JPH0656613B2 - Machine translation system - Google Patents

Description

The present invention relates to a machine translation system.

(Prior Art) In recent years, a system that automatically machine-translates an input original text by using a computer has attracted attention. This machine translation system basically divides an input original sentence into predetermined processing units such as words (phrases) by performing morphological analysis and syntactic analysis, and searches the translation dictionary for each processing unit to find the corresponding translated word ( It is configured to find a translated phrase) and combine it according to a predetermined translation rule to obtain the translated sentence.

However, with regard to natural language, it is extremely difficult to obtain an appropriate translated sentence because the semantic interpretation technology has not been sufficiently established. Therefore, there is still a need for the operator to edit an appropriate translated text.

For example, if there are multiple dependency relations in the interpretation of the original sentence and the translation result with the dependency relation expected by the operator cannot be obtained, the operator automatically specifies the dependency words and the system automatically Retranslation was performed according to the dependency relationship, and an appropriate translation was obtained.

(Problems to be Solved by the Invention) In the above machine translation system, when the original language is a language such as English having a phrase structure, the concept structure is determined and an appropriate translation is obtained only by specifying the dependent word. However, in the case of languages such as Japanese that have a case structure, the ambiguity cannot be resolved by simply specifying the dependent word, and the expected result may not be obtained. For example, consider a case where a Japanese sentence "This is his book." Is translated into English in a Japanese-English machine translation system. The phrase “his” can mean two things: “owned by him” or “created by him”. Therefore, each translation is "This is his
book. ”,“ This is the book by him. ”, and in order to obtain the translation that the operator expects, do not use the ambiguous expression“ his ”and“ hold it ”or“ his own ”. It had to be translated into a clear and uniquely interpretable expression such as "created." In the above example, the phrase “his” is “book” in both interpretations
However, an appropriate translation cannot be obtained by specifying the destination. For this reason, it was necessary to clarify the relation (case relation) that the phrase "his" has with "is a book", that is, possession or creation. The task of converting the original text expression into an expression that can only be interpreted is extremely troublesome and requires experience, and thus puts a heavy burden on the operator.

It is an object of the present invention to provide a machine translation system that has a simple configuration and simplifies the editing operation of an operator and can easily obtain an appropriate translation result.

[Structure of the Invention] (Means for Solving Problems) In order to achieve the above object, in the present invention, an input unit and an original sentence input via the input unit are separated into word units, and a translation dictionary is provided. A morphological analysis unit that searches for grammatical attributes of each word, a syntactic analysis unit that examines syntactic connection relations of each word examined by the morpheme analysis unit, and extracts syntactically conceivable connection relations; A structure conversion unit that examines and stores the case relations from the connection relation information extracted by the analysis unit and converts the case relations into the case relations of the target language, and the translation generation that generates the translated sentence based on the case relation information of the target language. Section, a display section that displays the original sentence and the translated sentence, and a phrase of the original sentence that can have a plurality of case relations is specified through the input section when editing the translated sentence displayed on the display section. Clarify the case relationship of In order to check the case relation information stored in the structure conversion unit, the case relation information of the specified clause is displayed on the display unit, and the case relation information is displayed on the input unit. A machine translation system comprising: a case conversion unit that prompts the structure conversion unit to retranslate in accordance with selected case relationship information when one of the information items is selected.

(Operation) In the above-described structure, when the translated sentence displayed on the display unit is edited, if the clauses of the original sentence that can have a plurality of case relationships are specified through the input unit, the case search unit In order to clarify the case relation of bunsetsu, the case relation information stored in the structure conversion unit is examined. Then, the case relation information of the specified phrase is displayed on the display unit. Here, when one of the case relation information is selected via the input unit, the case conversion unit
The structure conversion unit is prompted to re-translate according to the selected case relationship information. Then, the structure conversion unit re-translates, and the translation generation unit further generates an appropriate translation.

Embodiment An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is an overall block diagram of a Japanese-English machine translation system which is an embodiment of the present invention. This system includes an input unit 1 composed of a keyboard, an original sentence storage unit 2 for storing Japanese sentences keyed as original sentences, a translated sentence storage unit 3 for storing English sentences corresponding to input Japanese sentences as translated sentences, and controlling the entire system. Edit control unit 4, translation unit 5 that executes translation processing, translation dictionary 6 that contains knowledge information used for translation processing, display control unit 7 that controls the display of original sentences and English sentences, display unit 8 and original and translated sentences. And the like, and a printing unit 9 for outputting a hard copy and outputting it as a hard copy.

FIG. 2 is a diagram showing a key layout of the input unit 1. The input unit 1 is equipped with the following various keys in addition to the character keys for inputting Japanese sentences.

Translation instruction key: 100 Edit key: 101 to 106 Function key: 110 to 118 Cursor key: 130 to 133 Other keys FIG. 3 is a diagram showing a screen layout of the display unit 8.
The input source text is displayed in the source text display area on the left side of the screen. The translated text obtained as a result of the translation process is displayed at the position corresponding to the original text in the translated text display area on the right side of the screen. The edit area at the top of the screen can be used to display information necessary for various edits.

