JPH0194480A - Mechanical translating system - Google Patents

Abstract

PURPOSE:To obtain a translation with high accuracy and naturality by incorporating a syntactic and conceptional different point and its relation as knowledge, using the knowledge, and performing a translation-processing while recognizing an appropriate scope as one unit of the translation-processing according to the nature of an original. CONSTITUTION:The title system appropreately recognizes the one cycle unit of the translation-processing from the sentence column of the original stored in an original storing part 2 by the contents of a translation dictionary part 6 to accommodate necessary information for the translation-processing, fetches the different point of the grammatical organization between a first language and a second language, and the conceptional different point as the knowledge, sufficiently utilizes them, automatically adjusts the scope of the sentence to be the unit of the translation-processing according to the nature of the original, and performs the translation-processing. Thus, the translation with the high accuracy and naturality can be obtained.

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Industrial Application) The present invention relates to a machine translation system that can obtain translations with high accuracy and naturalness.

(Prior art) In recent years, a system that automatically machine translates an input original text using a converter has been attracting attention. It cuts out the details of the processing cycle, performs morphological and syntactic analysis on each unit, divides it into predetermined processing units such as words (phrases), and searches a translation dictionary for each of the 41 processing positions. It is configured to find corresponding translated words (translated phrases), etc., and combine them according to predetermined translation rules to obtain the translated text.

However, for example, English and Japanese! /) In translation between languages with large syntactic and conceptual differences, there is a limit to using a single sentence as the unit of processing. It is highly questionable whether there is enough information to construct a "sentence" as defined in the grammatical system, and conceptually, the concept of multiple sentences in the first language is expressed in a single sentence in the second language, and In reality, there are many cases where this is natural.

(Problem to be solved by the invention) As described above, in conventional machine translation systems where one sentence is the unit of processing, this framework poses a major obstacle to the goal of obtaining high accuracy and naturalness in the translated text. was. Therefore, the present invention aims to address the differences in the grammatical system of the first language and the second language.
The purpose of this invention is to provide a machine translation system that can obtain translations with high level accuracy and naturalness by looking at differences in ideas as knowledge of the translation department and making full use of them.

[Himei's composition]

(Means for Solving the Problems) The present invention comprises: an input unit for inputting an original text to be translated into D; an original text storage unit for storing the original text input from the input unit;
A translation dictionary unit that stores information necessary for translation processing, and the contents of this translation dictionary unit are used to appropriately recognize one cycle unit of translation processing from the sentence string of the original text stored in the original text storage unit, and perform translation processing. a translation section that executes the translation, a translation storage section that stores translation candidates obtained from the translation section i, a display section that displays the original text and translation candidates, and an I! +l
and a translation editing section that changes the translation candidates displayed on the display section according to the operation number in the input text section, and controls the translation process according to the nature of the original text. It is characterized by automatically adjusting the range of sentences (i.e., the number of phrases/clauses in a sentence) and performing translation processing.

(Function) The present invention incorporates the syntactic and conceptual differences between the Dou language and the 82 languages and their related features into the % translation dictionary section, and uses this knowledge to interpret the input original sentence string. By performing the translation process while recognizing the appropriate range of information that has an appropriate amount of information according to the nature of the original text, a highly accurate and natural translation can be obtained; It is something to do.

(Embodiment) FIG. 1 is an overall block diagram of a Japanese-English machine translation system as an embodiment of the present invention. This system consists of an input section l consisting of a keyboard, an original text storage section 2 that stores the English text entered manually as the original text. A translation storage unit 3 that stores an English sentence corresponding to the input Japanese sentence as a translation, and an editing control unit 4 that controls the entire system. A translation unit 5 that executes translation processing, a translation dictionary IF 6 that stores knowledge information used in translation processing. A display control section 7% for controlling the display of the original text, translated text, etc.; and a printing section 9 for transcribing the original text, translated text, etc. and outputting it as a hard copy. Original text paragraph position storage unit that extracts and stores the original text for each paragraph. Analysis data storage unit 11 that stores one piece of analyzed data during translation processing.
Tonyo PG has been created.

FIG. 2 is a diagram showing an example of the key arrangement of the input section l.

The input section 1 is equipped with the following keys in addition to the character keys for English writing.

Translation instruction keys: 100 Edit keys: 101-106 Function keys: 110-117 Cursor keys: 130-133 Other keys FIG. The input original text is displayed on the left side of the screen. The translated text obtained as a result of the translation process is displayed in the translated text display area on the right side of the screen as well as in a position corresponding to the original text. Additionally, the editing area at the top of the screen can be used to display information necessary for various types of editing.

