JPS62229472A - Mechanical translation system - Google Patents

Abstract

PURPOSE:To generate a proper translation by dividing an original text punctuated with punctuation marks such as comma to translate it again if analysis is impossible. CONSTITUTION:An analysis control part sends a candidates of a translation to a grammatical constitution analyzing part. If the grammatical constitution analyzing part fails in analysis of all candidates of the translation, candidates of the translation are divided properly by punctuation marks and are sent to the grammatical constitution analyzing part. The grammatical constitution analyzing part uses an analysis grammar 6c to analyze grammatical constitutions of candidates of the translation and generates structures of an English text. If the grammatical constitution analyzing part fails in analysis of all candidates of the translation the analysis grammar 6c is used to analyze grammatical constitutions of divided candidates of the translation, and they are synthesized thereafter to generate structures of the English text. A structure converting part uses a conversion grammar 6d to convert the structure of English to that of Japanese. A grammatical constitution generating part uses a generation grammar 6e to determine the order of words of Japanese in accordance with the structure of Japanese and converts them to a word string, and a morpheme generating part uses a morpheme generation grammar 6f to inflect words, thus completing the translation.

Description

[Detailed description of the invention] [Purpose of the invention] (Industrial application field) The present invention relates to a machine translation system.

(Prior Art) In recent years, systems that use computers to automatically translate input original texts have been attracting attention. This machine translation system basically performs predetermined processing such as morphological and syntactic analysis of the input original text (classifies words (phrases), etc. into 11 categories, and searches translation dictionaries for each processing unit. It is configured to find translated words (translated phrases), etc., and combine them according to predetermined translation rules to obtain the translated text.

In the translation process of such a machine translation system, when an original text is input, a morphological analysis unit converts the input original text into its original form using a translation dictionary, including words with changes in the endings, etc. Next, the dictionary search unit searches the translation dictionary for each word constituting the input original text based on this original word to obtain information such as its part of speech and translated word. Based on this information, the analysis control unit generates translation candidates. The syntactic analysis unit analyzes the syntax of this translation candidate using a translation dictionary and generates the structure of the original language. If the analysis fails here, the process returns to the analysis control unit, and the analysis control unit generates other translation candidates. Next, the structure conversion unit converts the structure of the original word into the structure of the translated word using a translation dictionary. If rejected, the process returns to the parsing section. Next, the syntax generation unit determines the word order of the translated word from the structure of the translated word according to the translation dictionary and converts it into a word string. The morpheme generator completes the translated sentence by changing the endings of words using the translation dictionary.

(Problem to be Solved by the Invention) In the translation processing of the machine translation system described above, delimiters such as commas and courtesy marks are used to indicate that the input source text is parallel, or to indicate insertion. If so, the parser must parse each source text separated by these delimiters together. This analysis process is very complex and takes a long time, and often results in failure or incorrect analysis results.

The present invention has a simple configuration, and is a machine translation system that can generate an appropriate translated text by dividing the original text separated by delimiters such as commas and translating it again when analysis becomes impossible. The purpose is to provide.

[Structure of the invention] (Means and operations for solving the problem) In order to achieve the above object, the present invention includes an input unit, an original text storage unit, a translation dictionary unit, a translation unit, It has a translated text storage unit, a display unit, and an editing control unit that controls these.The translation unit divides the input original text into predetermined processing units such as words (phrases) by morphological analysis and syntactic analysis, and then processes the processing units. When obtaining a translated text by searching the translation dictionary section for each word and finding corresponding translated words (translated phrases), etc., and combining them according to predetermined translation rules, if it becomes impossible to parse during the parsing process, the original text Separate them with delimiters, parse the syntax again, and synthesize the analysis results to obtain a translated sentence. Since the analysis process is performed after dividing the original text using delimiters in this manner, the burden of the analysis process is reduced and an appropriate translated text can be obtained.

(Example) Hereinafter, an example of the present invention will be described with reference to the drawings. 1st
The figure is an overall block diagram of an English-Japanese machine translation system as an embodiment of the present invention. This system includes an input section 1 consisting of a keyboard, an original text storage section 2 that stores the English text keyed in as the original text, a translation storage section 3 that stores the Japanese text corresponding to the input English text as the translated text, and controls the entire system. An editing control unit 4, a translation unit 5 that executes translation processing, a translation dictionary unit 6 that stores knowledge information used in translation processing, a display control unit 7 that controls display of the original text, translated text, etc., display unit 8, and the original text and translated text. and a printing section 9 for copying and outputting hard copies.

