JPH0728820A - Machine translation device - Google Patents

Abstract

(57) [Summary] [Purpose] A machine translation device can handle input sentences that are conventionally written in an error-prone notation, reducing the burden on the user during post-editing. . [Structure] The input Japanese sentence is pre-edited into a form suitable for machine translation, then translated into English by referring to the dictionary section in the translation section, and the translated English sentence is edited in the post-editing section. Sentences and English sentences are displayed on the display unit. At this time, if there is a sentence that is not translated correctly because the Japanese sentence contains a word that is commonly used but is incorrect in the orthography, operating the retranslation key causes the reading of the Japanese headword. If a headword having the same reading and the same part-of-speech is searched from the dictionary unit that also stores (S12) and the conditions such as the semantic information of the headword are satisfied (S14, YES),
The translation of the entry word is applied (S15).

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.

[0002]

2. Description of the Related Art In recent years, a machine translation system that automatically translates an input sentence using a computer has been receiving attention.
Such a machine translation device basically divides a Japanese sentence as an input sentence into predetermined processing units such as words and phrases by performing morphological analysis and syntactic analysis, and searches a plurality of translation dictionaries for each processing unit. And found the corresponding English translation,
By combining them according to a predetermined translation generation rule,
It is configured to obtain an English sentence as an output sentence.

In a conventional machine translation device, a variation due to a notation is registered as a different notation of a headword, and the dictionary is searched even with the different notation.
For example, even if "computer" is written as "computer", the entry word is registered as "computer" and "computer" is entered as a different notation so that "computer" can be drawn.

Further, in the conventional machine translation system, "early" is applied to "fast" and "fast" is applied to "fast" by distinguishing "fast" from "fast". It was Therefore, "The speed is fast."
peed is fast. ”

[0005]

However, in the machine translation system described above, there are a plurality of notations,
And, conventionally, when all of them were correct, I decided to translate one of them, but if the fluctuation of the notation is likely to be in the area of misuse, I will not accept the different notation,
It was processed within the conventional frame. Therefore, for example, if you enter "The speed is fast."
It was not possible to correctly translate it because it would output an incorrect translation such as "y." or it would be output without being translated as an unregistered word.

In order to correct this, the user may rewrite "early", which is judged to be inappropriate, as "fast" by post-editing, or correct the Japanese sentence as "fast" and re-translate it. It had to be done and the process was tedious.

The present invention has been made in order to solve the above-mentioned problems, and it is extremely common even if a word which is commonly used and is wrong in orthography is included in an input sentence. It is an object of the present invention to provide a machine translation device that can easily and correctly translate.

[0008]

To achieve this object, a machine translation apparatus of the present invention comprises a pre-editing unit for editing an input first language into a form suitable for machine translation, and a first language A dictionary unit that stores a part of speech corresponding to a headword and a translated word of the second language, a translation unit that uses the dictionary unit to translate the first language into the second language, and a second language that is a translation result is edited. A machine translation device comprising a post-editing unit for displaying and a display unit for displaying a first language and a second language, the reading storage area being provided in the dictionary unit and storing the reading of each headword of the first language. And a retranslation command means for instructing retranslation of the translated text translated by the translation part, and a headword having the same reading and the same part of speech is searched from the dictionary part based on the command of the retranslation command means. If the conditions such as the meaning information of the entry word are met, the finding By applying a translation of the word and a re-translation means for translating.

[0009]

In the machine translation apparatus of the present invention having the above configuration, the input first language is preedited into a form suitable for machine translation, and then the translation unit refers to the dictionary unit to convert it into the second language. The translated second language is translated in the post-editing section, and the first and second languages are displayed on the display section.

At this time, if there is a sentence which is commonly used but is incorrect in the orthography and is not correctly translated, the retranslation command means is operated to perform the retranslation. Based on the command, the retranslation means searches for a headword having the same reading and the same part-of-speech from the dictionary unit that also stores the reading of the headword in the first language, and conditions such as the semantic information of the headword meet. In this case, the translation of the headword is applied, and a sentence that is not translated correctly due to a different notation can also be translated correctly.

[0011]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings.

An embodiment in which the first language is Japanese and the second language is English will be described here.

