JPH08221424A - Machine translation system - Google Patents

Abstract

PURPOSE: To present the same translated sentence to a user present to the same sentence by avoiding uselessness such as double translation of the same sentence. CONSTITUTION: This system is constituted by connecting a translation data base device 21 provided with a translation data base 20 storing translation and a translation data base managing means 19 for retrieving the translation data base corresponding to a retrieving instruction and updating the translation data base corresponding to an updating instruction through a network 22 to a machine translation device 17. When an input is inputted to the machine translation device 17, a translation retrieval requesting means 12 executes the retrieval while using the translation data base device 21 so that it can be known whether the same translation was executed in the past or not, and when it was executed, that translated sentence can be extracted. Besides, a translation update requesting means 14 can execute the update of translation while using the translation data base device 21, if needed, based on the retrieved result.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a machine translation system, for example, applied to a system capable of executing machine translation in a distributed computer environment in which a plurality of computers (information processing devices) are connected by a high speed network such as a LAN. It is suitable.

[0002]

2. Description of the Related Art Using a dictionary holding word information and grammatical knowledge for parsing an input sentence and generating a translated sentence,
A method of constructing a machine translation device for translating an original sentence into a sentence of a target language has already been proposed (Japanese Patent Publication No. 3-51020).

Due to the ambiguity of natural language, the translated text output by the machine translation device may not be used as it is.
In such a case, post-editing work is usually performed manually to generate a final translated sentence. Alternatively, the machine translation is performed again after the words are registered using the function (user dictionary function) of registering the words in the dictionary for each user provided in the machine translation device. In practice, these two translation quality improvement tasks are often used together.

[0004]

By the way, when there are a large number of sentences to be translated, the sentences of different parts are assigned to a plurality of persons, and each person is assigned using a machine translation device. Sometimes the translation of sentences is done. For example, 1 such as the operation manual of the device
When translating a book, multiple people are responsible for translating the text of the assigned page. In this way, when the translation target is a large amount of text, the same text is often included in it,
Different people are in charge of the translation depending on the position where they exist. Each person often cannot recognize that the same sentence exists in a plurality of sentences assigned to another person. In such a work environment, there are the following problems.

(1) The same sentence is separately translated by a plurality of people using a machine translation device, and the translation work is redundant, resulting in a waste of work as a whole.

(2) Even if the translation result of the machine translation device for the same sentence is different due to the user dictionary registration function for each operator, or the same translation result is output, the final result is obtained by post-editing for each operator. The translation results may differ, and the same translation result may not be obtained.

Even if the same sentence appears a plurality of times in a sentence in which one person is in charge, the machine translation device has no means for finding the same sentence, and the operator is left to find the same sentence. However, the problem of work efficiency in (1) above due to duplication of translations and the problem of variation in translation results in (2) above may occur due to differences in the consciousness of the operator for each translation.

Further, when a machine-translated book is revised, the translation is executed again. In such a case,
Most sentences are translated in the past. In this case, since the machine translation device does not have a means for finding the same sentence, the operator compares the revised books and recognizes the sentence that needs to be translated, or re-executes the full-text translation.
In the latter case, the problem of work efficiency in (1) above and the above
The problem of variation in translation results in (2) arises.

Therefore, there is a demand for a machine translation system that can efficiently obtain the same translation result for a plurality of identical sentences existing in a sentence to be translated even if the translation times are almost the same.

[0010]

In order to solve such a problem, in the present invention, one or more machine translation devices and a parallel translation database device commonly provided in these machine translation devices are connected to a communication network. The machine translation system has been configured.

Then, each machine translation device is provided with (i) an input means for fetching an input including at least an original sentence, and (ii) a search instruction as to whether or not a parallel translation for the input fetched by the input means is stored in the parallel translation database device. A parallel translation search request means for giving a translation result to the parallel translation database device and receiving a search result from the parallel translation database device to determine the necessity of translation of the input original text, and (iii) a translation means for translating the input original text. (Iv) Parallel translation update request that gives an update instruction to the parallel translation database device including the input original text and the parallel translation information including the input translation, or the input original text and the parallel translation information including the translation translated by the translation means. Means and (v)
According to a search result from the bilingual database device, an output unit is provided which outputs a translation included in the search result or a translation translated by the translation unit as a translation for the input original sentence.

Further, the parallel translation database device incorporates (vi) a parallel translation database storing parallel translations, and (vii) input means of the machine translation device in response to a search instruction from any of the machine translation devices. It searches the bilingual translation for the input stored in the bilingual translation database, returns the search results, and updates the bilingual translation database in response to an update instruction from any machine translation device, waiting for multiple instructions. There is a parallel translation database management means that manages in a matrix and prioritizes the one that arrives first in the case of competition of the same type of instruction, and prioritizes the update instruction in the case of competition between the search instruction and the update instruction.

Here, the parallel translation database management means is
(A) When returning the search result that activates the translation in response to the search instruction to the machine translation device that issued the search instruction, make sure that the translation in the parallel translation database for the search instruction is "translating". (B) When a translation corresponding to the search instruction yields a parallel translation of “Translating”, the search instruction is reinserted at the back of the instruction queue, and (C) “Translating” is added. When an update instruction including a parallel translation corresponding to the translated parallel translation is given, it is preferable to update the parallel translation database and delete "during translation".

[0014]

In the machine translation system of the present invention, a parallel translation database storing parallel translations and a parallel translation database management means for searching the parallel translation database according to the search instruction and updating the parallel translation database according to the update instruction. The parallel translation database device included therein is connected to one or more machine translation devices via a communication network.

When an input is made to the machine translation device,
By the parallel translation search requesting means causing the parallel translation database device to execute the search, it is possible to retrieve whether or not the same translation has been executed in the past and, if it has, execute the translated sentence. Further, based on the search result, the parallel translation update requesting unit can cause the parallel translation database device to update the parallel translation, if necessary.

In this case, the parallel translation database management means manages a plurality of instructions in a queue, prioritizes the one that arrived first in the conflict of the same kind of instruction, and updates the instruction in the conflict between the search instruction and the update instruction. Therefore, even if multiple machine translation devices try to translate the same sentence at about the same time, the parallel translation database is updated by the other machine translation device in response to an update instruction from one machine translation device. Before the search instruction of, the other machine translation device can use the parallel translation immediately after being updated as the translation result.

