JPS6188367A - Translation system - Google Patents

Abstract

PURPOSE:To delete diversity of translation results in machine translation and to obtain precise translation results by indicating a part of speech of one word in an input sentence to be translated and translating the input sentence in accordance with the part of speech. CONSTITUTION:A keyboard 2a, which key-inputs the data related to a processing device 1a for the processing device 1a of the translation system, a memory 3a which stores the related data, a display device 4a which displays the data processed at the processing device 1a, and a translation module 5 which stores the data concerning translation are installed. At the module 5, an original, a consulting of the dictionary, sentence structure analysis, sentence structure generation, result buffers 51a-55a and a dictionary table 56a are installed. Before a translation original is inputted by an operation of the keyboard 2a, a flag, which analyzes sentence structure for an assembling of all pats of speech, is cleared. The part of speech of one word of the input sentence to be translated by the operation of the keyboard 2is indicated, and translation is executed in accordance with the part of speech and diversity of translation results is deleted.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of the Invention The present invention relates to an interactive translation method in which a machine and Iruma work together to create a correct translated sentence.

BACKGROUND TECHNOLOGY In general, machine translation is not performed through the process shown in FIG. The source text consisting of the input language to be translated needs to be analyzed during the translation process. There are three stages of analysis: morphological analysis, syntactic analysis, and semantic analysis.

Morphological analysis 1 Look up a dictionary for machine translation, extract grammatical information such as part of speech, translation information for each word, and use first person, first person, etc.
This is to analyze the tense of a number 1 sentence. Parse the syntax.

Modifications that indicate the dependent relationships between each word are used to analyze the structure of a Rope 1 sentence. Semantic analysis is the process of determining what is correct and what is not from multiple syntactic analysis results. Machine translation.

After obtaining the internal structure of the input language by analyzing it to one of the three levels, it converts it into the same level internal structure of the output language that makes up the translated sentence according to that structure, and then generates the output language. be. The accuracy of machine translation varies depending on the depth of this analysis level. It is not possible to translate sentence units by performing only morphological analysis, and the translation is limited to word units, as typified by so-called electronic translators. For those that perform syntactic analysis, all grammatically correct interpretations are output, but this results in output of multiple translation results, which increases the time and effort required for humans to determine the correct answer.

Although in principle it is possible to output only one correct translation result, it requires a huge amount of information to be stored in the machine, which is nearly impossible in reality.

Machine translation, which is the object of the present invention, is performed at least to the level of syntax analysis. In other words, it is possible to translate sentence by sentence, output a grammatically correct interpretation, and even perform semantic analysis.

Sot'L indicates machine translation that is not perfect and cannot narrow down the output to a unique one.

FIG. 8 t/'i is a block diagram showing the configuration of a conventional translation device.

A keyboard for manually inputting data in connection with the processing device BL1 2. A memory 3. in which data related to the processing device l is stored. A display device 4 for displaying the results processed by the processing device 1 and a translation module 5 in which data related to translation are stored are provided. In the translation module 5, an original text buffer 51 stores the input text. Dictionary lookup buffer 52 for storing dictionary lookup results. A syntax generation buffer 54 stores the results of converting the contents of the configuration analysis buffer 53, which stores the syntax analysis results of the input language, into the output language. It includes a result buffer 55 for storing the result of form-deaf element generation for the output language, and a table 56 consisting of dictionary one-sentence rules for machine translation.

Figure 9 #-t, when the input sentence is translated, Time fl
ies 1ike an arrow.

, the contents stored in each buffer 51 to 55 in the prior art are shown. FIG. 9(1) shows the contents of the original text buffer 51 and dictionary lookup buffer 52. Based on the words input to the original text buffer 51, the dictionary lookup buffer 52 extracts grammatical information such as part of speech and translation information corresponding to each word. @9 In the configuration analysis buffer 53 shown in Figure (2), the syntax of the input sentence is analyzed based on part-of-speech information given to the word. @9 Figure (3) shows the contents of the syntax generation buffer 54 and result buffer 55. The result of converting the contents of the parsing buffer 53 for the output language is stored in the syntactic generation buffer 54, and the translation result is stored in the result buffer 55 based on the generated f'L, North syntax.