Figure 4 shows 1% of the flow of dialogue translation processing by the editing control unit 4.
It is a flowchart showing Jt. In the interactive translation process, the editing control unit 4, which allows the operator to input the original text and edit the corresponding translated text, performs steps 81 to S.
4, it monitors whether there is any key input from the input unit 1 or whether a translation completion signal is received from the translation unit 5. conduct.

When the operator operates any of the function keys, this is detected in step S4, and a process corresponding to that function key is performed in step S6.

When the operator operates any of the editing keys, this is detected in step 831ζ, and the 4th one is detected in step S7.
4Ii & Performs processing corresponding to the key.

When the operator operates the cursor key 130 or other control keys, processing for kissing the cursor corresponding to the key operation and other processing are performed in step S5 via steps 81 to S4.

When Kazuya operates the character keys and inputs the original text, each character key is detected in step S2.

The corresponding character codes are sequentially set in the original text storage section (
Step S8) When the operator presses the translation instruction key 100 at any time, for example, when the human effort of the - sentence is completed, this key is detected in step S1, and the FIF4 collection control unit 4 causes the translation unit 5 to store the original text. supply the original text within the department;
Instruct the start of translation processing (step S9), na$P,
If input editing such as corrections, insertions, deletions, etc. is required during the original writing process.

After moving the cursor to a desired editing location using the cursor keys, this can be done using editing keys such as the insert key 101% and the delete key 102.

When the editing control section 4 detects a signal indicating the completion of translation for one processing unit, it sends the translation candidates and auxiliary information obtained from the translation section 5 to the display-side phloem 7 to display the translated text.

The display control unit 7 displays the translation candidate in the translation display area at a position corresponding to the input original text (on the left side of the screen), and if there is auxiliary information, displays the translated word with high brightness or inverted so that the operator can We are trying to make it possible to identify this fact.

In this situation, when the edit key is pressed, the edit control unit 4
detects this in step S3, and performs editing processing corresponding to each key on the word or phrase indicated by the cursor at this time (step S7).

For example, by operating the insert key 101, a character is inserted before the cursor position.

When the delete key 102 is operated, the character string in the range indicated by the cursor is deleted.

By operating the movement key 103, the range indicated by the cursor is moved.

By operating the time erase key 104, the above keys 101 and 102 are activated.
．． Nullifies the effect of 103.

Modifications can be made using the auxiliary information by operating the key 105.
Other dependencies of the word/phrase pointed to by the cursor display candidates.

Note that the following functions can be realized using one function key.

By operating the translated word display key 110, the translated word can be displayed for H+ζ in the translated sentence.

Resign! By operating the display key 111, a dictionary can be displayed for the word V% in the original text.

New words and phrases can be registered by operating the dictionary 1 registration key 112.

By operating the dictionary deletion key 113, words and phrases registered through dictionary registration can be deleted.

By operating the partial translation key 114, when translation fails, the partial translation is displayed.

The cursor keys include a cursor movement key 130 to move the cursor in each direction, a position change key 131 to change the prefecture to which the cursor moves, and an inter-area movement key 132 to move the cursor between display top areas.
It consists of enlargement/reduction keys, etc. for reducing the size of the cursor to the size of a character or enlarging it in units of words.

Therefore, the operator can interactively obtain a translated text by inputting the original text and editing the translated text as appropriate.

FIG. 5 is a flowchart showing the flow of translation processing in the translation section 5.

One paragraph in one sentence is taken as an example, and the translation is shown until all prefectural translations are completed.

First, in step Sol, -/clagraph is g! , and store it in the original paragraph storage unit.

In step 02, it is checked whether there is any untranslated sentence in the original paragraph storage.

If there are any remaining, proceed to step 03; otherwise, proceed to step 824.

In step 03, the untranslated sentences in the original paragraph storage section are set in an input buffer (not shown) in the editing control section 4. Next, in the morphological analysis section, the above-mentioned cover V
For the original text in the file, threatening words with changes in the ending, etc. are converted into their original forms (basic forms) using the conjugative inflection '496 a. (Step 04).

The dictionary search unit searches for the abbreviation/idiom dictionary f(6b) for each word that makes up the input original sentence, and obtains information such as its part of speech and translation (step 05). Based on the part-of-speech information for each word, a likely combination of parts of speech is extracted from among the many part-of-speech candidates, and the candidates are sent as input data to the syntactic analysis unit (step 06).

