JPH0264737A - Multilanguage generator - Google Patents

Abstract

PURPOSE:To separate the control of a device and a multilanguage description, to make an output image clear and to easily correspond even when a change occurs in a data structure by providing a multilanguage generating pattern storage means, a generating pattern extracting means, etc. CONSTITUTION:A multilanguage generating pattern storage device 103 stores a multilanguage generating pattern which is the model of a language to be generated such as a Japanese sentence generating pattern and a program generating pattern. A generating pattern extracting device 102 takes elements successively out of timber structure type input data held by an input data holding device 101 until there is no more element and inputs the multilanguage generating pattern corresponding to the elements. Pattern components included in the pattern are taken out successively from the head of each language generating pattern in the device 103, the next suitable element is taken out of the input data by its type, or concerning a program language or Japanese to show the meaning of the present pattern component, the one suitable for a context is selected and outputted. Consequently, even when the data structure occurs in the data structure, the device can correspond easily.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a multilingual generation device that generates Japanese sentences, programming languages, program structure diagrams, etc. from the same tree-structured data.

〔overview〕

The present invention stores multilingual generation patterns in a multilingual generation device that generates Japanese sentences, programming languages, program structure diagrams, etc. from the same tree-structured data, and generates multilingual descriptions of tree-structured data. By separately analyzing the description from control and creating and outputting language generation patterns, the output image can be understood at a glance, and it is possible to respond flexibly to changes in the data structure.

[Conventional technology]

To generate multiple languages, in the tree-structured data that is the source of multiple languages, select appropriate language elements for each keyword element,
For example, nouns, verbs, adjunct word sequences, imperative words, etc. may be associated and output in the permitted word order.

In conventional technology, each element is associated with an appropriate language element,
Procedural descriptions that actually list output instructions when outputting are known. (Reference Makoto Nagao: “Language Engineering” Artificial Intelligence Series 2, Shokodo 1983) With this conventional technology,
To generate multiple languages, separate procedural description output commands are prepared corresponding to the languages to be output.

FIG. 2 shows a hypothetical node that is an example of tree-structured data, and an example of a procedural description that is being generated in multiple languages corresponding to the hypothetical node 201 shown in FIG. 2 is shown below.

Description example print (° If, ') call ("Next node processing procedure stored in C0ND slot I") print ("When,") if <THfl'N Is there a child node in the slot? ? 〉then print (line feed code) ca l
1 (“Procedure for the next node stored in the THBN slot”) print (line feed code) if (Is there a child node in the εLSE slot?>
then print (“otherwise,’) ca
ll ("next node processing procedure stored in εLSE slot") In this example description, the print statement outputs appropriate characters, and the call statement repeatedly calls the next node processing procedure. The part inside “” is actually a kanji code.

The tree-structured data shown in FIG. 2 includes, as elements, a hypothesis node 201, an equivalent condition node 202, an execution node 203,
Addition node 206, assignment code 207, variable code 20
4.208.210.211, value code 205.209
have. The assumption node 201 has an equivalence condition node 202 and an execution node 203 as substructures, each of which has a CD
The N0 slot 212 and T)IBN slot 213 are connected to each other.

[Problem that the invention seeks to solve]

However, this conventional multilingual generation using procedural descriptions has the following drawbacks.

(a) The output image of Japanese sentences and programming languages is
Even if you look at the description, it is not easy to understand. In description example 1, output characters, line breaks, indentation, etc. are not easily recognized, so it is difficult to deal with even small changes.

(b) Because of the procedural description, the control of the multilingual generator and the procedural description are completely dependent. For this reason,
In the hypothetical node procedure of description example 1, an output command or a call statement must be executed after confirming that a specific condition is met using an if statement. therefore,
It was necessary to explicitly insert if statements and call statements into the procedural description, and even the control of the multilingual generator had to be described within the procedure.

(C) In order to perform multilingual generation, procedural descriptions must be created individually, so there are many overlapping parts in those descriptions, such as control of the multilingual generation device.

The present invention solves these shortcomings of the prior art, and separates device control and multilingual description to provide a clear output image and easily adapt to changes in the data structure. The purpose is to provide a multilingual generation device that can.

[Means for solving problems]

The present invention provides a multilingual generation device that generates and outputs a plurality of languages including Japanese sentences or a programming language from tree-structure data. a holding means; a multilingual generation pattern storage means for storing a multilingual generation pattern serving as a template for a language to be generated; and a multilingual generation pattern storage means for extracting each element from the input data held in the input data holding means and generating the multilingual generation pattern. A generated pattern extraction means extracts a pattern corresponding to each element from the storage means, a context search means searches for the context of the data held in the input data holding means, and the output of the context search means extracts the generated pattern. The present invention is characterized by comprising multilingual output means for outputting in a plurality of languages depending on the pattern extracted by the extraction means.

