JPH0528179A - Machine translation device - Google Patents

Abstract

PURPOSE:To prevent a character string written in a foreign language from appearing in a translation and to improve the translation by recognizing a character string written in Roman letters as a character string and replacing it by KATAKANA (cursive form of Japanese syllabary writing) when an unregistered word in a dictionary appears during the translation. CONSTITUTION:When an unregistered word appears during the machine translation of a foreign word character string inputted by an input means 1, whether or not a Roman letter recognition display part 8 can recognize the foreign word character string stored in an unknown word storage part 5 as a Roman letter is judged. If it is recognized as a Roman letter, a Roman letter/ KATAKANA conversion part 7 performs KATAKANA conversion using a Roman letter KATAKANA table 6, and outputs the translation result.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a machine translation device from a foreign language to Japanese.

[0002]

2. Description of the Related Art Conventionally, in a machine translation device, a word that has not been registered in a word dictionary cannot be understood as the semantic information of the word, so basically, it is processed only by outputting a character string as it is. It was

[0003]

If the translation processing is performed while leaving the words that are not registered in the word dictionary as they are, the quality of the output translated sentence will be deteriorated.

For example, the following foreign words "Shinkanse"
It is unnatural for a Japanese sentence to leave a character string such as “n”, “Karaoke”, “Fujiyama” in the translated sentence as it is (in this case, the Japanese sentence), and it is difficult to read the sentence. Bring results. Therefore, even if an attempt is made to register such unregistered words in the dictionary, it is impossible to register all of them in a limited storage area. In other words, the unregistered words in the dictionary had to be manually processed by post-correction. However, manual processing goes against the intention of using a computer for translation and performing the translation processing without human intervention.

SUMMARY OF THE INVENTION It is an object of the present invention to solve the above conventional problems and provide a machine translation device that outputs a high quality translated sentence.

[0006]

In order to achieve this object, an automatic translation apparatus of the present invention has an input means for inputting a character string of a foreign language sentence and an input memory for storing the character string input by the input means. Section, a word dictionary section that stores a Japanese translation corresponding to a foreign language character string, a translation execution control section that translates a foreign language sentence using the word dictionary section, and a translation execution control section that stores the word dictionary section. An unknown word storage unit that stores unregistered words, a Roman character Katakana table in which Roman characters and Katakana correspond one-to-one, a Roman character Katakana conversion unit that replaces Roman characters with Katakana, and an unknown word in the unknown word storage unit. A Roman character recognition control unit that recognizes as a Roman character string and controls whether to replace with Katakana,
The configuration includes a translated text storage unit that stores translated texts, and a translated text display unit that displays translated texts.

[0007]

With the above structure, the present invention reduces the number of alphabetic character strings appearing as unregistered words in a translated sentence by considering Japanese words that are not registered in the dictionary as Roman letters and replacing them with katakana.

[0008]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a functional block diagram of a machine translation device according to an embodiment of the present invention. In the figure, 1 is an input means for inputting a character string of a foreign language sentence, 2 is an input storage unit for storing the foreign language character string input from the input means 1, and 3 is a Japanese translation corresponding to the foreign language sentence. A word dictionary unit 4, a translation execution control unit that translates a foreign language sentence using the word dictionary unit 3 and an unknown word 5 that stores words that have not been registered in the word dictionary unit 3 by the translation execution control unit 4. A memory unit, 6 is a Roman character Katakana table in which Roman characters and Katakana correspond one-to-one, 7 is a Roman character Katakana conversion unit that replaces Roman characters with Katakana, and 8 recognizes whether an unknown word in the unknown word storage unit of 5 is a Roman character. A Roman character recognition control unit that controls whether or not to replace an unknown word with katakana, a translated text storage unit 9 that stores a translated text, and a translated text display unit that displays a translation result.

In the machine translation device having the above components, the operation of this device will be described below. As an example, the case of translating English into Japanese is explained.

First, the translation execution control unit 4 operates the English character string input by the input means 1 to execute translation. In the process of this translation, the word dictionary 3 is searched for each word that constitutes the input English character string. At this time, the translation processing is continuously performed for the words registered in the word dictionary, and the dictionary unregistered processing is performed for the words not registered in the word dictionary.

Hereinafter, the dictionary unregistered word processing performed when the word dictionary is looked up will be described with reference to the flowchart of FIG.

