JPS62267871A - Japanese word producing device - Google Patents

Abstract

PURPOSE:To show the noun phrase without losing the naturalness of the Japanese language by discriminating the character types of the translated words of a noun obtained from a dictionary means and deciding the connecting description between nouns in response to said character type in case a foreign language sentence includes a noun phrase containing plural nouns. CONSTITUTION:When a foreign language sentence includes a noun phrase containing plural nouns, the character types of the equivalents of the noun obtained from a dictionary means are discriminated. Based on this discriminated character type, the connecting description between nouns is decided for production of a Japanese sentence. For instance, a translated sentence producing part 26 decides whether each noun forming a noun phrase is equal to a single Chinese word, 2 Chinese words, a KATAKANA (square form of Japanese syllabary) character, a Japanese word or the description of a direct foreign language like English, etc. Then the part 26 produces the connecting description between nouns in terms of a surface layer in accordance with the descriving way of the noun, i.e., the character type of a word.

Description

DETAILED DESCRIPTION OF THE INVENTION Technical Field The present invention relates to an 11-word, 1-juku generation device useful for a 1-language translation device, particularly a 1-language translation device.

For example, when creating a corresponding II Kizume sentence from a foreign sentence such as English 1: h, the human-generated form iE1 element of the English sentence is analyzed, and its syntax is analyzed. Then, the sentence structure is converted, and a Japanese translation is then generated.

In Japanese sentences, as in Okachi, noun phrases consisting of multiple nouns are generally created by inserting the particle ``no'' between two adjacent nouns, inserting a midpoint ``・'', or inserting nothing at all. It is written in various ways, such as as a compound word of multiple nouns, or when words in a foreign language such as English are written as they are, they are listed with a space between them.

With conventional English-Japanese automatic translation devices, when such a noun phrase consisting of multiple nouns is translated into the corresponding Japanese, it is simply a list of multiple nouns or a particle between each noun. It was often written in an unnatural way as a single Japanese word, such as by inserting ``no''.

However, real Japanese sentences are quite complex. For example, in the case of the English noun phrase "a woman director."

To simplify the explanation, the Japanese words for "woman" are generally "Onna" and "Josei," and "dire" is the Japanese equivalent of "woman."
Assume that the Japanese words corresponding to "director" generally include "director" and "editor." In this case - in the crotch, 'a
Translations of the word ``woman director'' include ``female director,'' ``female editor,'' ``female director,'' and ``female editor,'' each of which may have slightly different usages or nuances in a sentence. Therefore, as in the conventional method, if you list multiple nouns in 'n' in the -rison, or just insert the particle 'no' between each noun, it will often be unnatural as a 11-hongo. .

the purpose Suekan II eliminates these drawbacks of the conventional technology.

More natural Japanese is generated regarding the notation of noun phrases1
The purpose is to provide a 1-tree, 1-h generation device.

Structure: In order to achieve the above object, the present invention includes a dictionary means storing dictionary data regarding Japanese translations corresponding to words in a foreign language, and a dictionary storing dictionary data regarding Japanese translations corresponding to words in a foreign language, and a dictionary regarding sentences in a foreign language whose morphology and syntax have been analyzed. and a Japanese language generating means that indexes the means and generates a corresponding Japanese sentence, and the Japanese language generating means is configured to generate a Japanese sentence that is better than the dictionary means when there is a noun phrase containing multiple nouns in the foreign language sentence. The present invention is characterized by a Japanese language generation device that determines the character type of the translated word of the noun that is entered, determines the connection notation between the nouns according to the character type, and generates a Japanese sentence. Hereinafter, the present invention will be specifically explained based on examples.

Referring to FIG. 1, there is shown the overall configuration of an embodiment in which the translation generation method of the automatic translation device according to the present invention is applied to an English-Japanese automatic translation device. It goes without saying that the present invention is effectively applied not only to an English-Japanese automatic translation device that translates English into one language, but also to an automatic translation device that translates other foreign languages into Japanese.

