JPH04105171A - Machine translation system - Google Patents

Abstract

PURPOSE:To make the meaning of an original sentence easy to understand by not giving a translated word to an original word indicated by a user and making it outputted as it is the original word. CONSTITUTION:The system consists of an input means 1, a sentence storage means 2, a translating means 3, a dictionary 4, a display means 5, an original sentence output instruction means 6, an original sentence output information storage means 7 and an original sentence output rule part 8. Then, a translated word is not given to an original word indicated by a user, it is outputted as it is the original word and besides, also a basic word with comparatively high frequency and with wide meaning is not rather translated and is outputted as it is in an original sentence. Thus, the meaning of the original sentence can be transferred precisely.

Description

[Detailed description of the invention] division of technical rights The present invention relates to a machine translation system.

A single word has multiple meanings, and when translated, it may result in multiple different translations. For this reason, one translation word is determined by giving priority to the translation words. However, when the translated word is chosen incorrectly from the original meaning, it significantly hinders the reader's understanding of the original text.

For example, humans can determine whether the meaning of a word such as "bank" is "bank" or "bank" based on information from the context of the sentence in which it is used. However, current machine translation technology cannot correctly select translated words. Therefore, for the word specified by the user, dictionary information for that word (part of speech, semantic features, co-occurrence information, etc.) is used, but the translated word is displayed as is, making it easier to understand the original text. Even if the user knew that the word "bank" had the meanings of both "bank" and "bank," the user would still use the word "bank" in the translation as "He is standing with his back to the bank." It is difficult to notice that J is ``Bank J.However, the meaning of ``Bank'' can be easily noticed from rbankJ.

In the example of LL asll, Na asll also has meanings such as "reason,""situation," and "concession." The translation results for each meaning are as follows.

Reason = “Some students study because they listen to music.” Situation: “Some students study while they listen to music.” “Some students study because they listen to music.” students study.

” Concession: “They listen to music, but how many students study?” It is impossible to correctly analyze and translate these meanings, and if you strangely give the wrong translation, you will end up referring to the original text entirely. I will do it. If so, they listen to music, some students study. ”
If translated as ``Hasll'', a user who can speak a little English can easily understand the meaning of the frequently occurring word ``Hasll'' from the context of the sentence. In particular, when it comes to technical terms in a specialized field, it is often easier for experts in that field to understand the translated text if the terminology appears in the original text.

The present invention has been made in view of the above-mentioned circumstances, and
For the original word specified by the user, the original word is used without any translation to make it easier to understand the meaning of the original text, and it is also possible to use basic words that are relatively frequent and have a wide range of meanings. Rather, the purpose was to provide a machine translation system that accurately conveys the meaning of the original text by outputting the original text without translation.

In order to achieve the above object, the present invention includes an input means for inputting a target language, a storage means for storing the target language inputted from the input means, and a storage means for storing the target language inputted by the input means. In a machine translation device consisting of a translation unit that performs translation, a dictionary unit necessary for the translation by the translation unit, and a display unit that outputs the translation result translated by the translation unit, An original text output instruction means for instructing to output the original text as it is, an original text high-power information means for storing information instructed by the original text output instruction means, and a word that is left unchanged from the original text does not appear in an unnatural form in the translation result. It has an original text output rule section for outputting the original text as it is, and when it is translated, it is translated by the dictionary section and the translation section, and the specified word is translated by referring to the original text output rule section. The feature is that the original text is output as a translated text.

Hereinafter, the present invention will be explained based on examples.

FIG. 1 is a block diagram for explaining one embodiment of a machine translation system according to the present invention, in which 1 is an input means, 2 is a text storage means, 3 is a translation means, 4 is a dictionary, 5 is a display means, 6 7 is an original text output instruction means, 7 is an original text output information storage means, and 8 is an original text output rule section. In this embodiment, a case of English-Japanese translation will be explained.

As the original text from the input means 1, as shown in (1) in FIG.
The sentence "t the bank," is input.The operator inputs the word #b bank in the original text from the original text output instruction means 6.
When u is specified, the original text output information storage means 7 outputs the word "b".
It is stored that "ank" has been instructed. When the operator instructs to start translation, the translation process shown in FIG. 3, which will be described later, begins.

The translation section 3 refers to the dictionary section 4 and uses the information to proceed. In morphological analysis, words are extracted from the input original text and information about the words is obtained from a dictionary. The translation process is performed as shown in FIG. 3 as follows. In syntactic analysis, the syntax of the original sentence is analyzed, the part of speech is determined at the same time, and the structure of the English sentence is generated. In an embodiment of the invention, II
bankI+ is interpreted as a noun. In the conversion,
Convert to Japanese structure. In generation, a translated word is selected, Japanese is used, and the Japanese word order is determined. Furthermore, in the generation unit, when selecting a translated word, it is determined from the content stored by the original text output information storage means 7 that the word "bank" is to be output as is in the original text.