FIG. 810 is a diagram showing a display screen by the display device 4 in the prior art. When an input sentence to be translated is input as shown in FIG. 1Oill, and a translation key provided on the keyboard 2 is operated.

Translation results as shown in FIG. 10 (2) to FIG. 1θ (6) are displayed. Each display screen Figure 10 (2) to 10th
The translated sentence shown in Figure (6) is the result of translating each word of the input sentence by recognizing it as having each part of speech as shown in Table 1 below.

These are all grammatically correct interpretations.

Of course, humans can judge that the interpretation shown in Figure 1θ (6) is correct. If the semantic analysis is complete, the translated sentence will look like the one shown in Figure 10 (6).

(-Arrows and flies don't look alike, huh? (b) Arrows have no ability to measure time.

(C) Flies never like arrows.

This knowledge needs to be stored in the machine. It is obvious that it is impossible for a machine to memorize all of this knowledge about the real world.

Therefore, it can be seen that it is inevitable that there will be solutions that should not be compatible at the conventional level of machine translation.

In this way, conventional machine translation systems have a drawback in that, since there are many solutions, it is necessary to repeatedly press the translation key until the correct answer is obtained. Furthermore, machine translation searches for all kinds of possibilities and finds all possible solutions, which increases the time it takes to get the correct answer. For this reason, a method is adopted in which the solutions are output in order, starting with those found.

Therefore, at the time when the first solution is output in Figure 10 (2), it is not known how many remaining solutions exist. Things like the number of remaining homophones in kana-kanji conversion cannot be displayed. For this reason, there is not only the disadvantage that the execution key must be pressed many times, but also the disadvantage that it is difficult to press the translation key many times to get the correct answer.

Therefore, the burden on the operator is quite large. From this point of view, it is desirable to reduce the total number of solutions by at least one by a simple operation.

In the translation results of the problems and statements that the invention is intended to solve, the translations shown in @g10 Figures (2) to (5) are caused by mistakes in determining the part of speech of multi-part speech words. I know that there is. That is, in the example of FIG. 10 (2), the part of speech is incorrectly identified as rfliesJ as a noun. Also, Figure 10 (3) and Figure 1
In the example in Figure 0 (4), ``The part of speech is incorrectly recognized as TimeJ as a particle, and the part of speech is incorrectly recognized as [flies J'' is a noun.Furthermore, @Figure 10 (+5)
In the example, the part of speech is incorrectly identified as ``s r flies J'' as a noun, and the part of speech is incorrectly identified as ``1ike J'' as a verb.

However, such errors can be easily noticed by a person who has the ability to judge the success or failure of a translation by looking at it.

An object of the present invention is to provide a translation method that eliminates the multiplicity of translation results in machine translation and allows Iruma to provide a simple instruction to obtain correct translation results.

Means for Solving the Problems The present invention is a translation method characterized by indicating the part of speech of at least one word of an input sentence to be translated, and translating the input sentence according to the specified part of speech. be.

According to the present invention, at least one part of speech of each word of the input sentence to be translated is specified for each word of the input sentence, and the input sentence is translated. , you can obtain correct translation results with simple operations.

Embodiment FIG. 1 is a block diagram showing the configuration of a translation apparatus in which the present invention can be implemented. In relation to the processing device 1a, there are a keyboard 2a for manually inputting data, a memory 3a for storing data related to the processing device 1a, and a processing device 1a.
A display device 4a for displaying the processed results and a P/R scale module 5a in which data related to translation are stored have the power provided. The translation module 5a includes an original text buffer 51a that stores input sentences, a dictionary lookup buffer 52a that stores the results of dictionary lookup, and a dictionary lookup buffer 52a that stores the results of syntactic analysis of the input language. II sentence analysis buffer 53
a h $+4 Syntax generation buffer 54 that stores the result of converting the contents of the sentence analysis buffer 53a into the output language
a, a result puff 5521 for storing the results of morpheme generation in a high-level language, a dictionary for machine translation, a table 56a consisting of grammar rules, and the like.

7 to 2 are flowcharts for understanding and clarifying the 6σ1 operation according to the present invention. Further, FIG. 3 shows an example of the display screen of the display device 4a according to the present invention. First, in step n11m, the primary candidate flag is cleared prior to inputting the translation original text. This flag is used to display an error message when no solution is obtained even after parsing all combinations of parts of speech, as will be described later. Next, when the operator operates the key 6 of the keyboard 2a, the words of the input sentence are input in step n2.