FIG. 1 is a block diagram functionally showing a hardware configuration of a machine translation device according to the present invention. The machine translation device 10 includes a data processing unit 12 including a CPU, an input unit 14 including a keyboard, and a CRT.
And the like, and a printing unit 18 such as a scanning printer.

In the data processing unit 12, a pre-editing unit 20 for editing the input sentence input from the input unit 14 into an appropriate form as necessary, and a second language receiving the input sentence from the pre-editing unit 20 A translation unit 24 for translating to, a dictionary unit 26 stored in a hard disk or an optical disk, and a post-editing unit 28 for editing the translation result sentence obtained by the translation unit 24 are realized by software and built-in. Has been done.

The dictionary unit 26 is composed of an analysis dictionary used during morphological analysis, syntax analysis and syntax generation, and a conversion dictionary used during syntax conversion.

In the translation unit 24, a morpheme analysis unit 30 that performs morphological analysis while referring to the analysis dictionary, divides each word, and adds word information and the like to each word, similarly to the analysis dictionary. While grouping words into phrases,
A syntax analysis unit 32 that performs syntax analysis such as dependency analysis, and a syntax conversion unit 34 that converts Japanese into an English structure based on the result of the syntax analysis, searches a conversion dictionary, and replaces each word with a translated word. And a syntax generation unit 3 that searches the analysis dictionary and generates a correct syntax as English based on the translation generation rule.
6 and a morpheme generation unit 38 for generating a morpheme such as a change in word form.

Next, the translation process in the machine translation device 10 thus configured will be described with reference to FIG.

First, when a Japanese sentence which is an input sentence, for example, "Computer speed is fast" is input by operating the character keys of the input unit 14, and the translation key is operated (S
1), the Japanese sentence is edited by the pre-editing unit 20 into an appropriate form as needed, and then sent to the translation unit 24 (S2). As shown in FIG. 3, the Japanese sentence sent to the translation unit 24 is morphologically analyzed by the morphological analysis unit 30 while referring to the analysis dictionary, divided into word units, and word information or the like is added to each word. (S3). The symbol / in FIG. 3 indicates the division position.

Then, as shown in FIG.
In 2, the words are collected into phrases while referring to the analysis dictionary, and syntax analysis such as dependency analysis is performed (S
4). Then, in the syntax conversion unit 34, as shown in FIG. 5, after converting Japanese into an English structure based on the result of the syntax analysis, the conversion dictionary is searched to convert each Japanese word into an English translation. Replace (S5). That is, here, each case element of Japanese is associated with an English case element, and transformed into an appropriate English structure.

Since "computer" is registered as a different expression in the dictionary of the heading "computer", if "computer" is written in the input sentence,
When searching the conversion dictionary, it is searched as "computer" and is replaced with "computer".

The sentence converted into the English structure by the syntax conversion unit 34 is delivered to the syntax generation unit 36.
First, in this case as well, an analysis dictionary is similarly searched for each word or group of words (phrase), and a correct syntax is generated as English based on the translated word generation rule (S
6). Subsequently, the morpheme generating unit 38 performs morpheme generation such as word form change, and the English translation "The speed of the com
puter is early. "is generated (S7). The English translation thus generated is freely edited by the post-editing unit 28 (S8) and displayed on the display unit 16 or output by the printing unit 18. (S9).

Next, in the machine translation device 10, the input sentence "Computer speed is high."
The control of retranslation when translated as "The speed of the computer is early." Will be described in detail with reference to FIG.

FIG. 6 is a schematic diagram showing a part of the internal structure of the conversion dictionary. "Early" is registered as a parallel translation in the Japanese heading "fast" and "fast" is registered in "fast". It is registered. FIG. 7 is a flowchart showing the retranslation syntax conversion process.

[The speed of the computer is earl
If the user determines that the translation is not appropriate and operates the retranslation key (not shown) of the input unit 14 to instruct the retranslation process, the translation process described above is performed. In the same manner, the morphological analysis process (the same process as S3) and the syntax analysis process (the same process as S4) are performed, and the syntax conversion process is executed.