Therefore, even when trying to translate the same sentence at almost the same time, it is possible to avoid the waste of double translation of the same sentence, and to present the same translated sentence to the user for the same sentence early. Like

Here, when the parallel translation database management means recognizes that the translation of the original text input by the machine translation device that received the search result is executed at the time of searching the parallel translation database, the original text of the parallel translation database is found. "Translating" is added to the content corresponding to, and when the same search instruction from another machine translation device is sent, the search result is transmitted after the completion of the translation. It is possible to further reduce the waste of duplicate translation when trying to translate the same sentence at almost the same time.

[0019]

【Example】

(A) First Embodiment Hereinafter, a first embodiment of the machine translation system according to the present invention will be described in detail with reference to the drawings. Here, FIG. 1 is a functional block diagram showing the configuration of the machine translation system of the first embodiment.

In FIG. 1, the machine translation system is
One or more machine translation devices 17 and a parallel translation database device 21 are connected to a network 22 such as a LAN.

Here, each of the machine translation device 17 and the parallel translation database device 21 is composed of an information processing device having a communication function such as a workstation and a personal computer, and functionally has the structure shown in FIG. . The parallel translation database device 21 may be configured on the same information processing device as any of the machine translation devices 17.

The machine translation device 17 includes input means 11, parallel translation database search means (represented by parallel translation DB search means in FIG. 1) 12, translation means 13, parallel translation database update means (represented by parallel translation DB update means in FIG. 1). 14),
It is composed of output means 15 and communication means 16.

On the other hand, the parallel translation database device 21 includes a communication means 18 and a parallel translation database management means (parallel translation D in FIG. 1).
B management means) 19 and a parallel translation database (represented by a parallel translation DB in FIG. 1) 20.

Hereinafter, each component 1 of the machine translation device 17
The functions of 1, ..., 16 will be sequentially described.

The input means 11 takes in an input to the machine translation device 17 and processes it. The input may be made by key input at that time, or may be already stored as a file.

Here, for simplification of description, the input from the user is the operation mode, the user dictionary information, and the sentence. In addition, as the operation mode, a “normal operation mode” which is a normal operation mode in which an input sentence is translated and a translation result is output as it is, and a bilingual translation including an original sentence and a translated sentence is registered in the bilingual database 20 as necessary. There are two "example translation modes" for searching whether a parallel translation corresponding to the input sentence already exists in the parallel translation database 20.
The input sentence may be the original sentence only, or the original sentence and the translated sentence, that is, the parallel translation. The user dictionary information includes a user dictionary name and a user dictionary version number, and is hereinafter referred to as a user dictionary ID.

The parallel translation database search means 12 generates a search command for the parallel translation database 20 using the original sentence of the input sentence as a key, transmits the search command to the parallel translation database device 20 via the communication means 16, and then returns the search result. It receives a reply and determines the necessity of translating the input source text.

Here, the search command includes the parallel translation and the user dictionary ID when the input sentence is the parallel translation, and includes the original sentence and the user dictionary ID when the input sentence is the original translation. Note that user dictionaries with the same user dictionary ID are
Assume that the same user dictionary is guaranteed.

Further, the parallel translation database searching means 12 is
The returned search result is received via the communication means 16, and the following processing is performed according to the content of the search result. That is, when the translation result (translated sentence) is received as the search result, the translation result is presented to the user via the output means 15. When "no data" is received as the search result, the translation means 13 is activated to translate the original sentence.

The translation means 13, when the translation is instructed by the input means 11 or the parallel translation database search means 12,
The original text captured by the input means 11 is translated. Further, at least the translated text (which may be a set of the translated text and the original text) is given to the output means 15 to be presented to the user. Furthermore, in a predetermined case, the original sentence and the translated sentence, that is, the parallel translation is given to the parallel translation database updating means 14.

The parallel translation database updating means 14 generates an update command for the parallel translation database 20 including the original sentence, the translated text, and the user dictionary ID, transmits the update command to the parallel translation database device 20 via the communication means 16, and the It waits for a reply of the update result "registration completed".
The parallel translation database updating means 14 activates the output means 15 when receiving “registration completed”.

The output means 15 is, as described above, the parallel translation database searching means 12 or the parallel translation database updating means 1.
Under the control of No. 4, at least the translated text (a set of the translated text and the original text may be presented) is presented to the user.

The communication means 16 transmits the search command from the parallel translation database searching means 12 to the parallel translation database device 21.
The search results sent back to the translation database are received and given to the parallel translation database search means 12. Further, the communication means 16 sends the update command from the parallel translation database updating means 14 to the parallel translation database device 21, receives the returned update result (“registration completed”) and gives it to the parallel translation database updating means 12. is there.

The communication means 16 cooperates with the communication means 18 of the parallel translation database device 21 to cooperate with the network 2
Of course, the establishment and disconnection of the connection link via 2 are also executed.

A machine translation device 1 including the above-mentioned components.
A common parallel translation database device 21 that 7 accesses as appropriate
Is installed for the purpose of avoiding duplicate translation of the same sentence and obtaining a unified translation result (translated sentence) of the same sentence.

The functions of the constituent elements 18, ..., 20 of the parallel translation database device 21 will be sequentially described below.

When the communication means 18 receives the search command via the network 22, the communication means 18 gives the search command to the parallel translation database management means 19. When the search result is given from the parallel translation database management means 19, the communication means 18 directs it to the machine translation device 17 as the search request source. The search result is transmitted. Further, when the communication unit 18 receives the update command via the network 22, the communication unit 18 gives the update command to the parallel translation database management unit 19. When the update result is given from the parallel translation database management unit 19, the communication unit 18 sends the update command to the machine translation apparatus 17 as the update request source. The update result is transmitted.

The parallel translation database 20 comprises an original text database 20a, a parallel translation management database 20b, and a translated text database 20c.

The original sentence database 20a stores a plurality of pairs of original sentences and original sentence identification information (hereinafter referred to as original sentence IDs). FIG. 2 shows an example of the structure of the original text database 20a and shows a tree structure. The part shown in Fig. 2 consists of three original sentences "" cat "isa command name.", "" C
at "concatenates argument files and output them to
standard output. '', `` with option -n, it adds the
line number to the output data. ", and the original text IDs are" 001 "," 002 ", and" 00 ", respectively.
3 ”is given.