The embodiment has an input section 10, through which an English text 12 to be translated into Japanese is input. For example, the human power department 10 may include a keyboard having character keys such as alphanumeric keys and mechanical keys, an optical character frame J111 "37) (OCR) for reading English text recorded on paper, and/or a magnetic disk. It may include a file storage device for reading English text recorded on a storage medium.

The English text manually created by filO is 1 part 1.
It is read into the editing section 14 and preprocessed for translation.

Here, we mainly perform sentence recognition and processing of unknown words.
This will be abandoned as the - part of the form y5 elementary analysis.

The edited English data is transferred to the form 1 elemental analysis 1-i111B together with the information obtained by the IIF editing. The morphological analysis unit 16 indexes the word dictionary 18 and divides it into sentences;
1. Analyze the morphological elements of English sentences, process unknown words, summarize various types of expressions such as proper nouns, time expressions, and number expressions, and process entire sentences such as additional questions and appositive recognition.

The form j8 elementary analysis rule is stored in the analysis rule file 3B.

The English sentence data subjected to the morphological analysis in this manner is transferred to the syntactic analysis section I 20 together with the dictionary information obtained by the morphological analysis. In this embodiment, the syntactic analysis part 1 20 uses cfg grammar (c
Apply the ontext-free grammar (context-free grammar) rules to English data to analyze sentences.
This is a functional unit that analyzes the surface structure in an all-up, right-to-left manner and finds all syntactic possibilities.

The English data parsed (barsed) by the parser 1 section 20 is sent to the parser 11 section 22 together with the analysis information. Here, a structural description is applied to select a solution from the surface bar results obtained by a parser. This creates a reliable parse tree for the English sentence and creates its structure.

These parsing rules are also stored in the parsing rule file 3B.

The parsed English data is transferred to the structure conversion unit 24 as parse tree data. The structure conversion unit 24 creates a syntax tree for a corresponding Japanese sentence from a syntax tree that is an intermediate structure of an English sentence, and converts the Japanese sentence into a basic Japanese structure that is easy to translate.

The syntax tree data representing the Japanese syntax tree whose structure has been converted in this way is sent to the translation generation unit 26, and the latter generates a translation. This is 1 from the tree structure of the Japanese syntax tree.
This is a function that generates one sentence from 111 onwards. First, 1,1
11 "In order to match 1 with that of 11 1:b,
j ce 1 Perform syntax generation to change the tree structure by permuting the ordinal order, and 1st form generation is performed to generate t in the syntax tree.
Translations are generated op-down and left-to-right.

The Japanese sentence data for which one translated sentence has been generated, that is, the translated sentence data, is sent to the post-editing section 30. The post-editing unit 30 uses the information used in the translation process to index the dictionary 1B and correct the translated data to complete a more natural Japanese sentence. This Japanese sentence data is transferred to the output unit 32 and translated "
This is outputted to the output section 32 as a frozen sentence 34. Output unit 32 includes, for example, a printer, a display, and/or a file storage device such as a magnetic disk.

The flow of a series of these translation processes is controlled by a control unit 38 that controls the entire apparatus.

cn word dictionary;! j18 stores lexical data for English and Japanese words in this embodiment, and the analysis rule file 36 stores rule data for morphological analysis and 4+4 sentence analysis.

An operation display section 40 is connected to the control section 38 . The operation display section 38 allows the operator to give various instructions to the device, such as operation keys such as translation instruction keys and cursor keys, input English text, translated Japanese text, intermediate data such as dictionary information, etc. It has a display and indicators that visually display various instructions to the operator.

Note that many of these operation display functions may be configured to be included in the keyboard if the input unit 10 is equipped with a keyboard, or in the display if the output unit 32 is equipped with a display.