First, a translation process is executed in step A, and a word dictionary is searched in step B at that time. Next, in step C, it is determined whether the word is registered in the dictionary. If the word to be searched is registered in the word dictionary, go to step D, and if not, go to step E. Step D
Then, information on the word is taken out from the word dictionary, added to the word, and then the process proceeds to step H. In step E, it is determined whether or not an English word, that is, a character string of an alphabet, matches a Roman character pattern for a word that is not registered in the word dictionary. That is, the character string is divided into two or three character strings starting with one vowel character or consonant using the Roman alphabet katakana table of FIG. . In step F, the character string determined to match the English character string as a result of step E is replaced with the Katakana character corresponding to the Roman character in the Roman character Katakana table of FIG. 3, and the process proceeds to Step H. In step G, if it is determined in step E that the character string of the English word does not match the Roman alphabet pattern, the word is left as it is as a dictionary unregistered word, and the existing unregistered word in the translation process is processed. In step H, it is determined whether to search the word dictionary. If it is not necessary to search the word dictionary, the dictionary unregistered word processing is terminated. If not, the procedure returns to step A to search the word dictionary.

Further, the actual processing will be described in detail with a specific example. Suppose there is a translation request for the following English text.

"She likes to sing karaoke." It is assumed that the word "Karaoke" is not registered in the dictionary in this example. Then, the existing machine translation device outputs the translation result "She likes to sing Karaoke." Here, it is judged whether or not the character string "Karaoke" can be recognized as a Roman character. That is, this character string can be divided into "ka", "ra", "o", and "ke", and matches the Roman character pattern of the Roman character Katakana conversion table in FIG. If each of the divided character strings is converted into katakana by the Roman character katakana conversion table of FIG. 3 and synthesized, it can be expressed as a katakana character string of “karaoke”, and the resulting translated sentence is
"She likes to sing karaoke." Is output.

As another example, consider the following English sentence. "I watch the TV program of NHK." As a result of translating this sentence, "I am watching a NHK TV program."
Is output. Here, "NHK" is a dictionary unregistered word,
The translated English string "NHK" is output as the translation result. The string "NHK" does not match the Roman pattern. That is, the character string cannot be divided into one vowel character, two consonant characters, or three characters, and the dictionary unregistered word processing of the existing machine translation device can be performed. It is expected that such a word "NHK" will appear as it is in the text even in the expression of a Japanese sentence.

Although an example of translation from English to Japanese has been given in the above embodiment, if a foreign language that does not hinder the operation of considering unregistered words as Roman letters and converting them to Katakana, it is converted to English. It is applicable without limitation.

[0017]

As described above, according to the machine translation device of the present invention, an unregistered word is converted by converting an alphabetic character string of a foreign language word determined to be a dictionary unregistered word into a Roman character and converting it into katakana. As a result, the character strings of the alphabet appearing in the translated text are reduced, and the translated text becomes easier to read.
In addition, it is not necessary to correct an unregistered word that appears in the translation result using an editor, and it is not necessary to register a foreign word that is not registered in the dictionary in the dictionary.

[Brief description of drawings]

FIG. 1 is a functional block diagram of a machine translation device according to an embodiment of the present invention.

FIG. 2 is a flowchart showing the operation of Roman character katakana conversion processing for unregistered words in the embodiment.

FIG. 3 is a romaji katakana table in the same embodiment.

[Explanation of symbols]

 1 input means 2 input storage unit 3 word dictionary unit 4 translation execution control unit 5 unknown word storage unit 6 Romaji katakana table 7 Romaji katakana conversion unit 8 Romaji recognition control unit 9 translation sentence storage unit 10 translation display unit

Claims (1)

[Claims] [Claim 1] Input means for inputting a character string of a foreign language sentence,
An input storage unit that stores a character string input by the input unit, a word dictionary unit that stores a Japanese translation corresponding to a foreign language character string, and a translation execution control that translates a foreign language sentence using the word dictionary unit. Section, an unknown word storage section that stores words that have not been registered in the word dictionary section in the translation execution control section, a Roman character katakana table in which Roman characters and Katakana correspond one-to-one, and Roman character katakana that replaces Roman characters with katakana. A conversion unit, a Roman character recognition control unit that controls whether to recognize an unknown word in the unknown word storage unit as a Roman character string and replace it with katakana, a translation storage unit that stores a translation, and a translation that displays the translation. A machine translation device having a display unit.