It is determined whether the translation key 7 is input to instruct translation. When inputting a word, the process moves to step n3 and the processing device fat
1a, the word is displayed by the display means 4a, and at the same time, the character code is sent from the processing device 1a to the translation module 5a. H at step n4! Translation module 5a The input character code is in source buffer 5
1a, and the dictionary in table 56a is used to perform morphological analysis. When the morphological analysis is performed and the process moves to step n5, the part of speech ranked #1 in the dictionary lookup result for the input word is displayed on the display screen. @3 figures +1
), the input word and the part of speech of the word are displayed. Proceeding to step n6, it is determined whether or not an acceptance instruction or a correction instruction has been given using key 8 or key 9 for the part of speech of the input word. If there is no instruction, the process moves to step nlo, where the part-of-speech display is deleted 71, and then the process moves to step n9, where all the dictionary lookup results are stored in the dictionary lookup buffer 52a. In step n6, when the operator issues an instruction to accept or correct the arrangement, the process moves from step n6 to step n7, and the instructed part-of-speech name is displayed on the display means 4a. In step n8, parts of speech other than those for which acceptance or correction was instructed are cut from the dictionary lookup result. The cut result is stored in the dictionary lookup buffer 52a in step n9, so the contents stored in the dictionary lookup buffer 52a in step n9 via Stella 7''n8 include only the part of speech for which the acceptance instruction or correction instruction was given. In this way, the present invention allows the operator to specify the part of speech when inputting words, while inputting the original text continues.

If there is no part of speech instruction by the operator, step n2mn3.
n4. n5. n6. There is no processing operation of nlO, n9 and n2 again, but if there is a part of speech instruction by the operator, step n2. n3. n4. n5+n6. n7+ n8. n
9, and the processing operation of step n2 is performed again. FIG. 3 (2) shows that from the state shown in FIG. 3 fil, an acceptance instruction is given to the part of speech of r Time J by operating the noun and key 8, and then the input word r
s J and its part of speech are displayed to show a state of mind. Further, from the state shown in FIG. 3 (21)
FIG. 3 (3) shows a state in which the part of speech of J was displayed as a noun, but no instruction was given by the operator, and the subsequently input word and its part of speech were displayed. While inputting the original text continues in this way, steps n2 to n9 or steps n2 to n6. The processing operations nlO and n9 are repeated.

When the input of the original text is completed and the operator issues a translation instruction by operating the translation key 7 in step n2,
Move to step nil. The third (Qi41) shows the original text in which word input has been completed and the operator has given an instruction to accept the part of speech of "Time" as a noun.

In step nil, the combination of parts of speech is initialized. Once the combination of parts of speech is initialized, syntactic analysis is performed in step n12 with reference to the parsing rules in table 56a.
The analysis result is stored in the syntax analysis buffer 53a. In step n13, it is determined whether or not the parsing was successful. If the parsing is successful, the process moves to step n14, where structural conversion is performed in the translation module 5a while referring to the conversion rules and production rules in the table 56a. Next, in step n15, syntax generation is performed.

The syntax generation buffer 54a is completed. Then, in step n16, the result buffer 55a is completed.

The result is displayed on the display means 4a in step n17. FIG. 3(5) shows the contents displayed by the display means 4a at this stage.

In step n1g, it is determined whether or not the next candidate key has been manually pressed by the operator in order to generate the next candidate when the translation result is incorrect. When the next candidate key is entered manually, the process moves to step n19, where the next candidate key 11 flag is set.

In this way, after the parsing is successful at least once and the translation result is displayed in step n12, when the person who performs the operation f' requests another translation result, the next candidate flag is set to zero. Step n1 again from step n19
Moving on to the syntactic analysis in step 2, the syntactic analysis buffer 53a stores the previous combination of parts of speech and the result of the syntactic analysis, and if the combination of 1 part of speech is unchanged from the previous time, the previous syntax %=X analysis result is searched. There is a female female. At this time, if there is no other solution for the same combination of parts of speech, the analysis will fail.
The process moves from step n13 to step n20.