By the way, the syntactic analysis unit 20 converts the cfg grammar rules stored in the file storage device 36 into English text for the English data that has been subjected to the morphological analysis.
ft to derive all possible syntactic solutions for that sentence. This solution is generally understood in the form of a structural tree. This means that the 11 words or phrases contained in each sentence are related to each other through subordination or co-occurrence relationships such as saturation relationships and case relationships, such as parent, r-
This indicates the dependent relationship of phase II: such as one grandchild. Each word or phrase occupies the position of a whistle point or node in the structure tree.

When the data defining the set of syntax-1- and the dependency relationship is obtained in this way, it is sent to the syntax analysis section II 24. This data is easily understood in the form of the syntax tree described above. This is further converted into 11 lines by the structure conversion unit 24:
The processing of the nodes in the structure tree is as follows:
-to-right Do it.

In the translation generation unit 26 of this embodiment, in the process of generating one main language sentence, the following process is performed on a noun phrase containing a plurality of nouns 1;-. In other words, for each group 1i14 that forms a noun phrase, whether it is a pleasant word with a - character, a word with two or more characters, Katakana, Japanese, English, etc. Depending on how the name 5:.4 is written, such as whether it is written as it is in the foreign language, or not, a connection notation between nouns is generated on the surface depending on the character type of the word.

In this embodiment, for example, a concatenation notation table is prepared as part of the analysis rule file 3B, and when such a noun phrase is to be generated, this concatenation notation table is referred to and a concatenation notation corresponding to the noun phrase is selected. It is configured. An example of a concatenation notation table is shown in FIG.

In this table, for each noun that forms a noun phrase, two adjacent nouns are both Chinese words (Kanji l) that have one character.
RlJ or ” code will be assigned. The code indicating this connected state is "1" which specifies the insertion of the particle "no", and "2" which specifies the insertion of the particle "no".
``3'' specifies the insertion of a midpoint ``・'', ``3'' specifies the insertion of one space for one character, and ``4'' specifies that no particle, midpoint, or space be inserted. Note that although English is used in this concatenated notation table, this generally includes not only English+i5 but also 11 words in which foreign language characters are used as they are in the 11-tree, :h sentence.

The translated sentence generation unit 26 in this embodiment expresses the generated Japanese sentence in shift JIS code. Therefore, the character types of the 11 Japanese and Chinese characters are determined using this character code, and a concatenated code is obtained by referring to the concatenated notation table according to the determination result. The processing flow is shown in FIG.

In this example, consecutive nouns X1 (i=1...
2. , , n).

Processing is performed in order from the first English noun (5o). First, the first English noun and the following English nouns are stored in registers N'A and NO (51, 51), and then the English words in registers NA and NB are stored. The dictionary 18 is indexed (53.55) and its Japanese translations are stored in registers JA and JB, respectively (54.58).

Therefore, the translation generation unit 2B performs a process called FUNI (JA, JR) based on the contents of registers JA and JB.
This process (FUNIA, JB) is a process that generates a concatenated notation that is uniquely determined from the notation relationship of the contents of registers JA and JB. For example, in this embodiment, register J'
The character type of the word is determined based on the character code of the first character of the word stored in "A" and "B," and the second character type is determined based on this.
A corresponding concatenation code is generated by referring to the concatenation notation table shown in the figure. For example, in the Shift JIS code, katakana characters are defined as 0!8340 to 018398. Note that when the character type determined as 1 is Chinese, the number of characters in the entire ff' word is also determined, and it is determined whether it is Chinese with one character or b66 with two or more characters. The processing result, that is, the concatenated code, is stored in register NO (
57).

Therefore, depending on the contents of register NO, the particle "no" is added to the Japanese word stored in register JA, and the midpoint "no" is added to the Japanese word stored in register JA.
Processing FUN2 (JA, No) such as adding ", adding a space for one character, adding nothing, etc. is performed, and the result is set in register YAKU. Register YAKU is a storage means in which the Japanese translations generated in this way are sequentially stored.

