JPH09204429A - Machine translation device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To realize precise translation by referring to inputted reading at the time of KANA(Japanese syllabary)/KANJI(Chinese character) conversion on the Japanese of different sound/same notation. SOLUTION: The character string of a Japanese sentence stored in a KANA/ KANJI conversion result buffer area 32 is segmented. When words having the same notation as the segmented character string exist and the reading differs in a translation morpheme analysis dictionary 49, reading stored in a reading input buffer area 31 is referred to at the time of KANA/KANJI conversion and a dictionary record is selected and translation is executed.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a machine translation apparatus which can improve efficiency of translation processing and obtain natural translated text.

[0002]

2. Description of the Related Art Conventionally, various developments have been made to obtain an accurate translated text in a machine translation device that automatically machine-translates an input original text (Japanese) using a computer and obtains the translated text (English etc.). Being tried.

This type of machine translation device is basically
Morphological analysis of the input original sentence (Japanese) consisting of [Kana] and [Kanji] is divided into predetermined language processing units such as words, and the translation dictionary that stores knowledge information necessary for translation processing is referenced. Then, the translated words (English, etc.) are obtained for each language processing unit, and these translated words are combined according to a predetermined translation rule to generate a translated text for the original text.

[0004]

However, some words written in kanji have two or more ways of reading. For example, "Yoko" and "Yoshi" for "Yoko"
There are street readings, each with a different meaning. "Enough"("Jupun" and "Jubun") and "Half-moon"
The same applies to (“Hangetsu” and “Hantsuki”). In the machine translation device as described above, it was not possible to determine which reading the input word had, so it was sometimes translated incorrectly.

The present invention has been made in order to solve the above-mentioned problems, and it stores a reading when a Chinese character is input, and uses the reading when there are a plurality of readings of the Chinese character. An object of the present invention is to provide a machine translation device capable of realizing highly accurate translation by translating into a word corresponding to the reading intended by the user.

[0006]

In order to achieve this object, a machine translation device according to claim 1 of the present invention comprises an input means for inputting a Japanese reading character string, and a reading of a word. A kana-kanji conversion dictionary that stores the notation for the reading, a kana-kanji conversion device that converts the reading character string input by the input device into a notation character string using the kana-kanji conversion dictionary, and a notation converted by the kana-kanji conversion device. Kana-Kanji conversion result storage means for storing character strings and information necessary for translating Japanese into a second language, such as Japanese notation, reading of the notation, translation of the notation in the second language, etc. A morpheme analysis means for morphologically analyzing the stored translation dictionary and the written character string stored in the kana-kanji conversion result storage means with reference to the information stored in the translation dictionary; The notation character string morphologically analyzed by the elementary analysis means is provided with a translation means for translating it into a second language by using the translation dictionary, and the kana-kanji conversion result storage means further converts the notation character string into a word. Each of them is stored in association with its reading, and the morpheme analysis means, when there are a plurality of words having the same notation as the display character string in the translation dictionary, the notation characters stored in the kana-kanji conversion result storage means. The morphological analysis is performed by referring to the reading of the column.

Therefore, when a plurality of words having the same notation as the target Japanese notation character string exist in the translation dictionary during morphological analysis, the reading of the notation when input is used. , You can determine which word it is and get the correct translation.

[0008]

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

The present invention relates to a machine translation device between Japanese and a second language. Here, a translation device between Japanese and English, that is, a Japanese-English translation device is taken as an example.

First, the overall configuration of the Japanese-English translation apparatus of this embodiment will be described with reference to FIG.

An input device 10 such as a keyboard as an input means for inputting a read character string or an instruction and an output device 50 such as a CRT for outputting a screen as shown in FIG. 2 are provided at the center for controlling the entire device. It is connected to the processing unit (CPU) 20.

The RAM 30 serving as a storage means includes a CPU 20.
Connected to, and a kana-kanji conversion result for storing the kana-kanji conversion character string of the kana-kanji conversion result obtained by kana-kanji conversion of the kana-kanji conversion string area for storing the input kana-kanji character string together with the accompanying information. A buffer area 32, a translation output buffer area 33 for storing a character string obtained by translating the kana-kanji converted character string,
A work area 34 for recording pointer information and the like is stored. Here, the reading input buffer area 31 and the kana-kanji conversion result buffer area 32 constitute the kana-kanji conversion result storage means of the present invention.

In the reading input buffer area 31, the reading input by the input device 10 is stored together with the character number from the beginning as shown in [Reading character string storage] of FIG. 3, for example. Further, in the kana-kanji conversion result buffer area 32, as shown in FIG. 3, the kana-kanji converted character string is marked for each word and is stored together with the reading leading character number corresponding to the word. .

ROM 40 for storing programs, dictionaries, etc.
Are connected to the CPU 20 and are roughly divided into those related to kana-kanji conversion and those related to translation.

The kana-kanji conversion programs include a kana-kanji conversion program 41, which is the kana-kanji conversion means of the present invention, and a kana-kanji conversion dictionary 48, which is the kana-kanji conversion dictionary of the present invention. As shown in FIG. 4, the kana-kanji conversion dictionary 48 stores the reading of each word, the notation (kanji, etc.) of the word, the part-of-speech information of the word, and the like. Also,
It also has a connection table that stores data that defines the connection relationship between words based on part-of-speech information.