The parsing unit analyzes the syntax using parsing grammar 6C.
Generates the structure of Japanese sentences. The threats whose analysis has failed are backtracked to the analysis control section (step 5O7).

If the analysis is successful in step SO7, the processing prefectural recognition grammar (6d) is used to check whether the data of the analysis result satisfies the conditions for being divided and performing the subsequent translation processing (step SO7). 5O8), if it is satisfied, the process moves to step 817; if it is not satisfied, the process moves to step SO9.

In step SO9, it is checked whether the analysis data of the preceding sentence remains in the analysis data storage section.

If there are any remaining, the process proceeds to step 810, and if there are none, the process proceeds to step 813.

In step S10, whether or not the following translation process should be performed by combining the main analysis data and the previous data in the analysis data storage section is checked using the processing prefectural recognition grammar (6d). 0 It is determined that it should be performed. If so, the two data are combined (step 515).

Proceed to step 813; if it is determined that there is no need for combination, proceed to step S11.

In step 811, the previous data in the analysis data storage section is sent to the structure conversion translation generation process, and the analysis data storage section is emptied.

The flow of the structural conversion/translation generation process (step 512) is shown in FIG. The structure conversion unit (step 5121) converts the Japanese structure into the English structure using the conversion grammar (6e). The W sentence generation unit (Stella 7”5122) determines the English syntax from the English !S structure according to the generative grammar (6f) and converts it into a Rin word string.The morpheme generation unit (Step 81
23) completes the translated sentence by changing the word form of V, abbreviation, etc. using the morpheme generation grammar (6g). When the translated text is completed, the scope information of the translated text and the corresponding original text is sent to the display control unit 7 through the editing side L4, and the original text is redisplayed alongside the translated text (step 8124).

I will tell you about this.

Again in Figure 5, after the translation process of the previous analysis data is completed in step 812, it is subsequently determined in step S13 whether or not it is necessary to store the two pieces of data in the analysis storage section and attempt to combine them with the next data. do.

At this time as well, the processing *tX recognition grammar (6d) is used. If it is determined that it is necessary, the main data is stored in the analysis data storage section (step S16), and the process returns to step S16).
Returning to O2, processing of the next statement is attempted. If it is determined that it is not necessary, the main data is sent to the structure conversion/translation generation process and a remaining translation process is performed (step 814), after which the process returns to step S02.

When it is determined in step S08 that it is necessary to divide one sentence, in step 817, as in step 09, it is checked whether the previous data remains in the analysis data storage section.

If there are any remaining pieces of data, the remaining translation processing is continued on the previous data in the same way as in steps 811 and 812, and the translated text is completed (step S22 and 823).If there are no remaining pieces of data, the process directly proceeds to step 81B, where processing prefectural recognition is performed. Perform prefectural division of this analysis data using grammar (6d).

Subsequently, in steps 819 to 821, the remaining translation processing is continued for each divided standalone to complete each translated sentence. Step 5 when all units have been translated
O2-Return, processing of next sentence is attempted.

If it is determined in step S02 that there are no unprocessed sentences remaining in the original sentence paragraph storage, it is checked in step 824 whether or not the analysis data storage is empty, and if the previous data remains, it is Continue the remaining translation processing to complete the translation, and if there are no remaining translations, the process ends.

Note that when inputting the original text using a keyboard in interactive translation, the text up to the point where the translation key is pressed is treated as a paragraph.

Figure 7 shows a simple example embodying the processing position recognition grammar.This grammar mainly consists of division rules for dividing a sentence and combination rules for combining multiple sentences. In general, conversion rules are listed according to a certain rule, and the application is tried in order starting from the first one.As for combination rules, if one applicable rule is met, the application of the following rule is not judged. On the other hand, regarding reduction rules, after a certain rule is applied, there are cases in which the process progresses to the first rule and cases in which it does not proceed. This difference is at the beginning of the rule.
They are distinguished by ``ku'' and ``1〉''. Here, 1ku” indicates that you will not proceed to the next rule after applying your rule, and 1〉
1 indicates proceed.

Next, I will give some examples of how to apply this rule.

For example, the division rule ■ can be applied to the analysis result of the sentence ■ "She is in the kitchen. Where is Tom?" Figure 8 shows the state of the data before and after the application of rule ■ as a tree. The meaning of ■ is, “If a declarative sentence and an interrogative sentence are connected in order, then separate the declarative sentence and interrogative sentence separately.”
Where is Tom? ” is generated.

