JPH03147179A - Machine translation device - Google Patents

Abstract

PURPOSE:To apply the same character attribute to a translated sentence by succeeding character attribute data in a word processing document, which is a translating original, as information to accompany respective morphemes during translation. CONSTITUTION:Data forms in a translation input file 4 and a translation output file 6 are made independent and while preparing exclusive input and output conversion programs 31 and 71 exclusive for a word processor to be used, the various word processors are used for input and output documents. Namely, as a word processor 1 for source sentence, the word processor optimum for a translation input language can be selected and as a word processor 9 for translated sentence, the word processor optimum for a translation output language can be selected. The character attribute data of the original inputted from the word processor 1 for source sentence are succeeded as the information to accompany the respective morphemes while executing a translation procedure. Thus, the same character attribute can be applied to the translated sentence in respect to the original, to which the character attribute is applied by the word processor 1 for source sentence, and outputted to the word processor 9 for translated sentence.

Description

TECHNICAL FIELD The present invention relates to a machine translation device, and more particularly to a machine translation device that translates an input document created by a word processor capable of giving display attributes to characters.

Prior Art Conventionally, in a machine-to-machine 1υ system, the translation section only translates the sentence ¥r (including sentences, names 31 and 1), so the input data to the translation section is the normal characters that make up the sentence. It consisted only of codes of and symbols.

Therefore, when translating a manuscript that contains function codes and character attribute codes created with a word processor (hereinafter referred to as word processor), only the character codes are extracted from the manuscript as a preparatory step, and input data is created. .

In such conventional machine translation systems, the character attributes that are transferred to the manuscript using a word processor are unmarked at the time of translation, so double-width characters and underlines in the input manuscript are not reflected in the translation result at all, and the input The disadvantage is that the character attributes of the word are given in the same way even when the translated word is illuminated, so after translation, it is necessary to re-light the translated text in a word processor and manually assign the character attributes and add corrections. .

Purpose of the Invention The present invention has been made to eliminate the drawbacks of the conventional ones as described above. We propose a machine translation device that can add the same character attributes to the translated text of a given manuscript and output it.

Structure of the Invention A machine translation device according to the present invention is a machine translation device for translating a manuscript created by a word processor for original text and consisting of character information and character attribute information indicating display attributes of the character information, the machine translation device comprising: a generating means for extracting the character information and the character attribute information from the manuscript to create translation input data; a generating means for generating a morpheme by solving the character ft'1 information; a translation means for creating translation output data from the generated morpheme and the character attribute information added to the morpheme; and converting the translation output data created by the translation means into a document data format of a word processor for translation. and a conversion means for creating translation data in which the character attribute information is reflected.

Embodiment Next, an embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing the structure of an embodiment of the present invention, and FIG. 2 is a diagram showing the operation of the embodiment of the present invention. In these figures, a manuscript input by an original text word processor (hereinafter referred to as an original text word processor) 1 is stored in an original text file 2.

Here, the word processor 1 for original text is composed of word processor software (hereinafter referred to as word processor software) 11, a keyboard 12, a CRT 13, and a printer 14.

The manuscript stored in the original text file 2 is converted into translation input data by an input conversion program 3 in an input conversion section 3, and sent to a translation input file 4.

That is, character data and character attribute data are extracted from the manuscript in the original text file 2 by the input conversion program 31, and stored in the character data section 4] and the character attribute data section 42 in the translation input file 4, respectively.

Translation input data consisting of character data and character attribute data is translated by a translation unit 5 and sent to a translation output file 6 as translation output data.

In other words, the character data in the character data section 41 of the translation input file 4 is in the form! ! ! The character data is analyzed by the elementary analysis unit 51, and the character data is segmented and the part of speech is recognized. Each segmented element (hereinafter referred to as a morpheme) is stored in the work file 155.

The character attribute data in the character attribute data section 42 of the translation input file 4 is added to the δ morpheme stored in the work file 155 by the character attribute expansion program 52, and is stored in the work file 1156.

The data stored in the work file 1156, that is, the three morphemes to which character attribute data have been added, are multiplied by χ・1 and parsed ~ Form! In the g3g generation unit 53, a translation procedure called 1M sentence analysis (tl - semantic analysis, style selection, syntactic generation, form!!! elementary generation) is performed and stored in the work file m57. This translation procedure is performed. Meanwhile, the character attribute data assigned to the convex hanging element is held.

The data stored in the work file m57, that is, the translated δ morpheme and the character attribute data of each of these morphemes are divided by the character attribute expansion program 54 and are respectively divided into the character data part 61 and the character attribute data of the translation output file 6. The information is stored in the section 62.

Translation output data, which is stored in a translation output file 6 and consists of character data and character attribute data, is synthesized by an output conversion program 71 in an output conversion unit 7, and converted into a document data format by a translation word processor (hereinafter referred to as translation word processor) 9. The translated data is converted into translated text data and sent to the translated text file 8.

