JPS6041166A - Machine translation system - Google Patents

Abstract

PURPOSE:To attain the machine translation of a sentence with no subject which is specific to Japanese by registering the complementary subject and its application condition, and if the sentence with no subject is detected in processing translation, by searching a storage to extract a corresponding subject and add to the sentence. CONSTITUTION:The complementary subject and a condition for applying the subject are set in the specific area of a main storage device as a variable. Japanese is read in, an idiom and a word dictionary are searched and Japanese-English translation processing based upon pattern matching is executed. The sentence with no subject refers to the specific area of the main storage device and the corresponding subject is extracted for the complement of the subject. If the translation is unsuccessful, the sentence on the way of translation is displayed. When the translation is successful, the translated sentence is displayed, and when no Japanese sentence is stored in the file, the translation is ended. Even if a context processing or deep meaning processing is not performed, the complementary subject is provided to a Japanese sentence with no subject fairly precisely in a specific field to attain precise machine translation.

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical Field of the Invention] The present invention relates to a method for machine-translating sentences without a subject, which is unique to Japanese.

[Prior art and problems]

Conventionally, there are two methods for translating sentences without a subject, which are unique to Japanese, into English: a method that uses the passive voice and a method that uses the subject of the preamble, but the former has the disadvantage of mistranslating complex sentences, and the latter has the disadvantage of having bullet points. It has the disadvantage of being mistranslated.

[Purpose of the invention]

An object of the present invention is to provide a machine translation method that can accurately machine translate sentences without a subject, which is unique to Japanese.

[Structure of the invention]

Therefore, in a machine translation system that machine translates Japanese into a foreign language, the machine translation method of the present invention is designed to prepare a plurality of pairs of subjects to be supplemented and application conditions for the subjects before performing machine translation. is registered in the storage device, and when a sentence with an omitted subject is detected during the translation process, the storage device is searched to retrieve the corresponding subject, and the retrieved subject is added to the sentence with the omitted subject. It is characterized by additional translation.

[Embodiments of the invention]

Hereinafter, the present invention will be explained with reference to the drawings.

FIG. 1 is a diagram illustrating the operation of a Japanese-English machine translation system for software reports to which the present invention is applied. Examples of software reports include the following.

Summary JSE C0PY performs copying.

1. Use the utility control statement shown below.

The opy ELECT EXCLUDE 2, C0PY control statement initiates unload, load, and compact operations. Input and output data sets are specified by the INDD and 0UTDD parameters.

3. 5ELECT control statement is...4. EXCL
The UDE control statement... When performing Japanese-to-English machine translation of a software report, the Japanese display 1 is used to perform Japanese translation and correction. The input Japanese text is stored in the Japanese text file 4. Next, in order to deal with sentences without a subject, which is unique to Japanese, subject registration is performed as described later, and then machine translation is performed using the translation system 3. The results of machine translation are stored in the English file 5. The English translated software report stored in the English file 5 is displayed on the display 2 and checked. The translation results are then distributed to software designers.

FIG. 2 shows the linguistic information packaging method. The program P consists of a machine translation program and a syntax driver, and the language information package PK consists of a word dictionary, an idiom dictionary, and a grammar dictionary.

FIG. 3 shows an outline of the translation system.

Syntax drivers and machine translation programs include word dictionaries,
Translate the translated sentence using an idiom dictionary and a grammar dictionary. As utilities, a word update utility, a phrase update utility, and a grammar update utility are provided.

FIG. 4 shows a display screen for subject registration. In the present invention, a user registers a subject for each mode in advance, and a translation system performs translation using the subject registered for each mode. Now, 1. Use the following utility control statement.

Assume that the following sentence is to be translated into English. In such a case,
On the display screen in Figure 4, for example, set the number 1 in the frame corresponding to the mode, set the number 1 in the first frame corresponding to the condition of the preamble, and set a period in the second frame. Then, set the alphabetic character SYSTEM in the frame corresponding to the subject. It goes without saying that the conditions for the preamble can be not only numbers and periods, but also regular character strings. When the translation system detects the presence of the character string "1.", it recognizes that the subject rsYsTEMJ is omitted in subsequent sentences containing it, and performs English translation by adding the subject rsYsTEMj. ``By supplementing the subject SYSTEMj, the above sentence becomes: 1. SYSTEM IS FOLLOWING U
TILITY CONTROL STATEMENT.

It is correctly translated into English.

FIG. 5 shows an example of means for subject registration. In FIG. 5, 6 to 8 are gates, 9 is a flip-flop, 1o is a subject registration key, 11 is a data register, 12 is a main memory, 13 is an arithmetic unit, 14 is a mode register, and 15 is an index section. are shown respectively. When subject registration key 10 is closed, flip-flop 9 is set and gates 6-8 are opened.