At step n20, it is determined whether or not all parts of speech have been combined in the dictionary reference buffer 52a. However, at this stage of the embodiment shown in FIG. 3, all part-of-speech combinations in the dictionary lookup buffer 52a have not been completed, so the process moves to step n21, a new part-of-speech combination is set, and the syntax analysis is again performed at step n12. will be carried out. As a result, if the syntax analysis is successful, steps n13 to n14 to n17 are executed. Figure 3 (6
), 1 is displayed on the display means 4a at this stage.

The contents are shown below. Since the translation result is also incorrect, the process moves from step n18 to step n19 according to the operator's next candidate instruction, and then returns to step n12 ('1ゝ・).
Move on to sentence analysis. At this time, if there is another solution for the same combination of parts of speech, the processing operations of steps n13 to n17 are performed. If the result is an error, the process moves from step n18 to step n19 according to the operator's instruction, and the syntax analysis in step n12 is performed again. Detailed IIJ after this
I will omit the detailed explanation of the work as it would be complicated, but
Generally, if there is another solution for the same part-of-speech combination, step n 19 * n l 2 * n 13
@nl4.

n 15. n 16. A processing operation is performed at step n17, the solution is displayed, and the operator's judgment at step n18 is awaited. Unless another solution exists.

Step n 19s n 12. After n13 and no processing operation 7'', the process moves to step nlo, where it is determined whether or not all part-of-speech combinations have been completed.If not, a new part-of-speech combination is selected in step n21 with δ. Then, the process returns to step n12. If the parsing is successful with the new part-of-speech combination, the processing operations are performed again from step n13 to steps n14 to n17, the solution is displayed, and the process waits for the operator's decision in step n18. If it fails, the process moves from step n13 to step n20.

A check is made to see if the combination of all parts of speech has been completed.

As shown in Figure 3 (7), when syntactic analysis is successful for a combination of parts of speech and the resulting translated sentence is correct,
Step n18 to step n2 depending on the operator's judgment.
The process moves to step 4, where it is determined whether or not to input the first sentence.
If the next sentence is to be input, the process moves to step n1, and if the first sentence is not input, the translation ends.

When all the parts of speech have been combined in step n20, the process moves to step n22 and it is determined whether the primary candidate flag is set. If the next candidate flag is set, the process moves to step nil; if it is not set, the process moves to step n231. Steps for this occasion n 12 * n 13 en20. The processing operation is performed in a loop called n21, and the primary candidate flag 1ii remains unset.

At this time, an error message is displayed in step n23.

In the embodiment shown in Fig. 43, when inputting a word, “Time
J ka name, ;N is displayed as 6, whereas ce1? The author instructs to accept the request, and in step n8 "Time
", the part of speech other than the noun, and therefore the particle cant."
Since the dσ1 word is stored by removing the part of speech of J, the translation results obtained are shown in Figures 3 (6) to 7 (7), and the m1 table ( 31*
The solution of f41 is canted, and the key +3, 7, 8.9 operations are reduced by two times to obtain the correct translated sentence.

Similarly to Fig. 3, Fig. 4 and Fig. 5 show the display means 4a.
The display screen is not showing anything. Figure 4 (11 and Figure 5 f11 show the input word and the step n5 described above)
The name of the part of speech displayed in is displayed. Figure 4 (21 to Figure 4 (3) shows a state in which the part of speech of ``flies J'' was not displayed as a noun, but the operator instructed it to be corrected as a verb, and Figures 5 (2) to 5 Figure (3) 16 r 1i
Whereas the part of speech of ke J was displayed as a preposition.

Indicates a state in which the operator has given an instruction to accept it. @Figure 4 (4)
In FIG. 15 (4), the original text for which word input has been completed and the part of speech specified by the operator are displayed.

If you correct the part of speech of ``flies J'' in Figure 4 as a verb, the translation result obtained is shown in Figure 4 (5), and the solution is uniquely determined, which is different from that shown in the prior art. The solutions in Table 1 (2) to Table 1 (5) are cut, and the correct translation result can be obtained with the first candidate.