In the case of such a tree structure, the word part on the head side common to the first two original sentences, that is, "" cat "" has the same branch, and these original sentences are stored independently. Compared with, the search can be executed faster.
"" Cat "is a command name."
When searching for the existence in a, the existence can be confirmed by tracing the branch from the "start" to the "end" (91), and further to the data "001" (94) stored therebelow. By tracing, it can be seen that the original text ID is "001".

As shown in FIG. 3, the translated text database 20c stores a plurality of pairs of translated texts and translated text identification information (hereinafter referred to as translated text IDs). In FIG. 3, the structure of the translated text database 20c is shown as a table structure, but like the original text database 20c, a tree structure is preferable in consideration of high-speed search.

The parallel translation management database 20b has the original text I
A plurality of sets (records) each including D, the translated text ID, the translated text time stamp, and the user dictionary ID are stored. FIG. 4 shows an example of the structure of the parallel translation database 20b, which has a table structure. The translated sentence time stamp is an identifier attached to each input unit input to the parallel translation database device 21. As described above, the user dictionary ID is a list of the input user dictionary name and its version number.
When the user dictionary is not designated, data indicating that is inserted, for example, "none".

In this embodiment, the translation ID of the parallel translation management database 20b is "under translation", as will be described later.
May be inserted. When the translated text ID is “translating”, the translated text for the original text defined by the parallel translation management data does not exist in the translated text database 20c, but at that time, one of the machine translation devices 17 is translating the original text. It means that there is.

Basically, the parallel translation database management means 19 accesses the parallel translation database 21 in accordance with the search command or update command given from the communication means 18, and the search result or update result is sent from the communication means 18 to the communication means 18. It is transmitted to the machine translation device 17.

The parallel translation database management means 19 manages the search command and the update command in the same queue 19a. When the search command or the update command is given, if the command queue 19a is not "empty", the received command is inserted into the queue 19a.

Here, for the same type of command, the command is inserted so that the command that arrives first has priority, and the command is inserted so that the update command has priority in the relationship between the search command and the update command. That is, when the search command is received, the command queue 19a
When the update command is received at the end of the update command, the update command is inserted at a position after the end update command of the command queue 19a and immediately before the start search command.

When the processing for a certain command is completed, the parallel translation database management means 19 takes out the command at the head of the command queue 19a and processes the command.

If the command extracted from the command queue 19a is a search command, the parallel translation database management means 19 basically searches the original text database 20a, the parallel translation management database 20b, and the translation text database 20c in this order. That is, if the original sentence database 20a is searched and the original sentence exists, the parallel translation management database 20b is searched using the obtained original sentence ID as a key, and if the translated sentence having the same user dictionary ID exists, the obtained translated sentence ID is obtained. The translation database 20c is searched using as a key.

This search is executed in order to confirm whether or not the parallel translation of the original text translated in the past exists in the parallel translation database 20. In other words, it is executed so that it is possible to determine whether or not new translation is necessary for the current input.

The parallel translation database management means 19 performs one of the following processes according to the retrieved result.

(1) When an original sentence is input, the input original sentence exists in the original sentence database 20a, and
When parallel translation management data relating to the original sentence having a user dictionary ID that matches the input user dictionary ID exists, the translation database 20c is searched based on the translation ID, and the translation with the latest translation time stamp is retrieved ( As a translation result), the machine translation device 1 of the search request source via the communication means 18
Reply to 7.

(2) When a parallel translation is input, the input original text exists in the original text database 20a, and
When parallel translation management data related to the original text having a user dictionary ID that matches the input user dictionary ID exists, the translation database 20c is searched based on the translation ID, and the searched translation and the input translation are matched. Then, the translated text is returned as a search result (translation result) to the machine translation device 17 that is the search request source via the communication means 18.

(3) When the bilingual translation is input, the input original text exists in the original text database 20a, and
When parallel translation management data related to the original text having a user dictionary ID that matches the input user dictionary ID exists, the translation database 20c is searched based on the translation ID, and the searched translation and the input translation are matched. If not done, the translation database 20 (translation management database 20
b and the translated text database 20c) are updated, and the entered translated text is returned as a translation result to the machine translation device 17 which is the search request source via the communication means 18.

That is, in this case, the parallel translation database 20 is also updated in the search stage.

(4) When an original sentence is input, the input original sentence exists in the original sentence database 20a, and
There is parallel translation management data related to the original text having a user dictionary ID that matches the input user dictionary ID,
If the translated text ID of the parallel translation management data is "translating",
The search is stopped and the queue of the search command is inserted at the end of the command queue.

(5) When the bilingual translation is input, the input original text exists in the original text database 20a, and
There is parallel translation management data related to the original text having a user dictionary ID that matches the input user dictionary ID,
If the translated text ID of the parallel translation management data is "translating",
The search is stopped and the queue of the search command is inserted at the end of the command queue.

Therefore, in the cases (4) and (5), the search result is not returned to the machine translation device 17 as the search request source at this stage.

(6) When the original text is input and the input original text does not exist in the original text database 20a, or the input original text is the original text database 20a.
If there is no parallel translation management data related to the original text that has a user dictionary ID that matches the input user dictionary ID that exists, the “translation in progress” is inserted into the corresponding translation ID of the parallel translation management database 20b, and “ “No data” is returned to the machine translation device 17 that is the search request source via the communication means 18. If the input original sentence does not exist in the original sentence database 20a, the original sentence is also registered.

(7) When a parallel translation is input and the input source text does not exist in the source text database 20a, or the input source text is in the source text database 20a.
If there is no parallel translation management data related to the original text that has a user dictionary ID that matches the input user dictionary ID that exists, the “translation in progress” is inserted into the corresponding translation ID of the parallel translation management database 20b, and “ “No data” is returned to the machine translation device 17 that is the search request source via the communication means 18. If the input original sentence does not exist in the original sentence database 20a, the original sentence is also registered.

If the command queue extracted from the head of the command queue 19a is an update command, the parallel translation database management means 19 updates (registers) the parallel translation database 20 in the order of the original text database 20a, the translated text database 20c, and the parallel translation management database 20b. To execute.

When the original text is registered in the original text database 20a, the parallel translation database management means 19 recognizes only the original text ID without performing registration if it is already registered. It is registered in the database 20a and is given a new original text ID.

When attempting to register a translated sentence in the translated sentence database 20c, the parallel translation database management means 19 recognizes only the translated sentence ID without executing registration if the registered sentence is already registered, and translates the translated sentence if not registered. It is registered in the database 20c and a new translated text ID is given.

