JP2896152B2 - Japanese text generation processor - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Summary] Regarding an apparatus for generating Japanese text such as machine translation, etc., it is possible to expand the types of words selected and determined by a user from processing results and efficiently execute the selection. The present invention is directed to a Japanese text generation processing apparatus, which comprises a sentence in a predetermined language, a sentence of a word node having one or more Japanese candidate words each including a word and control information on a connection immediately before and after the word. And a utilization table showing utilization of the term corresponding to each required style, and one of the candidates in the row of the word nodes is input to the output original text by a predetermined word designation input. A word is determined, the corresponding utilization table is determined by a predetermined style designation input, control information on a word following the word immediately before the candidate word, and a word immediately before the candidate word are determined. With utilization unit information of consisting of control information for the connection, configured to have an editing means for determining the word endings immediately before the candidate word.

[Industrial applications]

The present invention relates to an apparatus for generating Japanese text such as machine translation.

When a Japanese text is generated from another language by so-called machine translation, the user often selects a synonym. Also, it is convenient to be able to select styles such as the so-called "is" tone and the "isama" tone.

[Problems to be solved by conventional technology and invention]

FIG. 6 is a block diagram showing a configuration example of a machine translation device. When the translation processing unit 1 of the device reads the data of the original sentence 2 described in a predetermined language, the translation processing unit 1 analyzes the original sentence by searching the word dictionary 3 with a character string obtained by dividing the original sentence, and determines the grammatical properties such as the part of speech. Take out. This information and translation processing unit 1
Based on the grammar rules built in, determine the word order of the translation,
A Japanese word corresponding to each word is extracted from the word dictionary 3 and a Japanese text 4 is generated.

In the above translation processing, when there are a plurality of words in the translated sentence corresponding to the words in the original sentence, the translation processing unit 1 displays, for example, one candidate text on the display device 5 of the console 7 of the user, and When the user looks at it and designates one word by an appropriate predetermined operation using an input device 6 such as a keyboard, the synonym candidate words are sequentially displayed in accordance with the predetermined operation, and the user is presented with the candidate word of the synonym. One is selected, and thus the translation processing unit 1 completes one text by the selected word.

Therefore, as a result of the translation processing, the translation processing unit 1 creates word nodes of the list of synonyms (parts respectively enclosed by broken lines) as shown in FIG. 7, for example.
For example, a sentence based on the word in the first term of each word node is displayed as a candidate, and when the user designates a certain word as described above, the candidate words of the word node are sequentially displayed and selected.

Although synonyms can be selected in this way, for words that are utilized, for example, “buy it” in the example of FIG. 7, synonyms such as “buy it” and “buy it” are used. There can be phrases, and there is a synonym of "buy" such as "buy", so that a synonymous phrase can be considered for each of them, and the number of candidates increases. In the case of "bought", "bought", "purchased", "purchased"
Such expressions can be candidates.

For words that have such a conjugation, there are many combinations of the stem and the ending and the word that follows, and if all of them are saved, the word node becomes large, so the selectable words are limited to nouns, However, there is a problem that words having inflections such as adjectives, that is, declinable words are determined by the apparatus as in the case of FIG. 6, or a plurality of candidates must be limited even if they are allowed. .

An object of the present invention is to provide a Japanese text generation processing device that expands the types of words that can be selected and determined by a user from a translation result text even in the case of declinable words so that the selection can be performed efficiently. I do.

[Means for solving the problem]

 FIG. 1 is a block diagram showing the configuration of the present invention.

The figure shows the configuration of a Japanese text generation processing device. Reference numeral 10 denotes a prototype generation unit that processes a source text 2 and outputs a prototype text 11 in which synonyms are juxtaposed. Reference numeral 12 denotes a utilization table according to a user's instruction input. This is a post-editing unit that processes with reference to 13 and generates a Japanese text 14.

(Operation)

The prototype generation unit 10 of the Japanese text generation processing apparatus processes the sentence 2 in a predetermined language, and generates a sequence of word nodes having one or more Japanese candidate words including a word and control information of the word, Outputs the original text 11.

The post-editing section 12 has an inflection table 13 for each required body, which indicates the use of the corresponding verbs, and for the original text,
One candidate word in the row of word nodes is determined by a predetermined word designation input, and a corresponding utilization table is determined by a predetermined stylus designation input, and control information on a succeeding word of the word immediately before the candidate word is determined. The ending part of the word immediately before the candidate word is determined using the utilization unit information including the control information on the connection with the word immediately before the candidate word, and the Japanese text 14 is generated.

With this processing device, the types of words for which synonyms can be selected are expanded, the style can be selected, and post-editing including them can be efficiently processed separately from the processing such as translation of sentences.

〔Example〕

FIG. 2 is a block diagram of a machine translation apparatus showing an embodiment of the present invention. The translation unit 15 in FIG. 2 translates the original sentence 2 by the same configuration as the conventional translation processing unit 1 and word dictionary 3, but does not perform synonym selection processing. The prototype text 11 having the structure shown in FIG.

FIG. 3 (a) is an example of a prototype text consisting of word nodes according to the present invention of the sentence corresponding to FIG. 7, and the prototype text is constituted by a sequence of word nodes consisting of one or more candidate words as in the prior art. However, each candidate word has a configuration in which a control information column for storing control information such as a relationship between adjacent words on a sentence is arranged on the left and right of a column for storing words.