FIG. 2 is a diagram showing an example of the key layout of the input unit 1.

The input unit 1 includes the following various keys in addition to character keys for inputting English text.

Translation instruction keys: 100 Edit keys: 101-106 Function keys: 110-117 Cursor keys: 130-133 Other keys FIG. 3 is a diagram showing an example of the screen layout of the display unit 8. The input original text is displayed in the original text display area 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 at 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 edits.

FIG. 4 is a flowchart showing an example of the flow of dialog translation processing by the editing control unit 4. In the interactive translation process, the operator can input the original text and edit and control the corresponding translated text as appropriate. The editing control unit 4 performs steps 81 to S5.
, is there any key input from input section 1?
Alternatively, it monitors whether a translation completion signal is received from the translation unit 5, and performs corresponding processing if there is any key input.

When the operator operates any of the function keys, this is detected in step S5, and the process corresponding to the edit key is performed in step Sll.

When the operator operates the cursor key 130 or other control keys, processing for moving the cursor corresponding to the key operation and other processing are performed in step S13 via steps 81 to S5.

When the operator inputs the original text by operating the character keys, each character key is detected in step S3, and then the corresponding character code is sequentially set in the input buffer (not shown) in the editing control unit 4 (step S8). ). This input original text is displayed in the original text display area of the display unit 8 via the display control unit 7 (step S9). For example, when the operator finishes inputting a - sentence, the operator presses the translation instruction key 100. This key is detected in step S2, and the editing control unit 4 supplies the original text in the input buffer to the translation unit 5. and instructs the start of translation processing (step 510). If input editing such as correction, insertion, or deletion is required during input of the original text, move the cursor to the desired editing location using the cursor keys, then press the insert key 101 and delete key 102.
You can do this using editing keys such as .

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

The morphological analysis unit (S51) applies rules and
Use the Irregular Change Dictionary 6a to find changes in word endings, etc.jl
Convert i-word to its original form (basic form).

The dictionary search unit (S52) searches the LLI word/idiom dictionary 6b for each word constituting the input original text, and obtains information such as its part of speech and translation.

The analysis control unit (S53) performs a process of sending translation candidates to the syntactic analysis unit. Furthermore, if the syntax analysis unit fails to analyze all translation candidates, it performs a process of appropriately dividing the translation candidates using delimiters and sending them to the syntax analysis unit.

The syntactic analysis unit (S54) uses the analysis grammar 6C to analyze the syntax of the translation candidate and generates an English sentence structure. If the analysis fails, the process returns to the analysis control section. If the analysis of all translation candidates fails, the parsing grammar 6c is used to analyze the syntax of the separated translation candidates, and then these are combined to generate an English sentence structure.

The structure conversion unit (S55) converts the English structure into the Japanese structure using the conversion grammar 5d. If rejected,
Return to the parser.

The syntax generation unit (S56) determines the Japanese word order from the Japanese structure according to the generative grammar 6e and converts it into a word string.

The morpheme generation unit (S57) completes the translated sentence by changing the ending of the word using the morpheme generation grammar 6f.

Each of the above-mentioned processing units outputs a signal to the editing control unit 4 at the time of starting processing.

Referring again to FIG. 4, the editing control section 4 performs step S1.
When the translation completion signal is detected, the process proceeds to step S7, where the translation candidates and auxiliary information obtained from translation H5 are sent to the display control unit 7 to display the translated text (step S7). 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 translation with high brightness or in reverse video so that the operator can This allows you to identify that line.

From this state, when the edit key is pressed, the edit control unit 4
detects this in step S4, and performs editing processing corresponding to each key on the word or phrase indicated by the cursor at this time (step 511).

For example, by operating the insert key 101, a character is inserted before the cursor position. By operating the delete key 102, 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 cancel key 104, the key 10
1. Nullify the effects of 102.103. Modifications are made by operating the key 105, and auxiliary information is used to display candidates for modifications other than the word indicated by the cursor.

Note that by operating the @translation display key 110, which can implement the following functions using the function keys, the translation of a word in the translation can be displayed. By operating the dictionary display key 111, a dictionary can be displayed for the word in the current sentence. Dictionary registration key 11
By performing step 2, you can register new words and phrases. 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.

In addition, as a cursor key, there are cursor movement keys 13 (1° unit switching, which changes the unit in which the cursor moves, t-1--13) that moves the cursor in each direction.
1. An inter-area movement key 132 for moving the cursor between display areas, and an enlargement/reduction key for reducing the size of the cursor in units of characters or in units of words.