The translation of 11bank++ is rb a n kl. It does not matter whether the translation code is 1-byte or 2-byte as long as the character types are the same. In the Japanese language conjugation process of the generation unit, when referring to the original text output rule unit 8, there is a rule that when outputting a noun in the original text, it is ``Mashiyoyo'', so rbankJ is not used and is left as is. Since there are no rules for determining word order, the translation result when the original text is output is the * mark (1) in Figure 4, ie, ``He is standing with his back to the bank.''

FIG. 2 is a diagram showing the processing flow of the machine translation system according to the present invention. Below, each step will be explained in order.

Dear Mori: First, input the original text in the target language using the input means.

5tep2; Next, it is determined whether there is an instruction by the operator to output the original text.

If there is an output instruction in step 2 above.

Remembers that the original text was instructed to be output.

Company profile: Next, it is determined whether there is a translation instruction from the operator.

If there is a translation instruction, start the translation process.

Finally, output the translated text.

FIG. 3 is a block diagram for the translation process of step 5 in FIG. 1
5 is a generation unit (conjugation), 16 is a generation unit (word order), 17 is a dictionary, 18 is an original text output rule, and 19 is an original text output information storage unit. The morphological analysis unit 11 uses a dictionary to cut out words from the input original text. In the syntax analysis unit 12,
After obtaining each piece of information, the syntax of the original sentence is analyzed, the part of speech of the word is determined, and a tree structure of the English sentence is created. The conversion unit 13 converts the tree structure of the input original text into the tree structure of the target language, that is, the structure of Japanese. The generation units 15 and 16 generate the original text output rules 18 as shown in FIG. 5, and the original text output information storage unit 1.
9 to select the translated words, have them use the Japanese words, and decide on the word order.

As an embodiment of the present invention, a case will be described in which "Write a program tha" is instructed to output the original text.As shown in (2) in FIG.
t reverse search of the work
Suppose that the original text of ``ds in this 5orted file,'' is instructed to be translated.
The process is the same up to the translation word selection for generating the + example. Through syntactic analysis, "5ort" is interpreted as a verb. In the generation translation selection process, the original text output information storage means 7 refers to the information that the word "5ort" has "do" added to the form of "original". 5Ort” translation
"I will." In the conjugation process of generation, there is a rule that when outputting a verb in the original text, it should be ``variable''. Also, from the syntax information that 5Ort'' is passive, rsort
It is used as "was done". When determining the word order, the position is the same as the normal translation result, so the translation result when the output of the original result is instructed is the * mark (2) in Figure 4, that is, "This 5 or
``Write a program that transposes each word in the file t.''

Furthermore, as another embodiment of the present invention, a case will be explained in which the original text output is instructed to be "11aS".As shown in (3) in FIG.
As they listen to Iousic,”
Suppose that the original text of is instructed to be translated. The process up to the generation and utilization process of the above-mentioned "bank" example is exactly the same. Through syntactic analysis, sr as++ is analyzed as a subordinating conjunction.

If you refer to the source output rules when determining the word order for generation, there is a rule that when outputting a subordinating conjunction in the original text, it should be placed at the beginning of the clause with the conjunction, so the translation result when outputting the original text is specified as The xylem in (3) in the figure, namely ras
They listen to music, some students study.''

In this way, the original word specified by the user is displayed in its original form without any translation, making it easier to understand the meaning of the original text.

As is clear from the above explanation, according to the present invention, the original text output rule section does not simply output the original text, but allows the original text to be forwarded or utilized. By considering the word order, you can prevent words from appearing in a translated sentence in an unnatural manner.

In addition, when the original text and its translation are shown to the user in a sentence-by-sentence or sentence-by-sentence manner, since the original text and the translated sentence have the same character strings, when it is in a sentence-by-sentence format, the word-level correspondence is shown.
Also, when using sentences, it is possible to easily understand the correspondence between the original text and the translated text at the sentence level.

[Brief explanation of drawings]

FIG. 1 is a block diagram for explaining an embodiment of a machine translation system according to the present invention, and FIG. 2 is a diagram showing a processing flow of the machine translation system according to the present invention. FIG. 3 is a block diagram for translation processing, FIG. 4 is a diagram showing input sentences and translation results, and FIG. 5 is a diagram showing a rule table of the original text output rule section. DESCRIPTION OF SYMBOLS 1... Input means, 2... Text storage means, 3... Translation means, 4... Dictionary, 5... Display means, 6... Original text output instruction means, 7... Original text output information Memory means, 8...
- Original text output rules section. Patent applicant Rico Co., Ltd. - Diagram No.] Ward diagram ×: Conventional machine translation result O: Correct translation result Book: Translation result when original text output is instructed

Claims (1)

[Claims] 1. An input means for inputting a target language, a storage means for storing the target language input from the input means, a translation unit for translating the target language input by the input means, and necessary for translation by the translation unit In a machine translation device comprising a dictionary section and a display means for outputting a translation result translated by the translation section, an original text output instruction means for instructing to output a certain word in a target language as the original text; comprising an original text output information means for storing information instructed by the original text output instruction means, and an original text output rule section for preventing words left in the original text from appearing in an unnatural form in the translation result, A machine characterized in that it is instructed to output the original text as it is, and when it is translated, it is translated by the dictionary section and the translation section, and the instructed word is output as the original text as a translated text by referring to the original text output rule section. translation system.