Note that when the subject registration key 10 is pressed, a screen for subject registration as shown in FIG. 4 is displayed. The mode-data signal is set into mode register 14 via gate 6. Contents of mode register 14 and index section 1
The sum of the contents of 5 indicates the address of the main memory 7 device 12. The condition data of the preamble is set in the data register 11 via the gate 7, and the subject data is set in the data register 11 via the gate 18. The contents of the data conflict register 11 are written to an address on the main memory 12 determined by the index section 15 and the mode register 4.

FIG. 6 shows the flow of machine translation processing according to the present invention. The present invention is characterized in that the subject to be supplemented and the conditions for applying the subject are set as variables in a specific area AR of the main memory at the start of machine translation of . I will briefly explain the process flow of machine translation.

■ Set the subject and condition to be supplemented in variables.

Next, do ■.

■ Set the flag to 1. Next, do ■.

■ Check whether the flag is 1 or not. If Yes, ■
If the answer is No, do ■.

■ Perform Japanese-English translation processing using pattern matching.

Next, do ■. At this time, if there is no subject, the specific area AR of the main memory is referred to, the corresponding subject is extracted, and the subject is supplemented.

■ Check whether the text has been scanned to the end.

If Yes, perform [phase], and if No, perform ■.

■ Read Japanese. Next, do ■.

■ Check whether the Japanese text is no longer in the file. If Yes, the translation is finished; if No, ■ Look up the idioms and words in a dictionary. Next, do ■.

■ Set the skianing pointer to the beginning of the sentence and set the flag to O. Next, do ■.

[Phase] Check whether the translation was successful.

If yes, do @; if no, do ■.

■ Display the sentence that is currently being translated. Next, do @.

@ Set the flag to 1. Next, do ■.

0 Display translated text. Next, perform [phase].

FIG. 7 shows an example of syntax rules used in machine translation. The meanings of the symbols in FIG. 7 are shown in FIG. 10. FIG. 8 is a diagram explaining syntax rules. As shown in Figure 8, if there is a Japanese sequence of ten verbs for noun + particle, the corresponding English sequence is ten verbs to ten nouns, and this English sequence indicates the attributes of the verb. have for example,
If the Japanese sequence ``output to file'' is converted according to the syntax rules shown in FIG. 8, it will become as shown in FIG. 9. Group 3 #X in Figure 7, VP→#X, 2 5SWI (subject 1 to be supplemented)・-, vp→L@−, S SWI
(Subject 1 to be supplemented) 2 is a syntax rule added to the new nine in order to implement the present invention.

Figure 11 shows an example of the translation conversion procedure.
FIG. 2 shows an example of a word dictionary related to FIG. 11. The input Japanese text is input using katakana characters. In this example, the Japanese sentence ``Job Gaijimeri Shuryoshita Tokimatejotaiya Loop Jiyoutai Niochiiru Kotogaal'' has been input. After a Japanese sentence is input, a dictionary lookup is performed. Next, the syntax rules ND of group 1, ND→NDW12 are applied, and "1oop 5tatus
The noun J is obtained. The entire part of speech arrangement after conversion is ND X3 VP CF @, -ND CN N
D X5P HA becomes. Next, the syntax rules of group 2 ND , ON , ND −+ND are applied, resulting in “wait 5tatus
The noun orLoop 5tatus J is obtained. The entire part of speech sequence after conversion is: ND X3 VP CF @ -ND X5 VP H
It becomes A. Next, group 3's ND, X3. The syntax rule VP-+SS is applied and [Job termina
The sentence teabnormally j is obtained, and then the syntax rules of group 3 ND , X5 , VP −+vp are applied, and [mayfall to w
aitStatus or 1oop 5status
A sentence element with the verb attribute J is obtained. Next, group 3 @-, vp-+-@-s swt (complementary subject) 2
The syntax rule is applied, [S3'stem ma)
'fall to wait 5tatus or 1
The sentence oop 5tatu+s J is obtained, and finally the syntax rules of group 3 SS, CF, @ -, ss-+55wt, 4 are applied, and [When job termina
teabnormalLsystem may fa
ll to walt 5tatus orloop
The sentence st4%ya J is obtained.

FIG. 13 shows the format of the LOAD SUB command for reading the subject from the specific area AR (FIG. 6). This LOAD SUB command was newly established for the purpose of moving a set of data that fits the mode, the preamble condition, and the subject to be supplemented from a specific area AR that is invisible to the program to an area that is visible from the program. The LOAD SUB command is
It has an opcode section B2FE, a base register section, and a displacement section.

FIG. 14 is a diagram for explaining the LOAD SUB command. The first word (first word, 32-bit length) specified by the second operand of the LOAD SUB command contains the start address when moving data in the specific area AR, and the remaining word (second word, length) 32 pit length), the size of the specific area length is written.

Bits 0 to 7 of the first word are ignored, bit 0 of the second word
to 7 are padding characters (specific area AR if the specific area length that can be written for filling in the blanks is longer than the actual AH area length)
After all of the data has been moved, the remaining bytes are filled with the padding characters described above, and the condition code becomes 2''.