The parallel translation database management means 19 is operated when the update processing for the original text database 20a and the translated text database 20c is completed.
0b is updated, and the obtained original sentence ID and translated sentence ID,
The parallel translation management data including the input user dictionary ID and the time stamp at which the command is input is additionally registered in the parallel translation management database 20b, and when this registration is completed, the update result meaning "registration completed" is communicated. It returns to the machine translation device 17 of the search request source via 18. In the registration in the parallel translation management database 20b, the "translation in progress" portion is changed to the translated text ID in the case of a predetermined condition.

FIG. 5 shows an example of the structure of a message exchanged between the machine translation device 17 and the parallel translation database device 20, and this example relates to an update command. Although not shown, information defining a transmission source device, a transmission destination device, and the like is also attached to the head of the portion shown in FIG.

The message includes a message length indicating the total amount of messages to be sent and received, a message name indicating the type of search command, search result, update command, update result, etc.
It is composed of a data type representing a type such as an original sentence, a translated sentence, a user dictionary name, a user dictionary version number, a data length representing a data amount of the data type, and specific data of the data type. In some messages, a combination of a data type, a data length, and specific data (in some cases, no specific data exists) is repeated.

Next, the operation of the machine translation system of the first embodiment will be described in detail with reference to the flowchart of FIG.

For convenience of illustration, FIG. 6 shows a flow of processing from inputting one original sentence or parallel translation and presenting the translation result to the user. for that reason,
FIG. 6 is a flowchart in which the processing for the search command and the update command related to other input is omitted.

It is also possible to input a plurality of original sentences or bilingual translations collectively, and in this case, the flow is slightly different from the notation shown in FIG. Although illustration is omitted, the search command and the update command may be issued separately. Even in such a case, the translation result may be presented to the user when the translation results for all input sentences are complete. Further, the sending and receiving of the search command, the update command and the result thereof may be collectively performed.

When the input means 11 receives an input from a file or the like (step 101), the operation mode in the input information is checked (step 102).

When the operation mode is the "normal translation mode", the machine translation means 13 translates the input original sentence using the user dictionary, the system dictionary and the translation rules (step 121), and the output means 15 produces the translation. The translation result is presented to the user, and the series of processes for this input is completed (step 123).

On the other hand, when the operation mode is the "example operation mode", the parallel translation database searching means 12 generates a search command for searching the parallel translation database 20 using the original sentence as a key and sends it to the parallel translation database device 21. (Step 103), the parallel translation database device 21 receives this (step 104).

The received search command is inserted by the parallel translation database management means 19 of the parallel translation database device 21 at the end of the command queue 19a, and the processing is repeated until it reaches the head of the command queue 19a (step 105). ).

Now, when the search command related to the target input becomes the head of the command queue 19a and becomes the target of the search process, the parallel translation database management means 19
The parallel translation database 20 is searched (step 106).

Next, the parallel translation database management means 19
Whether the input original sentence exists in the original sentence database 20a regardless of whether it is a parallel translation input or an original sentence input, and whether parallel translation management data related to the original sentence having a user dictionary ID that matches the input user dictionary ID exists. It is determined whether or not (step 107).

If a negative result is obtained in this determination (corresponding to (6) or (7) in the above case), "translation in progress" is inserted into the corresponding translation ID of the parallel translation management database 20b, and "No data" is returned to the machine translation device 17 that is the search request source via the communication means 18 (step 1).
13, 111).

It is determined in step 107 that the input original sentence exists in the original sentence database 20a, and the parallel translation management data related to the original sentence having the user dictionary ID that matches the input user dictionary ID exists. When it is recognized (corresponding to (1) to (5) in the above case), it is further determined whether or not the translated text ID of the parallel translation management data is "translating" (step 109).

If the translated text ID of the parallel translation management data is "translating" (corresponding to (4) and (5) in the above case), the parallel translation database management means 19 cancels the search processing and returns the search command. Is inserted at the end of the command queue 19a (step 105).

If the translated sentence ID of the parallel translation management data is not "translating" (corresponding to (1) to (3) in the above case), the parallel translation database management means 19 determines whether or not the input is only the original sentence. If the input is a bilingual translation, it is determined whether the input translation matches the retrieved translation (step 10).
9). That is, at this stage, it is determined whether or not the parallel translation database 20 can be updated (the case (3) corresponds to the above case) (the cases (1) and (2) correspond to the above case).

If it is possible to update, the parallel translation database management means 19 updates the parallel translation database 20 (step 110), and then uses the input translation as a search result (translation result) to make a search request source via the communication means 18. To the machine translation device 17 (step 111). On the other hand, if it is not possible to update, the translated sentence with the latest translated time stamp extracted from the translated database 20c is returned as a search result (translation result) to the machine translation device 17 as the search request source via the communication means 18 (step 111).

Such a search result is displayed in the machine translation device 17
It is received by (the parallel translation database search means 12) (step 114). Parallel database searching means 12
Determines whether or not the parallel translation for the input original sentence is registered in the parallel translation database 20, based on the retrieved search result, that is, whether or not the translation for the input original sentence is necessary (step 115). .

If it exists (when the translation result is received), the output means 15 presents the translated text (which may be a combination of the original text and the translated text) included in the search result to the user and responds to the current input. A series of processing ends (step 12)
3).

On the other hand, when recognizing that the parallel translation corresponding to the input does not exist in the parallel translation database 20 (“no data”), the parallel translation database searching means 12 causes the translation means 13 to execute.
To translate the input text (step 11)
6). Then, the parallel translation database updating means 14 generates an update command including the input original sentence and the translated sentence obtained by the translation, and transmits the update command to the parallel translation database device 20 (step 117), and the parallel translation database device 21 receives this. (Step 118).

The received update command is inserted by the parallel translation database management means 19 of the parallel translation database device 21 at the end of the update command in the command queue 19a, and the processing is repeated until it comes to the head of the command queue 19a. (Step 119).

Now, when the update command related to the target input becomes the head of the command queue 19a and becomes the target of the update processing, the parallel translation database management means 19
The parallel translation database 20 is updated (step 120).
After that, the parallel translation database management means 19 transmits the completion of the update to the update request source, that is, the parallel translation database update means 14 of the machine translation device 17 that requested the update (step 121), and the parallel translation database search of the machine translation device 17 is performed. The means 12 fetches this update result (step 1
22).