This kind of process of ``ren'' is performed by the nickname, i!, included in one English name, one book, and four phrases. N1-Nn are executed in order from the beginning (59, 130), thereby the name rat J4
It is not a list of ``, ``7.4 r'', a middle point ``・'', a space, etc. are inserted into the appropriate νJ, making it a natural 11-line LIT'S.
A noun phrase is generated. The translation generated in this way is post-processed in the post-editing section 3o, and displayed on the operation display section 4o. '＋
It is displayed visually, and is also printed out as Japanese text 34 by the output unit 32, for example.

By the way, the above-mentioned processing FUNI (JA, JB) is configured to use character codes to determine the surface structure of Japanese nouns, that is, the character type. However, instead of doing this, or in addition to this, when receiving character type data as orthographic information regarding connected notation in ji, 1:18, and indexing the dictionary 18 using English words to obtain the Japanese translation. Next, we obtain this notation information and perform processing such as adding the particle ``no'' (as described above), adding the midpoint ``φ'', adding a space for one character, and not adding anything. In that case, for the word I;18, notation information of translation 1:5 is registered in the record of each entry as shown in Figure 4.Therefore, such a word, Hs When using the process FIINI (JA, J
In B), the notation information is extracted from the dictionary entries stored in registers JA and JB, and the second
A process is performed to obtain a concatenation code by indexing the concatenation notation table in the figure.

Words like this! ; i The process of concatenated notation using the notation information of a can perform more detailed processing that takes into account the meaning of the word, compared to the process using character codes. Instead of determining the character types of all nouns from character codes and indexing the concatenation table, it is also possible to obtain all the notation information from dictionary information. In addition, a method obtained from dictionary information may be used as an auxiliary method to the method of determining 'r1 from the character code. This provides the benefits of both fine-grained processing, which can be processed more appropriately using dictionary information.

Effect 1According to 191 Azuma, 191, in the production of 11 books 1;Δ, each name forms a noun phrase, 1; depending on the character type of I4, each name;
Processing such as whether or not to add "suke, i, 4 r" to the connection between 1 and 1 is performed. According to this method, the comparison inner cylinder "n
With this process, it is possible to write noun phrases in 11 wooden languages without losing their originality. Therefore, 11 more natural words are generated with respect to the notation of the noun phrase.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the overall configuration of an embodiment in which the Japanese language generation device according to the present invention is applied to an English language translation device. FIG. 3 is an explanatory diagram showing an example of a noun concatenation notation table stored in the analysis rule file. - Flowchart showing an example of noun phrase processing in the generation process. FIG. 4 is an explanatory diagram showing an example of idiom 1 given information about English words stored in the word dictionary in the same embodiment. Explanation of symbols of main parts 10, Input section 161. Shape y elemental analysis unit 18,,, Condolences 1 Satoru Dictionary 20.22. , Syntax analysis unit 24, , Structure conversion unit 26, , Translation generation unit 32, , Output unit 36, , Analysis rule 38, , Control unit 401. Operation display section

Claims (1)

[Claims] 1. Dictionary means storing dictionary data regarding Japanese translations corresponding to words in a foreign language, and indexing the dictionary means for foreign language sentences whose shapes and syntax have been analyzed. a Japanese language generating means for generating a corresponding Japanese sentence based on the foreign language sentence; 1. A Japanese language generation device that generates a Japanese sentence by determining a character type of a translated word of the noun, and determining a connecting notation between the nouns according to the character type. 2. In the device according to claim 1, the translated word is represented by a character code, and the Japanese language generation means determines the character type of the noun based on the character code of the first character of the noun. Characteristic Japanese generation device. 3. The Japanese language generation device according to claim 2, wherein the Japanese language generation means further determines the character type of the noun based on the number of characters forming the noun. 4. In the device according to claim 1, the dictionary means stores dictionary data regarding character types of Japanese translations for nouns, and the Japanese language generation means stores dictionary data obtained from the dictionary means. A Japanese language generation device characterized by determining the character type of a noun from data.