Regarding translation, the translation program section 42
And a translation morphological analysis dictionary 49 as the translation dictionary of the present invention
There is. As shown in FIG. 5, the translated morphological analysis dictionary 49 stores information such as Japanese notation, its reading, and part of speech. Here, the translation word (English) information is also provided, but the translation word may be configured to have another dictionary. Regarding the part of speech of the translated word (English), in this case, v is a verb, a is an auxiliary verb, d is a pronoun possessive form, and n is
Represents a noun.

The translation program unit 42 includes a morpheme analysis program 43, an analysis program 44, a conversion program 45, a generation program 46, and a morpheme generation program 4 which will be described later.
7 is stored.

Next, the operation of the Japanese-English machine translation apparatus configured as described above will be described with reference to the flowcharts of FIGS. 6 and 7.

When a key input from the input device 10 is received (S10) and the input is a read character string (S12, YES), the kana character code is stored in the RAM 30.
Is stored in the reading input buffer area 31 (S14).
For example, the reading string "My name is Yoko."
Is input, the kana character code (indicated by hiragana in the figure for easy understanding) is stored in the read input buffer area 31 together with the character number from the beginning, as shown in FIG.

If the input is neither reading (S12 / NO) nor conversion instruction (S16 / NO), other processing is performed (S18).

If the input is a conversion instruction (S16.YE)
S), Kana-Kanji conversion is performed. That is, based on the Kana-Kanji conversion dictionary 48 in the ROM 40 as shown in FIG. 4, the Kana-Kanji conversion program 41 stores the Kana character code stored in the reading input buffer area 31, for example, "My name is proton. Is converted into a kanji-kana mixed sentence (S20) and output as a candidate to the output device 50 (S22). A specific example is shown on the left side of the screen in FIG.

When the confirmation key is not pressed by the next key input (S24) (S26, NO), the process of re-dividing the next candidate or phrase is performed (S28).

If the confirm key is pressed (S26.YE
S): Finalize the original text for Japanese-English translation. That is, in the Kana-Kanji conversion result buffer area 32,
The conversion result is stored, and the character numbers from the beginning of the kana character code stored in the reading input buffer 31 in S12 are stored in association with each word (S20). For example, in the example of [Kana-Kanji conversion result storage] shown in FIG. 3, the word "I" in the Kana-Kanji conversion result is the first to third characters in the reading character string (the next word "no" is the reading character string 4). It is indicated from the th, so it corresponds to "I", and the word "proton" corresponds to
It is stored that it corresponds to the 9th to 11th reading character strings, that is, "Yoko".

Next, when the translation key is pressed (S32.YE
S), translation processing starts.

First, the translation morphological analysis dictionary 49 as shown in FIG.
Morphological analysis processing is performed (S36). Hereinafter, a detailed description will be given with reference to FIG.

First, the character string is cut out (S6).
1). That is, it is checked whether or not there is a word in the translation morphological analysis dictionary 49 in order from the longest character string from the front of the notation stored in the kana-kanji conversion result buffer area 32, and if there is no, it is shortened one by one from the back. Check if there is a word of length. For words whose endings are changed, such as verbs, it is possible to search in the basic form, such as searching in the final form (may be set to search in the stem).

For example, in the case of the original sentence "My name is proton.", First, "My name is proton." Is cut out (S61, S62, NO), and the translation morphological analysis as shown in FIG. 5 is performed. Search the Japanese notation in the dictionary 49 (S6
3), but there is no such notation (S64, NO),
Delete one of the strings and say, "My name is Yoko."
The same processing is performed. In this case, the character strings are deleted one after another, and when "my" is cut out (S61,
(S62, NO), and matches the notation in the translation morphological analysis dictionary 49 (S64, YES).

Next, it is checked whether or not the same notation as the cut out character string exists in the translation morphological analysis dictionary 49 (S65). In this case, since there is not more than one "my" notation (S65 / NO), the connection is checked with Japanese part-of-speech information (S69), and if the connection is OK (S65).
70, YES), based on the translation morphological analysis dictionary 49, the translated word and the part of speech corresponding to the word, together with the break of the word,
It is temporarily stored in the work area 34 (S71).

In this way, "my", "name",
"Ha" is cut out, and when the connection check is memorized, "proton" is cut out next (S6).
1, S62, NO). Then, the translation morphological analysis dictionary 49
Is present (S64, YES), but the notation "proton" is two and plural (S65, YES). Record "proton" registered in translation morphological analysis dictionary 49
The readings of “Kana-Kanji” and “Yoshi” are different reading words (S66, YES). Therefore, the Kana-Kanji conversion result notation stored in the Kana-Kanji conversion result buffer area 32 and the reading input buffer 31 The reading which is stored in is referred to (S67), and which reading word is selected (S68).