Upon recognizing the completion of the update, the parallel translation database updating means 14 activates the output means 15 to present the translated text obtained by the translation (the input original text and the set of the translated text may be presented) to the user, A series of processing for this input is ended (step 123).

The mode by such a series of processes is divided into cases for the “example operation mode” according to the input sentence, the presence / absence of the parallel translation database 20, the data to be updated, the translation presented to the user, etc. You can divide the case as follows.

In some cases, the same processing path is used in FIG. 6 in the following cases, but the search results, update processing, output contents, etc. are different from the other cases.

(A) When only the original sentence is input,
When the input original sentence does not exist in the original sentence database 20a, the input original sentence is translated, and the original sentence, the translated sentence obtained by the translation, and the parallel translation management data connecting these original sentence and the translated sentence are newly added to the parallel translation database 20. The translated text obtained by the registration and the translation is presented to the user. In this case, “translating” is temporarily registered in the parallel translation database 20.

The change of the processing step in this case is 101
~ 107, 113, 111, 114-123.

(B) When only the original sentence is input,
Although the parallel translation related to the input original text exists in the parallel translation database 20, if the parallel translation management data related to the original text having the user dictionary ID that matches the input user dictionary ID does not exist in the parallel translation database 20, it is input. When the translated original text is translated and the translated text obtained by the translation matches the translated text stored in the parallel translation database 20, new parallel translation management data (translated text time stamp and user Only the dictionary ID different from the existing one) is newly registered in the parallel translation database 20, and the translated text obtained by the translation is presented to the user. Also in this case, “translating” is temporarily registered in the parallel translation database 20.

The change of the processing step in this case is 101
~ 107, 113, 111, 114-123.

(C) When only the original sentence is input,
Although the parallel translation related to the input original text exists in the parallel translation database 20, if the parallel translation management data related to the original text having the user dictionary ID that matches the input user dictionary ID does not exist in the parallel translation database 20, it is input. When the translated original text is translated and the translated text obtained by the translation does not match the translated text stored in the parallel translation database 20, parallel translation management connecting the translated text obtained by the translation with the input original text and the translated translation The data is newly registered in the parallel translation database 20, and the translated text obtained by the translation is presented to the user. Also in this case, “translating” is temporarily registered in the parallel translation database 20.

The change in the processing step in this case is 101
~ 107, 113, 111, 114-123.

(D) When only the original sentence is input,
If the parallel translation of the input original text exists in the parallel translation database 20, and the parallel translation management data of the original text having the user dictionary ID that matches the input user dictionary ID exists in the parallel translation database 20, Registration and update operations for the database 20 are not executed, and the translated text extracted from the parallel translation database 20 is presented to the user.

The change in the processing step in this case is 101
˜109, 111, 114, 115, 123. It should be noted that the process may return from step 108 to step 105.

(E) When a bilingual translation is input and even if the input original text does not exist in the original text database 20a, the input original text is translated, and the translated text obtained by the translation and the input translation When and match, the parallel translation management data connecting the original text and the original text and the translated text already stored is registered in the parallel translation database 20, and the translated text obtained by the translation is presented to the user. Also in this case, “translating” is temporarily registered in the parallel translation database 20.

The change of the processing step in this case is 101
~ 107, 113, 111, 114-123.

(F) When a bilingual translation is input, and even if the input original text does not exist in the original text database 20a, the input original text is translated, and the translated text obtained by the translation and the input translation If does not match, the original sentence, the translated sentence obtained by the translation, and the parallel translation management data connecting the original sentence and the translated sentence are registered in the parallel translation database 20, and the translated sentence obtained by the translation is presented to the user. Also in this case, “translating” is temporarily registered in the parallel translation database 20.

The change of the processing step in this case is 101
~ 107, 113, 111, 114-123.

(G) When a parallel translation is input, and the parallel translation related to the input original text exists in the parallel translation database 20, but the parallel translation management data having the user dictionary ID that matches the input user dictionary ID is If it does not exist, the input original text is translated, and the translated text obtained by the translation,
When the translated text stored in the parallel translation database 20 matches, new parallel translation management data (translated text time stamp and user dictionary ID already existing) connecting the input original text and the already-stored translated text (equal to the translated text) Only) is registered in the parallel translation database 20, and the translated text obtained by the translation is presented to the user. Also in this case, “translating” is temporarily registered in the parallel translation database 20.

The change of the processing step in this case is 101
~ 107, 113, 111, 114-123.

(H) When a bilingual translation is input, the bilingual translation relating to the input original sentence exists in the bilingual translation database 20, but the bilingual translation management data having the user dictionary ID that matches the input user dictionary ID is stored. If it does not exist, the input original text is translated, and the translated text obtained by the translation,
If the translated text stored in the parallel translation database 20 does not match, the translated text obtained by the translation and the parallel translation management data connecting the input original text and the translated translation are registered in the parallel translation database 20 and obtained by the translation. The translated text is presented to the user. Also in this case, “translating” is temporarily registered in the parallel translation database 20.

The change in the processing step in this case is 101
~ 107, 113, 111, 114-123.

(I) When a bilingual translation is input, the bilingual translation related to the input original text exists in the bilingual translation database 20, and the bilingual translation management data having the user dictionary ID that matches the input user dictionary ID However, if the translated text defined by the parallel translation management data does not match the input translation, the input translation and the parallel translation management data linking the input original text and the translation are input to the parallel translation database 20.
The translated text that has been registered and entered is presented to the user.

The change in the processing step in this case is 101
˜111, 114, 115, 123. It should be noted that the process may return from step 108 to step 105.

(J) When a bilingual translation is input, the bilingual translation related to the input original sentence exists in the bilingual translation database 20, and the bilingual translation management data having the user dictionary ID matching the input user dictionary ID Exists and the translated text defined by the parallel translation management data matches the input translated text, the parallel translation database 20 is not registered and the input translated text (stored translated text) is presented to the user. .

The change in the processing step in this case is 101
˜109, 111, 114, 115, 123. It should be noted that the process may return from step 108 to step 105.

Next, the operation of the machine translation system of the first embodiment as described above will be explained using a concrete example.

When the three original sentences 1 to 3 shown in FIG. 7 are input to the machine translation device 17 in the "normal translation mode" without specifying the user dictionary, steps 101, 102 and 12 are executed.
The translation results are immediately presented to the user via 1, 123. FIG. 8 shows a case where the translation result is presented as a set of an original sentence and a translated sentence, and translated sentence 1, translated sentence 2, and translated sentence 3 correspond to original sentence 1, original sentence 2, and original sentence 3, respectively.