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

In the machine translation system configured as above, the operations of the analysis control unit (hereinafter referred to as S53) and the syntax analysis unit (hereinafter referred to as S54) will be explained in detail with reference to flowchart 1 shown in FIG. Explain. First, S53 selects one series from among several word series obtained by the dictionary search unit (S52) (601). Next S53
determines whether there is a delimiter (hereinafter referred to as A delimiter) added by the operator in advance (6
02). The A delimiter is a delimiter in the original text that is determined in advance by the operator in order to facilitate the analysis process. For example, “l saw a girl and a
When you want to translate the original text "boy,"
I saw ra gfrl and a boyJ
, ” When the original text is input by adding a delimiter (``''), S53 detects the delimiter and can immediately recognize the sentence break. If there is no delimiter in 602, S53 sends the entire sentence of the selected series to 554 (613). S54 performs syntactic analysis of this sentence (614). On the other hand, if there is an A delimiter (602), S53 sends the portion surrounded by the A delimiter to 354 (603). S54 performs syntax analysis of the part surrounded by the A delimiters (604). Syntax parsing failed at 604 and 614 (605, 615)
Sometimes S54 signals S53 that it has failed and S5
Upon receiving this notification, No. 3 checks whether there are any other word sequences obtained by the dictionary search unit (S52) (606, 619).
to decide. If there are other word sequences, one r11 word sequence is selected from them (601, 618), and the above process is repeated until the parsing is successful (605, 615). If the parsing is successful in 605, S54
3, and S53 stores the analysis result in an internal buffer (609). Then, it is determined whether there are any other delimiters (6013). If there are no other A delimiters (608), S53 sends the part other than the A delimiters to 354 (610), S54 parses this (611), and S54 parses it to S53. Send the results. S53 combines the analysis results stored in the buffer with the sent analysis results (612), stores the combined results (fi 16), and sends this result to the structure conversion unit (S57) (617). . 615
If the analysis is successful in , S54 sends this analysis result to S53, and S53 saves this analysis result (616
) and sends this result to the structure conversion unit (S55) (
617). On the other hand, if there are other A delimiters in 608, S53 extracts the information necessary for analyzing the part surrounded by the other A delimiters from the buffer in which the analysis results are stored (607), and sends the information to other A delimiters (607). Send the part surrounded by delimiters to 354 along with the necessary information (603)
. S54 analyzes the portion surrounded by the A delimiters with reference to this necessary information (604). In this way, when there are multiple A delimiters, each part is analyzed as described above, and in 612 they are combined to generate the structure of the original word.

If the syntactic analysis of all j1 obtained by the dictionary search unit (S52) in 606 and 619 fails for the word series, S53 uses a delimiter other than the delimiter (hereinafter referred to as 8
(referred to as a delimiter) is present (620). 8 delimiters are delimiters written in the original text from the beginning, such as commas, coach John marks, colons, and semicolons. For example, “That, I th
ik, is his best w.

In the sentence rk,'', the comma is the 8 separator.

If there is no 8 delimiter in 620, S53 is S5
4 to partially parse (63
2). Then, S53 again selects one series from the several word series obtained by the dictionary search unit (S52), and
54 (633).

S54 receives the partial translation instruction from S53, analyzes the partial translation until it can be partially analyzed, and partially generates the structure of the original language (634). This analysis result is saved by 553 (616) and sent to the structure conversion unit (S55) (61
7). On the other hand, if there are other 8 delimiters in 620, select one series again from several word series obtained by the Jiyo search unit (S52), and select 8 in the selected series.
The portion surrounded by the delimiter is sent to S54 (622).

S54 parses this (623), and if the parsing fails (624), S54 notifies S53 that the parsing has failed, and S53 receives this notification and sends the information to the dictionary security department (S52). It is determined whether there are other word sequences that have been found (631). 62 if there are other word series
Return to step 1 and select another word series (621), and repeat the above process until the analysis is successful. If syntactic analysis fails for all word sequences (631), the process proceeds to 632 and partial translation processing is performed (632, 633, 634). If the analysis is successful (624), S54 sends this analysis result to S53, and S53 stores this analysis result (625). Then, S53 determines whether there are other 8 delimiters (62
6). If there are no other 8 delimiters (626), S53
sends the parts other than those enclosed by 8 delimiters to 854 (
62 g).