The translation data stored in the translation file 8 is sent to a translation word processor 9 for display or printing. here,
The translation word processor 9 includes word processing software 91, a keyboard 92, a CRT 93, and a printer 94.

FIG. 3 is a diagram showing a specific example translated according to an embodiment of the present invention. FIG. 3(a) shows the manuscript created by the original text word processor 1, FIG. 3(b) shows the manuscript in the original text file 2, and FIG. 3(C) shows the character data part 41 and the translation input file 4. FIG. 3(d) shows the character data and character attribute data in the character attribute data section 42 in the work file 1.
Figure 3(e) shows each morpheme in work file 1.
55, each morpheme to which character attribute data has been added, Igl (f) represents each translated morpheme in the work file Ill57, and FIG. The character data and character attribute data in the character attribute data section 62 are shown in Figure (11).
3 shows the translation data in the translation file 8, and FIG. 3(i) shows the translation displayed on the translation word processor.

FIG. 4 is a diagram showing an example of the Sentence 8 data format.

In the figure, sentence 8 data formats A and B each have different character attributes, but the displayed contents are the same.

The operation of one embodiment of the present invention will be explained using these FIGS. 1 to 4.

First, when a sentence as shown in FIG. 3(a) is input into the original text word processor 1, this manuscript is stored in the original text file 2. This original text file 2 is a normal word processing document file, and the above text data displayed on the CRT 1.3 of the original text word processor 1 or the printer 14 includes character data, double-width codes 2a to 2C, and underline codes. The character attribute data of 2d to 2[ are mixed rE L[
See FIGS. 3(a) and (b) jlα]. However, in this embodiment, data in document data format A is inputted by the word processor 1 for original text. (See Figure 4).

Next, when the input conversion program 31 is started in the input conversion section 3, the input conversion program 31 inputs the original text from the original text file 2, and inputs the character data into the character data section 41 of the A key part cuff file 4. The attribute data is converted into a type and a byte position in the character data section 41 and output to the character/attribute section 42 of the translation section 4.

In other words, the original text in source file 2 contains the character data "The cat eats fish.", "Type double-width, start byte position 4 bytes [1, end byte position 9 PitroJ", and "type Angline, start byte position]"21<th item,
End byte position 17th byte" is converted into character attribute data. Ru. The first draft reading by the translation unit 5 is a process performed by the morphological analysis unit 51 as in a normal translation system.

Only the data of the form-deaf search solution [the character data section 4 in the translation input file 4 in the hill section 51] is input, and the same 121 word cut and part-of-speech recognition as in the normal translation system are completed.

In other words, in the character data section 41 of the translation input file 4,
Eating catfish. ” Word I in light of the character data
When JJ and recognition of parts of speech go together, "cat" (noun), "
``Ka'' (particle), ``Sakana'' (noun) ``wo'' (particle), ``eat'' (verb), ``ru'' (end) 1 ``.''
It is divided into each morpheme (period). This prefecture is sent to the work file +55 used during translation [
See Figure 3(d)].

, bt+, When the character attribute assigning program 52 is activated by Sawabe 5, the character attribute CI'''p program 52 creates a work file 155 which is the output of the morpheme M ln unit 51.
Input the data of Il+ and the character attribute data of the character attribute pig part 42 in Sho, Station Café Aire 4, and enter the character f
iz f Raw data is Lj for 6 morphemes.

That is, 1 [3 elements 11'+
! ) In the IF output data, in the character data, 0 to 1 byte II "draw", 2 to '3 byte [ ] "ga", 4 to 9?
Ku・ito 11's "fish j, 1" (1-11 bytes "1's"wo" 12-15 ba (1-Ll's "eat", 16-17
"ru" in height 11, "." in byte 11 from 18 to 1 is fire i7--form deaf element [see Figure 3(d) Q(j
l.

On the other hand, in the character attribute data in the character attribute data section 42 of the translation input file 4, 4 to 9 bytes [ ] are double-width,
Since the 12th to 17th bytes are unlined,
4 to 9 bytes ``Data that the morpheme ``fish'' that occupies the 1 position of 1 is a double angle! j, eh, we can get the data that "tabe" in the 12th to 15th bytes and "ru" in the 16th to 17th bytes have an angline. This result is sent to the work file [156] [see Figure 3 (e)]
.

In addition, the translation procedure in the syntax analysis) 11 to the morpheme generation unit 53 is the same as in a normal translation system, and the work file U5
6 is input, and translation proceeds in the order of 18 sentence analysis - semantic analysis - style selection - syntax generation - morpheme generation.

During this time, the character attribute data assigned to each morpheme by the character attribute assignment program 52 is retained throughout the translation procedure.