Also, if you specify the part of speech of rlikeJ as a preposition as shown in Figure 5, the translation result obtained will be as shown in Figure 5 (6).
~t'Lfc shown in FIG. 5 (8), and 7' (the solution in Table 1 (5) is canted and the number of key operations is reduced by one) as shown in the prior art.

In any of the cases shown in FIGS. 3 to 5, the number of translation candidates can be reduced compared to the prior art, and accuracy can be obtained quickly.

In the present invention, it is only necessary to give an instruction to accept or correct the part of speech of one word at the time of input, but of course the instruction is not limited to one word, and instructions may be given to a hake idiom that has multiple words.

For example, in the embodiment shown in FIG.
It is also possible to accept the part of speech of `` as a noun and then accept the part of speech of r 1ike J as a preposition in the state shown in Figure 3 (3). By doing this, the solutions of Table 1 (3) to Table 1 (5) shown in the prior art will be cut out. This means that key operations will be omitted three times.

Figure fE6 shows the contents stored in each pan 7a 51a to 55a in the present invention. FIG. 6 shows the contents of the +11Vi original text buffer 751a and the dictionary lookup buffer 52a. Based on the words that have been input into the original text buffer 51a, the dictionary lookup buffer 52a retrieves grammatical information such as part of speech and translation information corresponding to each word. The parsing buffer 53a shown in FIG. 6(2) then parses the input sentence based on part-of-speech information given to each word.

FIG. 1(3) shows the contents of the syntax generation buffer 54a and the result buffer 55a. Based on the parsed content, a syntax in the output language is generated in the syntax generation buffer 54a, and the content of the translated sentence is stored in the result buffer 55a based on the structure.

In the above embodiment, the parts of speech stored in the dictionary lookup buffer 52a are displayed, and the operator selects the correct part of speech from among the displayed parts of speech f'. The operator may input and instruct the part of speech of the word to be input as the ninth word for translation by operating a key. Translation is performed based on the input and instructed language authority, and this also allows for quick and accurate translation. In this simple embodiment, the part of speech does not need to be stored in the t1 pull buffer 52a, and instead of the part of speech in step n5 of FIG. t
'I don't care if the part of speech is not input by the first person.

Effects of the Invention As described above, the present invention solves the problem of the large number of translation results in machine translation, and allows one person to give a simple instruction to obtain correct translation results.

[Brief explanation of the drawing]

Figure 1 shows the configuration of the translation device. Figure @2 is a flowchart for explaining wJP\ according to the present invention, Figures 3 to 5 are diagrams showing display screens in the display means 4, and Figure @6 is a flowchart for explaining wJP\ according to the present invention.
Figure 7 is a diagram for explaining the process of machine translation. FIG. 8 is a block diagram showing the configuration of a translation device according to 1#i conventional technology, FIG. 9 is a diagram showing the contents stored in each buffer 51 to 56 in the prior art, and FIG. 10 is a display device 4 in the prior art. It is a figure showing a display screen by. 5.5a--Translation module, 51-55.51a-
55a...Buffer, 56.56a...Table agent Patent attorney Kei Nishi Partial engraving of Figure 1 drawing (no change in content) Ichino Engraving of ζ7 drawing: (
: N) ζノ ~ノ -'1 No change) G Figure 7 Figure 8 Procedural amendment 'J! (Method) February 13, 1985, Commissioner of the Japan Patent Office ・4.7゛1, Indication of case vIMI+? ? 59-21021) 4 2. Name of the invention ■ Sawa method 3. 1' with the person making the amendment! Section J Applicant Address: 22-22, Aiike-cho, Abeno-ku, Irisaka City Name (50
4,) Sharp Co., Ltd. Representative Saeki Asahi 4, Agent 1t Address Shinkoki Building Kokusai 1 “ELEX” 1-13-38 Nishihonmachi, Nishi-ku, Osaka (+525-5985 1N)
TAPT J International “AX (’;llll&Gff (
(>6) 538-o247 subject f06>538-026
3 (Representative) □ Name tr Physician (7555) Kei Nishikyo Part 5,
Date of amendment order: January 29, 1985 (shipment date) -] 6. Amendment lit.
Engraving of m + Or4 (no changes to the original). J-

Claims (1)

[Claims] A translation method characterized by specifying the part of speech of at least one word of an input sentence to be translated, and translating the input sentence according to the specified part of speech.