FIG. 4 is a flowchart showing the flow of the dialogue translation process by the edit control unit 4. In the interactive translation process, the operator can appropriately control the editing of the translated sentence corresponding to the input of the original sentence. In steps S1 to S5, the edit control unit 4 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, and if there is any key input, Perform the corresponding processing.

When the operator operates any of the function keys, this is detected in step 5, and the process corresponding to the function key is performed in step S12.

When the operator operates the cursor key 130 or other control keys, the process for moving the cursor corresponding to the key operation and other processes are performed in step S13 through steps S1 to S5.

When the operator operates the character keys to input the original sentence, each character key is detected in step S3, and then the corresponding character code is sequentially set in the input buffer (not shown) of the edit control unit 4 (step S8). . This input original sentence is displayed in the original sentence display area of the display unit 8 via the display control unit 7 (step S9). When the operator presses the translation instruction key 100 at an arbitrary time, for example, when one sentence is input, this key is detected in step S2, and the edit control unit 4 supplies the original sentence in the input buffer to the translation unit 5. ,
The start of translation processing is instructed (step S10). In addition,
If input editing such as correction / insertion / deletion is required in the middle of inputting the original text, move the cursor to the desired edit position with the cursor key, and then use the edit keys such as the insert key 101 and the delete key 102. It can be performed.

FIG. 5 is a flowchart showing the flow of translation processing in the translation unit 5. The phrase analysis unit (S51) converts the input original sentence into phrases by using the word obtained from the dictionary search unit (S52) and the adjunct word connection / inflectional ending change dictionary 6a. The dictionary search unit (S52) searches the word dictionary 6b for words that can form the input original sentence. In addition, information such as the part of speech and translation is also calculated. Analysis control unit (S5
3) performs a process of sending translated sentence candidates to the syntactic analysis unit. The syntax analysis unit (S54) analyzes the syntax of the translated sentence candidate using the analysis grammar 6c, and still generates a structure in which semantic ambiguity is inherent. If the analysis fails, the process returns to the analysis control unit. The semantic analysis unit (S55) uses the semantic analysis grammar 6d to generate a Japanese structure with no ambiguity in interpretation. The structure conversion unit (S56) converts the Japanese structure into English using the conversion grammar 6e. If it is rejected, the procedure returns to the parsing section. The syntax generation unit (S57) determines the English word order from the English structure according to the generation grammar 6f and converts it into a word string. The morpheme generation unit (S58) changes the word endings and the like by the morpheme generation grammar 6g to complete the main translated sentence.

Each of the above processing units outputs a signal to the edit control unit 4 when the processing is started.

Referring again to FIG. 4, when the editing control unit 4 detects the translation completion signal in step S1, the editing control unit 4 proceeds to step 7 and displays the translation candidate and the auxiliary information obtained from the translation unit 5 in the display control unit 7
The translated text is displayed on (step 7). Display control unit 7
Displays the translation candidate at a position corresponding to the input original sentence (on the left side of the screen) in the translation display area, and if there is auxiliary information, the operator can identify it by highlighting or highlighting the translation. I am trying.

From this state, if the edit key is pressed, the edit control unit 4
Detects this in step 4, and performs edit processing corresponding to each key for the word or phrase pointed by the cursor at this time (step S11). For example, by operating the insert key 101, before the cursor position. Insert a character. By operating the delete key 102, the character string in the range indicated by the cursor is deleted. By operating the move key 103, the range designated by the cursor is moved. Cancel key 1
The operation of 04 cancels the effect of the keys 101, 102, and 103.

The following functions can be realized by the function keys. By operating the translated word display key 110, the translated word can be displayed for the word in the translated text. By operating the dictionary display key 111, the dictionary can be displayed for the word in the current sentence. By operating the dictionary registration key 112, new words / idioms can be registered. Dictionary delete key 1
By the operation of 13, it is possible to delete the words / idioms registered by the dictionary registration. When the translation fails by operating the partial translation key 114, the partial translation is displayed.
By operating the dependency change key 115, the case relationship can be displayed for the words in the original sentence.

Further, as the key-sol key, a cursor moving key 130 for moving the cursor in each direction, a unit switching key 131 for switching a unit in which the cursor moves, an inter-region moving key 132 for moving the cursor between respective display regions, and a cursor It is composed of a scaling key for reducing the size in character units or expanding in word units.

Therefore, the operator can interactively obtain the translated text while appropriately inputting the original text and editing the translated text.

In the system configured as described above, the modification processing will be described in detail with reference to the flowcharts shown in FIGS. 6 (A) and 6 (B). First, the operator looks at the original sentence displayed on the display unit 8 and moves the cursor to the phrase for which the dependency or case relation is to be changed (6
01), and press the dependency change key 115 (602)
Then, the edit control unit 4 retrieves the case relationship information stored in the semantic analysis unit (S55) of the translation unit 5 (604).