For example, in the case of being stored as shown in FIG. 3, the character string "proton" of [Kana-Kanji conversion result] is stored together with the number "9", and the next word of "proton" is "12". It is memorized with the number. Therefore, since it corresponds to the "9" th of [reading character string storage], the number obtained by subtracting 1 from "12", that is, "11"
It is determined that the kana corresponding to the second is the reading corresponding to "proton" (S67).

Since the reading "YOKO" has been determined here, this reading "YOKO" is included in the word written as "proton" in the translation morphological analysis dictionary 49 shown in FIG.
The word with is selected (S68).

Next, the connection is checked (S69), and if it is determined that the connection is OK (S70, YES), the translated word "Yoko" and the part of speech "n (" Noun) ”is stored in the work area 34 together with the word break.

By repeating the above processing, "my""name"
"Ha", "Yoko", "Dai", "."
It is temporarily stored in the work area 34 together with the part of speech of the translated word,
When there is no character string to be cut out (S61, S62 ・ N
O), the morphological analysis process ends. The above morphological analysis processing steps constitute the morphological analysis means of the present invention.

Returning to FIG. 7, as is well known from here on, a parsing is performed by the analysis program 44 to obtain the structure of the part of speech sequence, its dependency relationship, etc. (S38).

After that, the conversion program 45 converts the original Japanese structure into the translated English structure (S40).

Next, the generation program 46 rearranges the translated word candidates of each word previously stored in the work area 34 based on the syntactic structure of the translated text, in the order of words according to the language form of the translated text, An English translation of the Japanese original is obtained (S42).

Then, the morpheme generation program 47 performs transformation processing such as changing the ending of the basic form of each translated word candidate that constitutes the translated sentence candidate (S44), and completes the translated sentence as an appropriate language expression, and outputs the output device 50. To output as shown on the right side of the screen in FIG.

After that, if the user's key input is the confirm key (S50, YES), the translated sentence is confirmed (S54), post-editing is performed (S56), and the translation is completed.

As described above, in the machine translation device of this embodiment, there are two different pronunciation words (eg, "proton" and "yoko" and "youshi"). ), Since the translation is performed by referring to the reading, the word that became “Yoko” as a result of Kana-Kanji conversion of the reading character string “Yoko” has the translated word “Yoko” and the reading character string “Yoshi”. Kana-Kanji conversion and translation of the word "proton" results in "proton". Therefore, as in the past, it is possible to realize a highly accurate translation without being mistakenly translated as another word having the same notation but different reading.

The present invention is not limited to Japanese-English translation, but can be applied to a machine translation device for translating Japanese into another language.

[0041]

As is apparent from the above description, according to the machine translation device of the first aspect of the present invention, when translating Japanese into a second language, words with different allophones are used. Refers to the reading when Japanese is input, so it is possible to correctly add the translation of the word, prevent mistranslation due to difference in reading, search for another translation from unintended translation results, It is possible to translate smoothly without re-translating.

[Brief description of drawings]

FIG. 1 is a block diagram of a control unit of a machine translation device of this embodiment.

FIG. 2 is a diagram showing an example of a display form of a screen of an output unit of the machine translation device of this embodiment.

FIG. 3 is a diagram showing an example of stored contents of a kana-kanji conversion result storage means of the present embodiment.

FIG. 4 is a conceptual diagram of contents of a Kana-Kanji conversion dictionary according to the present embodiment.

FIG. 5 is a conceptual diagram of the contents of a translation morphological analysis dictionary of this embodiment.

FIG. 6 is a flowchart showing the first half of the flow of processing of the present embodiment.

FIG. 7 is a flowchart showing the latter half of the processing flow of this embodiment.

FIG. 8 is a flowchart showing a flow of morphological analysis processing according to the present embodiment.

[Explanation of symbols] 10 input device 20 central processing unit 30 RAM area 31 reading input buffer area 32 kana-kanji conversion result buffer area 40 ROM area 41 kana-kanji conversion program 43 morphological analysis program 48 kana-kanji conversion dictionary 49 translation morphological analysis dictionary

Claims (1)

[Claims] 1. An input unit for inputting a Japanese reading character string, a Kana-Kanji conversion dictionary storing a reading of a word and a notation for the reading, and a reading character string input by the input unit, the Kana-Kanji character. Kana-Kanji conversion means for converting into a notation character string using a conversion dictionary, Kana-Kanji conversion result storage means for storing the notation character string converted by the Kana-Kanji conversion means, and necessary for translating Japanese into a second language A translation dictionary that stores information such as Japanese notation, reading of the notation, and translations of the notation in a second language, and the notation character string stored in the Kana-Kanji conversion result storage means A morpheme analysis means for performing a morpheme analysis while referring to the information stored in the second phrase using the translation dictionary for the notation character string morpheme analyzed by the morpheme analysis means. In a machine translation device including a translation unit that translates into words, the Kana-Kanji conversion result storage unit stores the converted notation character string in association with the reading of each word, and the morphological analysis unit includes the translation unit. When a plurality of words having the same notation as the display character string are present in the dictionary, the morphological analysis is performed by referring to the reading of the notation character string stored in the kana-kanji conversion result storage means. Machine translation device.