[Effect]

A generated pattern extracting means sequentially extracts elements from the tree-structured input data held in the input data holding means until no elements are left. Enter the multilingual generation pattern corresponding to the currently retrieved element. The pattern components included in the pattern are sequentially extracted from each language generation pattern in the multilingual generation pattern storage means from the beginning, and depending on the type, the appropriate next element is extracted from the input data, or the current pattern component is extracted. Select and output the programming language or Japanese that expresses the meaning according to the context.

〔Example〕

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

FIG. 1 is a block diagram showing the configuration of a multilingual generation device according to an embodiment of the present invention. In FIG. 1, the multilingual generation device of this embodiment includes an input data holding device 101 that holds input tree-structured input data, and a multilingual generation pattern memory that analyzes this input data and corresponds to each element. Device 10
a generation pattern extraction device 102 that extracts generation patterns from each Japanese sentence generation pattern, program generation pattern, Japanese annotated program generation pattern, etc. stored in 3 and outputs it to a multilingual output device 105; and an input data holding device. a context search device 104 that analyzes and searches the context of input data held in 101 and outputs the result to a multilingual output device 105; and a multilingual generation pattern and context search device 104 from this generation pattern extraction device 102. and a multilingual output device 105 that performs multilingual output from data from.

Next, we will explain the operation of this embodiment of the device using an example of outputting Japanese text, a program, and a program with Japanese annotations for the tree-structured data that is the source of multilingual data shown in Figure 2. In particular, we will output Japanese text. This will be explained using an example. Similar processing is performed for other languages as well.

FIG. 3 shows Japanese sentence generation patterns 401 to 405 that are stored in the multilingual generation pattern storage device 103 and are abstracted from Japanese sentences to be generated.

The hypothetical node Japanese sentence generation pattern 401 in FIG. 3 has a hypothetical node name 410 at the beginning as a component.

Hypothetical node name 410 is provided to recognize that this pattern description corresponds to hypothetical node 201 in FIG.

The code 411.412 following this assumed node name 410
．． 413.414 and 415 rmoshi *C0
ND toki #new-1ine *T) The IEJ part is a component of a pattern that is referred to when actually generating Japanese sentences from a hypothetical node. The pattern components 412, 415, and 418 with "*" at the beginning have the lower structure of the hypothetical node 201 in FIG.
12, THEN slot 213, and ELSE slot. However, at this time, the 8LSE slot has a [ ] in the pattern, so
There is no need for ELSE slots to exist below the hypothetical node 201 in FIG. In other words, it moves to the next process without outputting anything. Pattern component 411.41
3.417 respectively correspond to the appropriate Japanese words expressing the processing of the hypothetical node. The pattern is a label representing the Japanese language to be output (in this example, a label in Roman letters)
It is expressed as "moshi"411 and 't
oki"413 and °'5oudenakereba"4
17, "if,""when," and "otherwise," are actually output.

FIG. 6 is a flowchart explaining the operation of this embodiment. According to this flowchart, from the tree-structured data example in FIG. .

Set Y to Z. ” The process of inputting Japanese sentences will be explained step by step.

(2) Input the tree-structured data shown in FIG. 2 into the input data holding device 101 (step 5100O).

■ If the node to be traced no longer exists (step 5100I), the process ends.

■ Hypothetical node 2, which is the top node, from the tree-structured data held in the input data holding device 101
01 is taken out (step 51002).

■ Japanese sentence generation pattern 4 corresponding to hypothetical node 201
01 is extracted from the multilingual generation pattern storage device 103 (step 51003).

■ Step 51 of extracting the component “moshi i” 411 from the beginning of the generation part of the Japanese sentence generation pattern 401
004).

■ Component "mosh i '"411 is'°*"
Since the process has not started with (step 51005), the process is transferred to the multilingual output device 105.

■ The multilingual output device 105 outputs the Japanese word "moshi" corresponding to "moshi i" (step 5100).
6).

■ Extract the next component book C0ND "412 in the Japanese sentence generation pattern 401 (step 51004)
.

■ Component 'Book C0ND "412" starts with '*', so the child node is searched (step 5100).
? ).

■ It can be seen from the component "red OND" that the equivalence condition node 202, which is a lower structure from the hypothetical node 201, is connected below the C0ND slot (step 510).
01, step 51002).

If the equivalent condition node 202 is processed in the same manner as described below, the Japanese conditional sentence at the hypothetical node will be output. The Japanese sentence that is output at this time is "X is equal to 2."
It is.

■ Next component of Japanese sentence generation pattern 401 'tok'
i'' 413 outputs ``toki,'' on the multilingual output device 105 similarly to ■ (step 51006).