Here, the case relation information generated by the semantic analysis unit (S55) will be described in detail with reference to FIG. For example, when translating the original sentence "I saw a movie with him." After morphological analysis and syntactic analysis, the semantic analysis unit (S55) focuses on the verb "view = movie = he = me". Connection related information is sent. Here, the structure of Japanese is generated without disambiguation of the interpretation. In the case of this original sentence, there are two possible structures. Fig. 8 (A) shows the structure of "Watching movie ()", "Watching with him ()", "I watched ()", and Fig. 8 (B) shows It shows the structure of "I saw a movie ()", "A movie with him ()", and "I saw ()". (A) and (B)
Compared with, it is different from and shows that there are two possible case relations of the phrase “with him”. The semantic analysis unit (S55) thus generates and stores the Japanese structure in consideration of the case relation. If there are multiple possible structures,
Have either one translated.

Returning to FIG. 6 again, when there are a plurality of case relationships stored in the semantic analysis section (S55) (605), the edit control section 4 causes the display section 8 to display the plurality of case relationships (60).
6). Here, when any one of the case relations is selected and instructed by the operator via the input unit 1 (607), the edit control unit 4 judges whether or not the original case relation is the same (60).
8) If they are not the same, the process proceeds. Next, the edit control unit 4 determines whether there are a plurality of dependents (609), and if there are a plurality of dependents, the words of the dependents are displayed with high brightness. Then, when the operator designates a word to be related to via the input unit 1 (610), the edit control unit 4 judges whether the instructed word is appropriate as a related destination (61).
1) Change the dependency if appropriate (61
2). On the other hand, if there is one contact at 609, then 6
Change the case relationship at 12. The edit control unit 4 displays the result of the change as described above via the display unit 8 (6
13) The dependency change process ends.

Hereinafter, an operation of the operator and a display method of the case relationship will be described with reference to the diagram showing the display screen in the dependency change processing shown in FIG. The translation of the original sentence "I saw a movie with him."
First, when the translation is completed, the original sentence and the translated sentence are displayed (S7).
1). The operator saw this translation and read the translation "I watched a movie with him."
If you want to change to the translation, move the cursor to "with him" and press the dependency change key
The case relationship of "to" is displayed (S72). Here, in the first translation, "to" is interpreted as "issho", so "issho" has a marker. The operator operates the cursor key to move the marker to "and" and presses the confirmation key (S73). By this, it is re-translated and an appropriate translation result is obtained (S74).

As described above, in the present embodiment, when the dependency is changed, attention is paid to the Japanese case structure in which the ambiguity cannot be resolved by simply specifying the dependency word, and this case relationship is displayed in an operator-friendly expression. Therefore, the operator can intuitively grasp the case relation of the original sentence, and the editing operation becomes easy.

[Effects of the Invention] According to the present invention, it is possible to simplify the editing operation of the operator when changing the dependency, and easily obtain an appropriate translation result.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention, FIG. 2 is a diagram showing a key arrangement of an input section shown in FIG. 1, and FIG. 3 is a screen layout of a display section shown in FIG. FIG. 4, FIG. 4 is a flow chart showing the flow of processing of the editing control unit shown in FIG. 1, FIG. 5 is a flow chart showing the flow of processing by the translation unit shown in FIG. 1, and FIG. 6 (A). 7B are flowcharts showing the flow of the dependency change processing by the editing control unit, FIG. 7 is a view showing a display screen in the dependency change processing shown in FIG. 6, and FIG. 8 is shown in FIG. It is a figure which shows the case relationship produced | generated by the translation part. 1: Input section, 2: Original sentence storage section, 3: Translation storage section, 4: Edit control section, 5: Translation section, 6: Translation dictionary, 7: Display section, 8: Printing section

 ─────────────────────────────────────────────────── ─── Continuation of front page (72) Inventor Shinji Amano 1 Komukai Toshiba-cho, Sachi-ku, Kawasaki-shi, Kanagawa Inside Toshiba Research Institute Co., Ltd. (56) Reference JP-A-61-90268 (JP, A)

Claims (1)

[Claims] 1. An input unit, a morpheme analysis unit that divides an original sentence input through the input unit into word units, searches a translation dictionary to check grammatical attributes of each word, and the morpheme analysis unit. A syntactic analysis unit that examines syntactic connection relations of each word to extract syntactically conceivable connection relations, and a case relation is searched and stored from the connection relation information extracted by the syntactic analysis unit, and the case relation is targeted. A structure conversion unit for converting into a case relation of a language, a translated sentence generation unit for generating a translated sentence based on the case relation information of the target language, a display unit for displaying the original sentence and the translated sentence, and a translated sentence displayed on the display unit. When a phrase of an original sentence that can have a plurality of case relations is specified through the input unit at the time of editing, in order to clarify the case relation of the phrase, the case relation information stored in the structure conversion unit is changed. Examine the case of the specified phrase The case retrieving unit for displaying the related information through the display unit, and one of the case relation information is selected through the input unit, so that the retranslation according to the selected case relation information is performed by the structure conversion unit. A machine translation system comprising: a case conversion unit that prompts the