First, the Japanese structure of the input sentence is converted into an English structure based on the result of the syntactic analysis (S11).
Subsequently, the conversion dictionary shown in FIG. 6 is searched to find a Japanese word having the same reading and the same part of speech for each of the Japanese words (S12). Then, for the first Japanese word, a word having the same part of speech and the same reading is searched (S13, YES), and if conditions such as semantic information are satisfied (S14, YES), the word is set as a candidate. A Japanese word is replaced with an English translation based on the conversion dictionary (S15). On the other hand, if Japanese words with the same part-of-speech and the same pronunciation are not searched (S13, NO), or if conditions such as semantic information are not met (S14, N).
O), each word in Japanese is replaced with an English translation by using the word from the previous translation as a candidate (S16).

Next, when the replacement is completed for one word, it is determined whether the replacement is completed for all Japanese words (S17). In this case, since the replacement has not been completed for all the words (S17, NO), the process shifts to S13, the search result for the next Japanese word is determined, and thereafter, the above-mentioned processes (S14-) are performed. S1
4) is performed and the above-mentioned processes are repeated until all Japanese words are replaced with English translations.

Then, in the same way as the above-mentioned translation processing, the syntax generation unit 36 searches the analysis dictionary for each word or group of words (phrase), and constructs the correct English syntax based on the translated word generation rules. The syntax generation process (process similar to S6) is generated. Subsequently, the morpheme generation unit 3
A morpheme generation process (process similar to S7) such as a change in word form is performed in 8, and as a retranslation result, as shown in FIG.
The English translation "The speed of the computer is fast." Is generated and output.

With this configuration, the machine translation apparatus 10 stores the Japanese headword reading data in the conversion dictionary of the dictionary unit 26, and the user issues a retranslation instruction. In this case, during the syntactic conversion process, you can search for a word that has the same reading as the Japanese word with the same part of speech, and if the conditions such as semantic information are met, re-translate that word as a translation candidate. Thus, it is possible to deal with an input sentence that is conventionally written in a notable mistake, and it is possible to reduce the burden on the user during post-editing.

Although the part of speech is not limited at the time of retranslation in the above embodiment, the part of speech may be limited depending on the degree of influence.

[0030]

As described above in detail, according to the machine translation apparatus of the present invention, the dictionary stores the reading data of the headword of the first language, and it is the same when the user issues a retranslation instruction. An input sentence that is conventionally written in an error-prone notation by searching for a word that has the same reading as the part-of-speech word and retranslating that word as a candidate if the conditions such as semantic information meet Can be dealt with, and the burden on the user at the time of post-editing can be reduced.

[Brief description of drawings]

FIG. 1 is a block diagram showing the configuration of a machine translation device to which this embodiment is applied.

FIG. 2 is a flowchart showing translation processing control of this embodiment.

FIG. 3 is a schematic explanatory diagram of a morpheme generation process of the present embodiment.

FIG. 4 is a schematic explanatory diagram of a syntax analysis process of the present embodiment.

FIG. 5 is a schematic explanatory diagram of a syntax conversion process of the present embodiment.

FIG. 6 is a diagram showing a part of the internal structure of the conversion dictionary of the present embodiment.

FIG. 7 is a flowchart showing retranslation processing control of this embodiment.

FIG. 8 is a diagram showing a post-editing screen of the present embodiment.

[Explanation of symbols]

 10 machine translation device 12 data processing unit 14 input unit 16 display unit 18 printing unit 20 pre-editing unit 24 translation unit 26 dictionary 28 post-editing unit

Claims (1)

[Claims] 1. A pre-editing unit that edits an input first language into a form suitable for machine translation, and a dictionary unit that stores a part of speech corresponding to a headword in the first language, a translated word in the second language, and the like.
A translation unit that translates the first language into a second language using the dictionary unit, a post-editing unit that edits the second language that is the translation result, and a display unit that displays the first language and the second language. A machine translation apparatus comprising: a reading storage area provided in the dictionary section for storing readings of respective headwords of the first language; and a retranslation instruction for instructing retranslation of a translated sentence translated by the translation section. Means and the retranslation command means, based on the command of the retranslation command means, search for a headword having the same reading and the same part of speech from the dictionary part, and if the conditions such as the meaning information of the headword meet, A machine translation device, comprising: a retranslation unit that translates by applying bilingual translation.