■ As the process continues, the next component "$newline" 414 is extracted in the same way as in ■ (step 5).
1004).

■ In the multilingual output device 105, 'llnewli
ne'' 414 outputs a new line code. At this time, components with #''' at the beginning are output by the context search device 104.
outputs an appropriate output by searching the context of the data held in the input data holding device 101.

In this case, the context search device 104 checks whether there are multiple child nodes below the execution node 203.

Since it currently exists, the context search device 104 outputs a line feed code to the multilingual output device 105 and sends information for executing rate reduction from now on (step 51005).

■ Components of Japanese generation pattern 401' Book T HB
N "415 also traces the execution node 203 with the context search device 104 in the same way as ■, and processes in the same manner as ■ and below, then
The Japanese executable sentence of T)IEN at the hypothetical node is output.

■Finally, we process the component "LSE" 418, but since there is no IELSE slot in FIG.
] The part surrounded by brackets is not output.

Through these processes, "If X is equal to 2, Add 1 to Y.

Set Y to 2. ” The statement is output.

Moreover, FIG. 4 is a diagram explaining a program generation pattern for outputting a programming language. The generation pattern shown in FIG. It is output from the device 105.

rlFX=2 COMPUTB Y= Y+1 MOVB Y TO2 END-IF, J Fig. 5 is a diagram explaining the generation pattern of a program with Japanese annotation text, and the generation pattern shown in Fig. 4 allows the above-mentioned ■ By the same operation as ~@, the next Japanese annotated program language is output from the multilingual output device 105.

rlFX=2 / book where X is equal to 21 / CDMPUTB Y= Y +1 / book Y1 equals 1. Set the book/MOVB Y TOZ/book Z1koY. Book/END-IP, J In this way, the multilingual generation patterns are stored in the multilingual generation pattern storage device 103, and the pattern corresponding to the input data is extracted by the generation pattern extraction device 102 and sent to the multilingual output device 105. (1) Referring to the multilingual generation hypothetical node pattern 401.601.801 shown in Figures 3, 4, and 5,
You can see that the pattern can express line breaks and indentation using images that match the output language. The output Japanese is not a kanji code, but is replaced with something that can be understood at a glance, such as Roman characters.

(2) In the multilingual generation pattern description, a component with an * at the beginning (for example, this CDN0412, etc.) represents a slot name, and the child node to be followed next is extracted from the input data holding device according to the slot name. Therefore, the description is simpler than the conventional method of calling a procedure corresponding to a child node. Components prefixed with # (for example, $newline, #i, etc.) call a processing procedure defined outside the pattern so that the output can be selected according to the context. This improves the descriptive nature of this multilingual generation process without introducing control into patterns. [Slots enclosed in square brackets are [ ] when their child nodes do not exist.
Nothing is output inside this. In the conventional method, the inside of the procedure was complicated, but in the multilingual generation pattern, the above method made the control and multilingual description independent.

(3) When it is desired to generate a program in a new programming language, according to the present embodiment, this can be easily done by simply adding a generation pattern representing the syntax of the new programming language.

〔Effect of the invention〕

As explained above, the multilingual generation device of the present invention has a Japanese sentence generation pattern and separates the control of the device from the multilingual description, so the output image of the Japanese sentence can be changed just by looking at the description. It is possible to realize a multilingual generation device that is clear, can easily cope with changes in the data structure that is the source of sentences, and has improved operability in multilingual generation compared to conventional devices.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing the configuration of an embodiment of the present invention. FIG. 2 is a diagram showing an example of the data structure of input data. FIG. 3 is an explanatory diagram showing an example of a Japanese sentence generation pattern. FIG. 4 is an explanatory diagram showing an example of a program generation pattern. FIG. 5 is an explanatory diagram showing an example of a program generation pattern with Japanese score annotations. FIG. 6 is a flowchart explaining the operation of the device of this embodiment. 101... Input data holding device, 102... Generation pattern extraction device, 103... Multilingual generation pattern storage device, 104... Context search device, 105... Multilingual output device.

Claims (1)

[Scope of Claims] 1. A multilingual generation device that generates and outputs a plurality of languages including Japanese sentences or a programming language from tree-structured data, comprising: a multilingual generation pattern storage means for storing a multilingual generation pattern serving as a template for a language to be generated; and a multilingual generation pattern storage means for storing a multilingual generation pattern that is a template for a language to be generated; a generation pattern extraction means for extracting a pattern corresponding to each element from the multilingual generation pattern storage means; a context search means for searching the context of the data held in the input data holding means; and an output from the context search means. and multilingual output means for outputting in a plurality of languages according to the pattern extracted by the generation pattern extraction means.