These control information fields are left blank when no control information is required especially in the case of nouns, but as in the case of the two word nodes at the right end in FIG. In the case of a word node, set the information indicating the type of use in the control information column to the right of the stem of a synonym such as "buy" and "buy", and indicate the word following the verb etc. In the word node of, in the control information column on the left of each word, by selecting this word, information for determining the inflected form such as a verb is set as shown in the figure, so that both can be arbitrarily combined Utilization unit information that regulates the use of declinable words in a combination of. The control information is set by, for example, a numeral, but is shown by a name indicating the content for easy understanding.

In the case of an adjective, similarly, as shown in FIG. 3 (b), the "thick book" is a form in which the node of the stem "thick" and the node of the noun "book" are related by the inflection unit information. . In addition,
In the case of ending with a verb, as in the example of FIG.
A utilization node information is formed by placing a dummy node to the right of a verb word node.

A utilization table 13 shown in FIG. 1 is provided to extract the ending determined by the utilization unit information. Conjugation table 13 is negated if it is a verb for each conjugation type set in the control information column to the right of the verb,
It is assumed that a table as shown in FIG. 4 indicating the ending of the inflected forms such as past, end, union, etc. is provided for each necessary style.

When the post-editing unit 16 receives the prototype text 11 having the above structure, the post-editing unit 16 displays a Japanese sentence including, for example, the first word of each word node on the display device 5 of the user's scan desk 7 as in the related art. Waiting for input from the input device 6, in the case of a noun, candidate words in the word node are sequentially displayed for each noun designated by the user, for example, at every predetermined "next designation" key input. If there is no other input, it is assumed that the candidate word displayed immediately before is selected, and the word of the displayed sentence is replaced.

In the case of a verb, for example, when a word following the stem or the verb is selected in the same manner as described above, the ending is determined in combination with the currently determined candidate word of the adjacent word node, and the previous word is determined using the combination. Replace that part of the sentence. At this time, the processing is performed in consideration of the style designation that has been input by the user in advance or defined as a default value.

FIG. 5 shows an example of post-editing processing by the post-editing unit 16,
When a word is newly selected in the processing step 20 as described above, the word is stored in the processing step 21, and whether or not the word node of the word has control information is identified in the control information column in the processing step 22. If both the left and right control information fields are blank, the process branches to processing step 27, where the display of the corresponding word of the sentence displayed by the selected word is changed, and the process ends. This corresponds to the case of a noun or the like that is not related to a word that is used.

If there is control information, proceed to processing step 23,
The information in the control information column of the selected word and the information in the corresponding control information column of the word currently selected in the adjacent word node on the side opposite to the control information column are taken out to form utilization unit information.

In a processing step 24, a necessary utilization table 13 is determined according to a separately designated stylistic designation, and in a processing step 25, an end determined by the utilization type and the utilization form designation is determined, read out, and stored from the utilization table.

By connecting the endings determined before or after the word selected in the processing step 26 above, the display of the corresponding part of the display sentence is replaced, and the synonym selection processing of one word is completed, thus requiring the user. Is completed, the Japanese text 14 consisting of the selected word is completed.

By the above processing, for example, from the prototype text of FIG. 3 (a), for example, it is assumed that a sentence such as, for example, If it is instructed to change to "", the post-editing unit 16
Based on the utilization unit information determined from the combination with, the hypothetical form of the supposition form of the use of the variable is determined as the ending, and is changed to -if purchased.

If the user changes "ba" to "case", the post-editing section
16 takes the ending of the past tense of the use of a combination of “purchase” and “case”, and makes it like “if purchased”. In the above sentence example, the ending part inserted based on the utilization part information by the above processing is underlined for the sake of explanation.

In the above description, a prototype text is generated by the translation unit 15 that executes machine translation.For example, even in an automatic summarization system that analyzes a large number of documents and creates a summary thereof, the summary result is converted to the prototype text of the present invention. Output as
The present invention can be similarly applied to the case where the digest is edited by the editing unit as described above after being separated from the digest process.

〔The invention's effect〕

As is apparent from the above description, according to the present invention, in a machine translation apparatus or the like, the types of synonyms that can be selected and determined by the user from the text of the processing result can be expanded, and the selection processing can be executed efficiently. There is a significant industrial effect of improving the performance and economics of the device.

[Brief description of the drawings]

 FIG. 1 is a block diagram showing a configuration of the present invention, FIG. 2 is a block diagram showing an embodiment of the present invention, FIG. 3 is an explanatory diagram of a prototype text of the present invention, FIG. FIG. 5 is a flowchart of the process of the present invention, FIG. 6 is a block diagram showing a conventional configuration example, and FIG. 7 is an explanatory diagram of conventional synonym selection. In the figure, 1 is a translation processing unit, 2 is an original sentence, 3 is a word dictionary, 4 and 14 are Japanese texts, 5 is a display device, 6 is an input device, 7 is a console, 10 is a prototype generation unit, and 11 is a prototype. Text, 12 and 16 are post-editing units, 13 is a utilization table, 15 is a translation unit, and 20 to 27 are processing steps.

Claims (1)

(57) [Claims] 1. A translation means for outputting an original text from a sentence in a predetermined language as a sequence of word nodes having one or more Japanese candidate words each including a word and control information on a connection immediately before and after the word. And a utilization table showing utilization of corresponding verbs for each required style, and determining one of the candidate words in the row of the word nodes by inputting a predetermined word with respect to the output original text. And a utilization unit comprising: control information relating to a succeeding word of the word immediately preceding the candidate word; and control information relating to connection with the word immediately preceding the candidate word. A Japanese text generation processing apparatus comprising an editing unit that determines an end of a word immediately before the candidate word using the information.