On the other hand, the three original sentences 1 to 3 shown in FIG.
When the original sentence 3 is input to the machine translation device 17 with the “example translation mode” added without the user dictionary being specified, the machine translation device 17 and the parallel translation database device 21 respectively respond to each of the original sentence 1 to the original sentence 3. The translation result shown in FIG. 8 is obtained in the same manner as in the case of being input to the machine translation device 17 in the "normal translation mode" by performing the processing of the above-mentioned case (a) involving translation sentence generation and parallel translation registration. To the user.

In this case, however, the parallel translation database 20 is searched and updated, and the content shown in FIG. 2 is registered in the original text database 20a.
The content shown in FIG. 3 described above is registered in the translated text database 20c, and the content shown in FIG. 4 described above is registered in the parallel translation management database 20b.

Here, the three original sentences 1 to 3 shown in FIG. 7 are input to the two machine translation devices 17 with the "example translation mode" added without specifying the user dictionary. Suppose that these were executed at about the same time. In this case, in the parallel translation database device 21, "translation in progress" is registered in the parallel translation database 20 in the processing of the search command arriving first, and based on "translation in progress" in the processing of the search command arriving later. After waiting for the processing, the parallel translation is registered in the parallel translation database 20 based on the update command from the machine translation device related to the search command that arrived first, and then the processing for the search command that arrived later is executed.

Generally, in the translation using only the system dictionary (even if the user dictionary is used), a good translation is rarely obtained, and the translated sentence is often post-edited. On the other hand, looking at the output (translated sentence) of machine translation, instead of directly changing the translated sentence, after changing the original sentence, registering a word in the user dictionary, or performing both at the same time, again By translating, pre-editing work may be performed to obtain a better translated text.
Further, pre-editing and post-editing may be used together.

Hereinafter, an example in which pre-editing and post-editing are used together will be described. Here, as shown in FIG. 9, original sentence 1 is corrected to original sentence 4, translated sentence 2 is corrected to translated sentence 4, and these three parallel translations (original sentence 4 and translated sentence 1, original sentence 2 and translated sentence 4, and original sentence 3) are corrected. It is assumed that the translated sentence 3) is input to the machine translation device 17 with the "example translation mode" added without specifying the user dictionary.

At this time, the machine translation device 17 and the parallel translation database device 21 perform the processing in the above case (f) involving the formation of a new translated sentence (translated sentence 5) for the parallel translation consisting of the original sentence 4 and the translated sentence 1. In the above case (i) with additional registration of the translation for the translation 2 consisting of the original sentence 2 and the translated sentence 4
Processing is performed, and the parallel translation database 20 is not updated for the parallel translation consisting of the original sentence 3 and the translated sentence 3. The above-mentioned case (j) is performed, and as a result, the translation result shown in FIG. 10 is presented to the user. To do.

By such processing, although not shown, the original sentence database 20a is updated to have the original sentence 4 added to the contents of FIG. 2, and the translated sentence database 20c has two records added to the contents shown in FIG. 11 is updated, and the parallel translation management database 20b is updated to the content shown in FIG. 12 in which two records are added to the content shown in FIG.

Each database 2 of the parallel translation database 20
When 0a, 20b and 20c have such contents, as shown in FIG.
The four original sentences 4, the original sentence 2, the original sentence 5 and the original sentence 3 shown in 3 are
The user dictionary shown in FIG. 14 (user dictionary ID = ws.00
1) is specified and the machine translation device 1 in "example translation mode"
7 is input.

At this time, the machine translation device 17 and the parallel translation database device 21 perform the processing in the case (b) in which only the parallel translation management data is added to the original sentence 4 and the translated sentence is translated into the original sentence 2 and the original sentence 3. When the translated text data and parallel translation management data are added (c), the original sentence 5 is processed by the above-mentioned case (a) with translation generation and parallel translation registration. The translation result shown in 15 is presented to the user.

By such processing, although not shown, the original sentence database 20a is updated to have the original sentence 4 in addition to the contents of FIG. 2, and the translated sentence database 20c is updated to the contents of FIG. Is updated to the content shown in FIG. 16 in which three records are added to the content shown in
The parallel translation management database 20b is updated to the content shown in FIG. 17 in which three records are added to the content shown in FIG.

Therefore, according to the first embodiment, the parallel translation database is provided to store the parallel translations translated in the past, and the parallel translations for the original texts having the same contents are stored in the parallel translation database 20 when the original texts or the parallel translations are input. It is determined whether or not the translation is executed depending on the determination result, and the input parallel translation, or the input original sentence and the translated translation thereof are registered in the parallel translation database 20. Therefore, it is possible to omit the waste of translating the same sentence in duplicate, and to present a unified translated sentence for the same sentence to the user.

When the update command is prioritized over the search command and translation is executed, "translation in progress" is registered in the parallel translation database 20 so that the processing of the search command which has reached "translation in progress" can be waited. As a result, even if multiple machine translation devices try to translate the same sentence at about the same time, it is possible to eliminate the waste of translating the same sentence redundantly, and to provide the user with a unified translation of the same sentence. Can be presented.

Further, according to the first embodiment, since the "normal translation mode" and the "example translation mode" are provided as the operation modes, the user can select the operation mode as appropriate and the usability is improved. If the user wants to present the translation result early or knows that the same sentence has never been translated (for example, immediately after the start of translation), select "Normal translation mode" If you want a unified translation result, you can select the "example translation mode".

Furthermore, according to the first embodiment, since the user dictionary ID is included in the bilingual translation data of the bilingual translation database 20, the input user dictionary ID and the user dictionary ID are different even if the bilingual translation already exists. Since the translation is performed even in the case, it means that the latest user dictionary with a better translation result is respected, and the translation result can be presented in which the translated words are consistent and unified.

(B) Second Embodiment Next, a second embodiment of the machine translation system according to the present invention will be described in detail with reference to the drawings. The functional block diagram of the second embodiment can also be represented by FIG. 1 according to the first embodiment described above.

In this second embodiment, the parallel translation database 20
Is always executed only when an update command is given.