The combination rule ■ consists of a question and its answer, like a dialogue sentence 1
It can be applied to the results of parsing a set of sentences. Using this rule, a question sentence and a response sentence can be combined to identify omitted phrases in the response sentence. Figure 9 shows how the rule is applied to the pair ``Did he go to school?'' and ``Yes, he went.'' It is omitted from the response sentence as shown in the figure after application. The identified phrases can also be easily identified by combining them with the question sentence. This combined data undergoes a subsequent structural conversion process and is subject to generation rule restrictions during the generation process, such as @Did he go t.

5chool? The translated text Yea+, he dld,' is output.

For example, the combination rule ■ is applied to the following two sentence analysis results. ■``Study hard!'' ■``If you don't, you won't get into university.'' The purpose of this rule was to avoid the fact that as long as the sentences were translated separately, the resulting translation would be more than just the sentences. By combining, it is possible to translate more than a single sentence or syntax according to the grammar of two languages. For example, the result of applying 0 is @5tudy hard,
Other wise you vrill fa
ill to enter any unknown
y,” @5tudyhard, and you ca
n enter a university,”, @
5tudyhard so that you may
enter a university,'.

”If you study hard, you a
an enter a university alty,”
The translated text will be output.

Figure 410 and Figure 11 show how the translated sentences corresponding to the input sentences ■, 0 and ■ are displayed and at the same time the original sentences are re-displayed. <12%1> and (12,2) are reasserted in accordance with < 1 again
The number 5-16> is displayed.

As described above, this device performs translation while recognizing the differences between the first language and the second language and the characteristics of each, so it is possible to obtain high-quality translations.

Note that the present invention is not limited to the implementation sequence described above; for example, a rule for recognizing the processing unit (fL) is incorporated into the parsing grammar - all υ data of the paragraph is placed in the syntactic analysis section. As you progress, you may perform the Yume-i-kiri 9 extraction, in other words, recognize and analyze the units at the same time.If you use this method, the translation will be completed without having to display the original text again to you. This means that it can remain displayed at the same time as i-recognition.

It is to be understood that the present invention can be implemented with various modifications without departing from the spirit thereof.

[Fanming effect]

As described above, according to the present invention, an i-sentence document can be converted into a grammatical feature of the second language within the grammatical framework of the second language, without being limited to the grammatical framework of the first language. It is possible to translate documents into a second language while incorporating them. Therefore, the resulting translation will be of high quality.

[Brief explanation of the drawing]

11A is an overall block diagram of one embodiment of the present invention, FIG. 2 is a diagram showing the key arrangement of the input section used in one embodiment of the present invention, and FIG. 3 is a diagram of the display section used in one embodiment of the present invention. A diagram showing an example of a screen layout, FIG. 4 is a flowchart showing a sequence of processing steps of an editing control unit used in an embodiment of the present invention, and FIG. 5 is a flowchart of a translation process used in an embodiment of the present invention. FIG. 6 is a flowchart showing the flow from the structure conversion process to the translated text display process during translation processing used in an embodiment of the present invention, and FIG. Figures 8 and 9 are diagrams showing the contents of the processing unit recognition grammar used in the embodiment in comparison with different versions, respectively. Figures 10 and 11 are diagrams showing the state of the data in a tree shape, respectively, by applying one rule in the process 141 Risei Sensho Grammar used in one embodiment of the present invention, and redisplaying it on the screen together with the translated text. Actual layout 2. FIG. 3 is a diagram showing a comparison with before redisplay. 1... Input section, 2... Original text storage section, 3... Translation memory section, 4... Edit'J 1flJ department, 5... Translation section, 6... Translation section, 8...Display section, 9...Print section, 10...Original paragraph single-sided storage section, 11...Analysis data storage section. Agent Patent Attorney Nori Chika Kei Weight Hikaru Matsuyama Figure 11

Claims (1)

[Claims] an input unit having at least a character key, an edit key, and a cursor key; an original text storage unit that stores an original text to be translated inputted from the input unit; and a translation dictionary unit that stores knowledge information used in translation processing; a translation unit that executes translation processing of the input original text using the contents of the translation dictionary unit; a translation storage unit that stores the translated text obtained as a result of the translation processing; and a display for displaying the original text and the translated text. and a translation for controlling these and for editing the original or translated text displayed on the display section and controlling the cursor on the display surface of the display section in response to the operation of the edit key or cursor key. an editing unit, the translation unit recognizes one cycle unit of translation processing from the original sentence string stored in the original text storage unit using knowledge in the translation dictionary unit;
A machine translation system that adjusts the range of the original text that is used as a unit of translation processing according to the characteristics of the original text.