If the specific area AR does not exist, data will not be moved and the condition code will be 13". At this time, it is unknown whether an access exception will occur or not. If the specific area length specified by the second operand is If it is shorter than the actual AR area length, it wraps around and the remaining data is moved from the start address, and the condition code becomes '1'. This command is
This is an interruptible instruction, and if an interrupt occurs at the end of a unit of operation, the number of bytes in the second @ specified by the second operand is updated by the number of bytes processed. Condition codes are unpredictable. When the instruction is resumed, it is re-executed at the point it was interrupted. The condition code is 0, 1.2.3 is true, 0 means normal completion, 1 means the specific area length specified by the second operand is shorter than the actual AH area length, 2 means the second operand. The designated specific area length is longer than the actual AH area length, and 3 indicates that the specific area AR does not exist. Program exceptions include designated exceptions and program exceptions (storage to second operand, read from specific area AR).

Next, we will explain the LOAD SUB instruction using a concrete example.
01 The content of address 1034 (first word) in main memory is X '1
1001000 v, main memory (7) address 1035 (second
Assume that the content of the word) is Xma22000100'. This LOAD SUB instruction uses general-purpose register 12 as a pace register for the second operand, and sets its displacement to 0OA (10 in decimal), so the second operand starts at address 1024+10, that is, address 1034. Bits 8 to 31 of the first word starting at address 1034
From this, it can be seen that the movement destination of the specific area AR starts from address 4096. X'22' in bits 0 to 7 of the second word starting at address 1035 are padding characters, and bits 8 to 31 indicate that the size of the specific area is 256.

FIG. 15 shows an example of the data structure of the specific area AR. In the same figure, rM:01J has a mode number of 1.
It shows that. Condition 1 of the preamble is that (1) exists, and condition 1 of the preamble is "...". When all of the conditionals in the sentence before mode l are satisfied, the subject rTHESYSTEMJ is added. Also,
When all conditions 1 to m of the preamble of mode 2 are satisfied, the subject rWEJ is added.

FIG. 16 shows the contents of the main storage area after the LOAD SUB command is completed. The specific area AR has a data structure as shown in FIG. 15, and the destination area is from address 4096 to address 4351. The content of address 4096 is X
V61D47AFOV-f? Aruka, xvelv is EB
CDIC:I-)"T/T available, XvD4vhaEBCD
The IC code is M, and the EBCDIC code is 9. XYFO is 0.

Note that before the LOAD SUB command is executed, the area from address 4096 to address 4351 is r4040...40.
contains blank data.

〔Effect of the invention〕

As is clear from the above explanation, according to the present invention, it is possible in certain fields to accurately supplement the subject of Japanese sentences without a subject, even without context processing or deep semantic processing. .

[Brief explanation of drawings]

Figure 1 is a diagram explaining the operation of the Japanese-English machine translation system for software reports to which the present invention is applied, Figure 2 is a diagram showing the language package method, Figure 3 is a diagram showing the outline of the translation system, FIG. 4 is a diagram showing a display screen for subject registration, FIG. 5 is a diagram showing an example of means for subject registration, and FIG. 6 is a diagram showing the flow of machine translation processing of the present invention.
Figure 7 is a diagram showing an example of syntax rules that can be used in machine translation, Figure 8 is a diagram explaining syntax rules, Figure 9 is a diagram showing a specific example of conversion based on syntax rules, and Figure 10 is a diagram showing a concrete example of conversion based on syntax rules. A diagram explaining the meaning of the symbols in Figure 7, Figure 11 is a diagram showing an example of the conversion procedure,
FIG. 12 is a diagram showing an example of a word dictionary related to FIG. 11,
Figure 13 is a diagram showing the format of the LOADSUB command, Figure 14 is a diagram showing the format of the LOADSUB command.
Figure ti is a diagram explaining the LOADSUB command, Figure 15 is a diagram showing an example of the data structure of the specific area AR, and Figure 16 is a diagram explaining the LOADSUB command.
FIG. 3 is a diagram showing the contents of the main storage area after the OADSUB instruction ends. 1...Japanese display, 2...Display,
3...Translation system, 4...Japanese file, 5...
・English file, 6 to 8...Gate, 9...Flip-flop, 10...Subject registration key, 11...
・Data register, 12...Main storage device, 13...
・Arithmetic unit, 14...Mode write register, 15...
Index section. Patent Applicant: Fujitsu Ltd. Patent Attorney Kyo Tani 4 Part Z Figure 3 Figure Fang Wa Figure 8 Figure IZ Piece 13 Figure O 2 O Duke Rat' Yu 74 Figure

Claims (1)

[Claims] In a machine translation system that machine translates Japanese into a foreign language, before machine translation, the subject to be supplemented,
Multiple sets of subject application conditions are registered in a storage device, and when a sentence in which the subject is omitted is detected during the translation process, the storage device is searched and the corresponding subject is retrieved. A machine translation method that performs translation by adding a given subject to sentences where the subject is omitted.