the rules of the knowledge base section applied during translation, and an input sentence and/or translation result. The apparatus also includes an output display section that displays the accuracy output from the conversion table.

[Effect]

In the natural language machine translation device of the present invention, the conversion table outputs the accuracy of the translation based on the rules of the knowledge base section that were used during translation, and outputs the accuracy of the input sentence and/or
Alternatively, the output display section displays the accuracy output from the conversion table together with the translation result.

〔Example〕

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing an embodiment of a natural language machine translation device of the present invention. In FIG. 1, this embodiment includes an original human resources section 10, a memory circuit section 11, a dictionary 12, a knowledge base section 13, an analysis engine section 14, and a memory circuit section 15.
, a conversion table 16 , and an output display section 17 . The functions of each part are as follows.

Input sentence (text) to be translated by Original Human Power Department 10i
A part for inputting information, such as a keyboard, an optical reader, etc.

Memory circuit section 11: This is a section that receives and stores input sentences from the original human resources section 10.

Dictionary 12: This is a part that stores information regarding vocabulary.

Knowledge base part 13: A part that stores rules necessary for analyzing language events, and as shown in FIG. 2, it has a condition part and an action part which is the process when the condition is satisfied , has a label that identifies the grammatical item that the rule is trying to analyze, and a rank that indicates the accuracy of the rule. The classification of labels and ranks is as shown in Figure 3, for example, label A is a rule group for analyzing parallel relationships, label B is a rule group for recognizing the object of the main phrase, Rank 1 has the highest accuracy.

For example, a condition such as whether or not there is an appositive pronoun across a parallel conjunction can determine the existence of a parallel relationship with high accuracy, so such a case will be ranked l. however,
Rank indicates the degree of accuracy within the same label, and rules with different labels cannot be directly compared.

Analysis engine section 14; dictionary 12 and knowledge base section 13
This is a part that performs morphological analysis and syntactic analysis, which are translation processes, using the information of This can be realized using the English analysis section of the machine translation device shown in ``.

Storage circuit section 15: This is a section that stores the analysis results and/or input sentences of the analysis engine section 14.

Conversion table 16: This table outputs the accuracy of a rule set in advance according to the label and rank of the rule of the knowledge base unit 13 applied for translation by the analysis engine unit 14. For example, as shown in FIG. 4, a degree of accuracy indicating the accuracy of translation according to the rule is set for each label and rank. That is, although the accuracy differs depending on the label even for rules with the same rank, the conversion table 16 allows the accuracy to be set for all rules using a uniform standard.

Output display unit 17: This is a unit that displays the accuracy output from the conversion table 16 together with the input sentence and/or translation result output from the storage circuit unit 15 on a display or print surface.

The operation of the above embodiment will be explained below.

First, an input sentence to be translated is input by the original human resources section 10 and stored in the storage circuit section 11 .

The double-folding engine section 14 performs morphological analysis such as part-of-speech analysis and meaning analysis on the input sentence stored in the memory circuit section 11 using the dictionary 12, and further uses the rules of the knowledge base section 13 to perform morphological analysis such as part-of-speech analysis and meaning analysis. Perform parsing and translation. On this occasion,
Conversion table 16 for label and rank of applied rule
At the same time, the analysis results and the like are output to the storage circuit section 15.

The conversion table 16 outputs the accuracy corresponding to the given label and rank, and the storage circuit section 15 outputs the accuracy corresponding to the given label and rank.
4 and the accuracy given from the conversion table 16 are stored, and the output display section 17 displays the accuracy together with the input sentence and/or the analysis result.

〔Effect of the invention〕

As explained above, in the natural language machine translation device of the present invention, accuracy indicating the accuracy of translation is displayed together with the input sentence and/or the translation result, so parts with high accuracy can be left to the machine. can be ignored as,
Since only the portions with low accuracy can be subject to editing or post-editing, the number of objects to be examined is greatly reduced, which has the effect of greatly increasing work efficiency.

[Brief explanation of drawings]

Figure 1 is a block diagram showing an embodiment of the natural language machine translation device of the present invention, Figure 2 is a diagram showing an example of the contents of the knowledge base section in Figure 1, and Figure 3 is the label of Figure 2. and FIG. 4 is a diagram showing an example of the contents of the conversion table in FIG. 1. In the figure, 10...Original human resources section 11...Storage circuit section 12...Dictionary 13...Knowledge base section 14...Analysis engine Section 15: Memory circuit section 16: Conversion table 17: Output display section FIG. 2 shows an example of the contents of the knowledge base section

Claims (1)

[Scope of Claims] It has a dictionary that stores information regarding vocabulary and a knowledge base section that stores rules necessary for analyzing linguistic events, and uses information in the dictionary and knowledge base section to translate input sentences. In a natural language machine translation device that performs translation, a conversion table that outputs accuracy indicating the accuracy of translation with respect to the rules of the knowledge base section applied during translation, and output from the conversion table together with the input sentence and/or translation result. A natural language machine translation device comprising: an output display unit that displays the accuracy of the translation.