S54 parses the syntax of this other part and sends the result to S54 (629). S53 combines the analysis results already saved in the buffer with the sent analysis results (630) and saves this result in the buffer (616).
At the same time, this result is sent to the structure conversion unit (S55) (6
17). On the other hand, if there are other 8 delimiters in 626, S53 retrieves the information necessary to analyze the part marked 17M by the other 8 delimiters from the buffer in which the analysis results are stored (628), and the other 8 delimiters. The part surrounded by delimiters is sent to S54 along with the necessary information (622
). 854 refers to this necessary information and analyzes the portion surrounded by 8 delimiters (623). In this way, when there is an inverse number of 8-delimiters, each part is analyzed as described above, and in 630, they are combined to generate the structure of the original word.

To explain using a specific sentence, for example, “Tbat, 1
thlk, is his best work,”
When inputting the original sentence, S53 selects the word series, separates them with commas, and sends them to S54 based on the word strings with parts of speech sent through the morphological analysis unit (51) and dictionary search unit (52). send. In this example, first, '1think'
S54 parses this and sends this result to S53. S53 saves this result.

Next, 553 is “That is his best w”
ork” to 854. S54 analyzes this and sends S5
Send this result to 3. S53 combines this analysis result with the stored analysis result to generate the structure of the original word.

As mentioned above, in this embodiment, the analysis control unit (S53
) instructs the syntax analysis unit (S54) to break up sentences in advance, so the processing load on the syntax analysis unit (S54) is reduced.

Note that the present invention is not limited to the English-Japanese machine translation system of the above embodiment, but can also be applied to other machine translation systems. Furthermore, in the above embodiment, after parsing all translation candidates fails, the original text is separated using delimiters and then parsed.
You can first parse the original text by separating it with a delimiter, and if that fails, analyze it all at once without separating it.

[Effects of the Invention] As described above, according to the present invention, when analysis becomes impossible, the original text separated by delimiters such as commas is divided and translated again, making it possible to generate an appropriate translated text. can.

[Brief explanation of drawings]

FIG. 1 is an overall block diagram of an embodiment of the present invention, FIG. 2 is a diagram showing a key arrangement of an input section used in an embodiment of the present invention,
FIG. 3 is a diagram showing an example of the screen layout of the display unit used in an embodiment of the present invention, FIG. FIG. 5 is a flowchart showing an example of the flow of translation processing used in an embodiment of the present invention. FIG. 6 is a flowchart showing an example of the flow of analysis control processing. Input section 2: Original text storage section

Claims (2)

[Claims] (1) An input section, an original text storage section that stores the original text to be translated that is input from this input section, a translation dictionary section that stores knowledge information used for translation processing, and a translation dictionary section that uses the contents of this translation dictionary section. a translation unit that executes translation processing of the input original text, a translation storage unit that stores the translated text obtained as a result of translation, and a display unit that displays the original text and the translated text; and an editing control section that performs editing processing of the original text or translated text displayed in the translation section, and when the translation section becomes unable to analyze the text, after dividing the original text with delimiters, the analysis processing is performed for each part. , a machine translation system characterized by synthesizing analysis results to obtain a translated text. (2) The translation unit includes a morphological analysis unit that converts words with changes in endings etc. into their original forms using the contents of the translation dictionary unit for the input original text, and the original form of the words changed by the morphological analysis unit. a dictionary search unit that searches the contents of the translation dictionary for each word that makes up the input original sentence and obtains information such as its part of speech and translated words; an analysis control unit that divides the candidates as appropriate and sends them; and parses the syntax of the translation candidates sent from the analysis control unit using the contents of the translation dictionary unit; if the analysis fails, a syntactic analysis unit that obtains other translation candidates from the analysis control unit, parses the syntax, and generates the structure of the original word; and a syntactic analysis unit that generates the structure of the original word by parsing it, and uses the contents of the translation dictionary unit to convert the structure of the original word generated by the syntactic analysis unit into the structure of the translated word. a structure conversion unit that converts the translated word into a word string; a syntax generation unit that determines the word order of the translated word according to the contents of the translation dictionary unit based on the structure of the translated word converted by the structure conversion unit and converts it into a word string; a morpheme generation unit that completes a translated sentence by changing the endings of words of the converted word string using the contents of the translation dictionary unit, and the syntactic analysis unit,
If the syntax analysis of all the translation candidates given by the analysis control unit fails, the analysis control unit performs a process of appropriately separating the translation candidates using delimiters and sending them to the syntax analysis unit. A machine translation system according to claim 1.