Therefore, (1) each morpheme after translation generated by the sentence analysis-morpheme generation unit 53 is rAJ, "△" fca
tJ, r△J, reatsJ, “△J.

rfishJ, r, J, and the character attribute data that says "eat sJ has an angle line, rf
Character attribute data indicating double-width is added to tshJ. This result is sent to the work file llll57 [3rd
See figure (f)].

As the final procedure of the translation section 5, the character attribute expansion program 5
4 is started, the character attribute development program 54 inputs the data of the work file 1157, which is the output of the translation procedure by the syntax analysis to morpheme generation section 53, and stores each morpheme in the character data section 61 of the translation output file 6. , the character attribute data that each morpheme has is f! By rI and character data section 61
The data is expanded into byte positions within the translation output file 6 and output to the character attribute data section 62 in the translation output file 6.

In other words, in the data of work file ■57, the morpheme r
eatsJ changed the character attributes of Angline to [fishJ
has the character attribute of double-width, so each morpheme is sent to the character data section 61, and the character attribute types "angline" and "double-width" and the shape part X (reat
sJ, rfishJ or the byte position occupied in the character data section 61 "Start hide position 6th byte, end byte position 9"
byte" and "start byte position 11th byte, end byte position 9th byte" are combined and sent to the character attribute data section 62 [see FIG. 3(g)].

Next, when the output conversion program 71 is started in the output conversion unit 7, the output conversion program 71 converts the translation output file 6
inputs the data, synthesizes the character data in the character data section 61 and the character attribute data in the character attribute data section 62, converts it into data in document data format B of the translation word processor 9, and outputs it to the translation file 8 ( (See Figure 4).

That is, use a model of the translation word processor 9 that has character attributes in an on/off format, and convert the data in the translation output file 6 into data in document data format B.
See Figure 3(h)].

However, the Angline On Code 8a is a control code that outputs subsequent characters with an Angline, and the Enlarged On Code 8C is a control code that outputs subsequent characters in double-width.

Also, Angline Off Code 8b and Expanded Off Code 8
d is a code for canceling those control codes, and the remaining codes are normal character codes.

When this translation is displayed on the CRT 93 of the translation word processor 9 or the printer 94, it becomes as shown in FIG. 3(1).

As shown in the example above, the way character attributes are held in the word processor Buncho differs depending on the word processor model, and as shown in Figure 4, the internal data differs depending on the same display by χ・1. There are many things. In this embodiment, the data format in the translation input file 4 and the translation output file 6 is changed to the word processor model (not tZ 11, but an independent It is said that

In addition, input conversion program 3 dedicated to the word processor used
1 and an output conversion program 71, the &FI word processor can be used for input documents and output documents.

In other words, the word processor most suitable for the translation input language can be selected as the original text word processor 1, and the word processor most suitable for the translation output eight words can be selected as the translation word processor 9. Of course, the same word processor can be used for both the original text and the translated text.

Furthermore, when changing the word processor to a different model, if you replace either the input conversion program 31 or the output conversion program 71, which are word processor dependent parts, with those for the new word processor,
The translation program can be used without modification.

In this way, by inheriting the character attribute data of the manuscript input from the original word processor 1 as information accompanying the morphemes on both sides of the translation procedure, the character attributes can be changed in the original word processor 1. The same character attributes can be added to the translated text that is i4 applied to the original document, and then output to the word processor 9 for translated text.

DETAILED DESCRIPTION OF THE INVENTION According to the present invention, character attribute data in a word processing document, which is a translation manuscript, is inherited as information accompanying each morpheme during translation, so that character attribute data can be transferred on the input side. This has the effect that the same character attributes can be added to the translated text of the obtained manuscript and output.

[Brief explanation of the drawing]

Fig. 1 is a block diagram showing the configuration of an embodiment of the present invention, Fig. 2 is a diagram showing the operation of an embodiment of the invention, and Fig. 3 is a concrete example translated according to the embodiment of the invention. The figure shown in FIG. 4 is a diagram showing an example of the document data format. Explanation of symbols of main parts 1... Word processor for original text 2...
Original text file 3... Input conversion section 4...
... Translation input file 5 ... Translation section 7 ... Output conversion section 6 ... Translation output file 8 ... Translation file 9 ... Translation word processor 51...
・Form tension part 11

Claims (1)

[Claims] (1) A machine translation device that translates a manuscript created by a word processor for original text and consisting of character information and character attribute information indicating a display attribute of the character information, the machine translation device translating the character information and the a generating means for generating translation input data by extracting character attribute information; a generating means for generating morphemes by analyzing the character information; a translation means for creating translation output data from the character attribute information; and a translation means for converting the translation output data created by the translation means into a document data format of a word processor for translation, and converting the translation data to which the character attribute information is reflected. A machine translation device comprising: a conversion means for creating a machine translation device.