That is, in the first embodiment, when the search command is given, the search result is
In the case of (3), the parallel translation database 20 was updated at that stage. However, in the second embodiment, even when the search result corresponds to the case (3), the update at this stage is not executed, and the parallel translation database 20 is updated when an update command is given thereafter. And

Therefore, the parallel translation database management means 19
When the search result corresponds to the case (3) in the first embodiment, the following case (3a) is performed.

(3a) When parallel translation is input, the input original text exists in the original text database 20a, and
When parallel translation management data related to the original sentence having a user dictionary ID that matches the input user dictionary ID exists, the translated sentence database 20c is searched based on the translated sentence ID, and the searched translated sentence and the input translated sentence are If not, the input translated text is returned as a translation result to the machine translation device 17 of the search request source via the communication means 18.

FIG. 18 shows the overall processing of the system of the second embodiment, and the same or corresponding steps as in FIG. 6 described above are designated by the same reference numerals.

The processing up to step 108 is the first
It is similar to the embodiment. In the first embodiment, the condition determination is performed after this and the parallel translation database 20 is appropriately updated, but in the second embodiment, the process immediately proceeds to step 111 and the search result is transmitted.

The processing (steps 114 and 115) performed by the machine translation device 17 to determine whether or not an original sentence related to the input exists in the parallel translation database 21 is the same as in the first embodiment.

However, in the second embodiment, even if there is a parallel translation related to the input in the parallel translation database 20 (“no data”).
Even if a result other than the above is received), the process does not immediately proceed to the translation result presentation processing step 123, and it is determined whether or not the update processing of the input bilingual translation without translation is necessary (step 109). ). That is, it is determined whether or not the case (3a) is among the cases where the search result other than “no data” is given. For example, at the time of parallel translation input, such determination can be performed depending on whether or not the search result is the input translation itself.

Then, if the search result is (3a), the process proceeds to the update command transmitting process of step 117, and if the search result is other than case (3a) (whether "no data" has already been confirmed). ), And the process proceeds to the translation result presentation processing step 123.

Also in the second embodiment, the processing after step 117 is the same as that of the first embodiment, and the description thereof is omitted.

As described above, in the second embodiment, when the parallel translation database 20 needs to be updated, the update command is always issued to update the parallel translation database 20.

The second embodiment is different from the first embodiment in the flow of processing when the input bilingual text is registered in the bilingual translation database 20 without translating the input original text. The second embodiment can also achieve the same effect as the first embodiment.

(C) Third Embodiment Next, a third embodiment of the machine translation system according to the present invention will be described in detail with reference to the drawings. The functional block diagram of the third embodiment can also be represented by FIG. 1 according to the first embodiment described above.

In the third embodiment, in addition to the above-mentioned "normal translation mode" and "example translation mode", a "translation insertion mode" is provided as an operation mode.

The “translation input mode” is an operation mode in which the translation is registered in the translation database 20 and neither the translation database 20 is searched nor the input original sentence is translated. For example, when the final parallel translation obtained by post-editing only the translated text is stored in the parallel translation database 21, the processing time is shorter in the “parallel translation input mode” than in the “example translation mode” that involves a search operation. So practical.

FIG. 19 shows the overall processing of the system of the third embodiment, and the same or corresponding steps as those in FIG. 6 described above are designated by the same reference numerals.

In the third embodiment, in the operation mode determining step 302a, it is determined which one of the three modes. If it is the "normal translation mode", the processing from step 121 onward is performed as in the first embodiment.
If it is the "example translation mode", the processing from step 103 onward is performed as in the first embodiment.

In the case of the "translation input mode", the processing from step 117 onward is performed. That is, an update command including the input original sentence and translated sentence is immediately created and transmitted to the parallel translation database device 21 to be updated (input). Note that in the user presentation process (step 1
23a), the translation result is presented in the "normal translation mode" and the "example translation mode", and the "translation input mode" is displayed.
Then, indicate that registration is complete.

According to the third embodiment, the same effect as that of the first embodiment can be obtained. In addition to this, the third
According to the embodiment, since the “translation input mode” for simply registering (inputting) the parallel translation in the parallel translation database 20 is also provided, the usability for the user can be further improved.

(D) Fourth Embodiment Next, a fourth embodiment of the machine translation system according to the present invention will be described in detail with reference to the drawings. Here, FIG. 20 is a functional block diagram showing the fourth embodiment, and the same or corresponding portions as in FIG. 1 are designated by the same reference numerals.

The fourth embodiment is a system in which a document search device 30 is further connected to the network 22.

The document retrieving device 30 includes a document storing means 33 for storing documents and a document retrieving means 32 for retrieving a predetermined document from the document storing means 33 in response to a request from an external device.
And communication means 31 for establishing and disconnecting a link with an external device and exchanging information.

The machine translation device 17 of the fourth embodiment is provided with a document search request means 34 for requesting the document search device 30 to search for a document and receiving the returned document.
Upon receiving the search document, the document search requesting unit 34 inputs the received document to the input unit 11 in the "example translation mode".

That is, in the fourth embodiment, the document search requesting means 34 is operated by the document search device 3 in response to a search instruction.
The document extracted from 0 is immediately input to the machine translation configuration to obtain the translation result. The input means 11 has
It is also possible to input by the same input method as in the first embodiment.

According to the fourth embodiment, the same effect as that of the first embodiment can be obtained. In addition to this, the fourth
According to the embodiment, even when a document (a group of a plurality of sentences) desired to be translated exists in another device, the translation result of the document can be promptly obtained without the user separately inputting the search instruction and the translation instruction. Can be obtained.

(E) Other Embodiments (1) In the above embodiment, a system provided with the "normal translation mode" and the "example translation mode", or the "normal translation mode", the "example translation mode" and the "example translation mode" The system with "translation input mode" is shown, but the system with "example translation mode" and "translation input mode",
The system may have only the "example translation mode".

(2) In the above embodiment, an English-Japanese translation system was shown, but it is needless to say that it can be applied to a translation system between other two languages, and the system can handle translation directions of two or more. It may be.

(3) In the above embodiment, the unit of judgment of the input identity and the unit of updating the parallel translation database are shown as one sentence. However, the input of the same input is made in units of a plurality of sentences (for example, paragraphs). It may be possible to judge (search) the sex and update the parallel translation database.

For example, when translating a revised version of the operation manual in which most of the sentences translated in the past match those in the past, the processing in such a large unit is preferable.

(4) The user dictionary ID has only to define the attributes of the user dictionary, and the temporal information is not limited to the version number and may be the update date or the creation date.

Further, a plurality of user dictionaries may be designated.
When a plurality of user dictionaries are designated, it is desirable to set the case where the IDs of all the user dictionaries are the same as the condition of the identity regarding the user dictionaries.

(5) In the above embodiment, when the original sentence and the user dictionary ID are the same, but the translated sentences are different, the translation is registered in the parallel translation database 20 as different parallel translations. Only one parallel translation having the same number may be registered in the parallel translation database 20. In this case, since what is input to the parallel translation database device 21 later is often used for post-editing processing,
All you have to do is register it. By doing so, the translated text time stamp in the parallel translation management database 20b can be eliminated and the storage capacity can be reduced.

For example, when registering the fifth record from the top in FIG. 12, the second record from the top in FIG. 12 is deleted. At this time, the second record from the top in FIG. 11 is also deleted.

(6) Further, the system may be such that the user dictionary ID is not used for the search process and the update process in the parallel translation database device 21.

(7) The option items of the machine translation function other than the user dictionary ID may be used for the search process and the update process in the parallel translation database device 21 instead of or in addition to the user dictionary ID. For example, setting item information as to whether or not to translate “you” in the English operation manual into “you” may be included in the parallel translation database 20 and used for the search processing and update processing in the parallel translation database device 21. .

(8) The structures of the original sentence database and the translated sentence database are not limited to the tree structure. For example, an array with a hash key as an index can be used as a data structure.

(9) In the system having the “parallel translation input mode”, the system may not allow the parallel translation input in the “example translation mode”.

(10) In the above embodiment, even when the bilingual database 20 is updated, the original text database 20 is updated.
Although it is shown that a search such as a is performed again, if the same original text is registered when returning "no data" in the search result, the original text ID is also included in the search result, and the original text is also included in the update command. The ID may be included so that the original text search may be omitted when updating.

(11) In the above embodiment, when the update is required, the translated text is presented to the user after the update is finished, but the user's presentation process may be executed first. . In this case, a message asking the user whether or not to execute the update may be displayed on the screen presented to the user, and the user may select whether or not to execute the update. Further, even when the translated text is presented to the user later, a message asking the user whether or not to execute the update may be displayed, and the user may select whether or not to execute the update.

[0164]

As described above, according to the present invention, a parallel translation database is provided to store previously translated parallel translations, and at the time of input, whether the parallel translation corresponding to the input is stored in the parallel translation database or not. Depending on the judgment result, the translated text extracted from the bilingual database without executing translation is presented, and the input bilingual text, the input original text and its translated text, and the translation in progress are indicated. Since the information to be registered is registered in the bilingual database, it is possible to eliminate the waste of duplicate translation even if the translation of the same sentence occurs at almost the same time, and to provide the user with a unified translation of the same sentence. A machine translation system that can be presented can be realized.

[Brief description of drawings]

FIG. 1 is a functional block diagram of a first embodiment.

FIG. 2 is an explanatory diagram of a storage example (part 1) of the original text database according to the first embodiment.

FIG. 3 is an explanatory diagram of a storage example (No. 1) of a translated text database according to the first embodiment.

FIG. 4 is an explanatory diagram of a storage example (No. 1) of the parallel translation management database according to the first embodiment.

FIG. 5 is an explanatory diagram of the structure of a message sent and received between both devices of the first embodiment.

FIG. 6 is a flowchart showing the overall processing of the first embodiment.

FIG. 7 is an explanatory diagram of an input example (No. 1) of the first embodiment.

FIG. 8 is an explanatory diagram of an output example (No. 1) of the first embodiment.

FIG. 9 is an explanatory diagram of an input example (No. 2) of the first embodiment.

FIG. 10 is an explanatory diagram of an output example (No. 2) of the first embodiment.

FIG. 11 is an explanatory diagram of a storage example (No. 2) of the translated text database according to the first embodiment.

FIG. 12 is an explanatory diagram of a storage example (No. 2) of the parallel translation management database according to the first embodiment.

FIG. 13 is an explanatory diagram of an input example (No. 3) of the first embodiment.

FIG. 14 is an explanatory diagram of an example of a user dictionary according to the first embodiment.

FIG. 15 is an explanatory diagram of an output example (No. 3) of the first embodiment.

FIG. 16 is an explanatory diagram of a storage example (No. 1) of the original text database according to the first embodiment.

FIG. 17 is an explanatory diagram of a storage example (No. 1) of the original text database according to the first embodiment.

FIG. 18 is a flowchart showing the overall processing of the second embodiment.

FIG. 19 is a flowchart showing the overall processing of the third embodiment.

FIG. 20 is a functional block diagram of a fourth embodiment.

[Explanation of symbols]

11 ... Input means, 12 ... Parallel database search means (parallel search request means), 13 ... Translation means, 14 ... Parallel database update means (parallel update request means), 15 ... Output means, 16, 18 ... Communication means, 17 ... Machine translation device, 19
... parallel translation database management means, 19a ... command queue, 20 ... parallel translation database, 20a ... original text database, 20b ... parallel translation management database, 20c ... translated text database, 21 ... parallel translation database device, 22 ... network.

Claims (5)

[Claims] 1. A machine translation system in which one or more machine translation apparatuses and a parallel translation database apparatus common to these machine translation apparatuses are connected to a communication network, wherein each of the machine translation apparatuses is at least An input means for fetching an input including an original sentence, and a search instruction as to whether or not a parallel translation for the input fetched by the input means is stored in the parallel translation database device are given to the parallel translation database device, and a search result from the parallel translation database device is given. Parallel translation search request means for receiving and determining the necessity of translation of the input original text, translation means for translating the input original text, parallel translation information consisting of the input original text and the input translation, or input To the parallel translation database device, the update instruction including the translated original text and the parallel translation information composed of the translated text translated by the translation means. Depending on the search result and the input content from the parallel translation database device, the translation included in the search result or the translation translated by the translation means is output as the translation for the input original sentence. The parallel translation database device having output means, and the parallel translation database storing the parallel translation, and the parallel translation for the input captured by the input means of the machine translation device in response to a search instruction from any of the machine translation devices. Is stored in the parallel translation database, returns the search result, and updates the parallel translation database in response to an update instruction from any of the machine translation devices, and waits for a plurality of instructions. It manages in a queue and gives priority to the one that arrived first in the competition of the same kind of instruction, and also the competition between the search instruction and the update instruction. Machine translation system, characterized in that it includes a translation database management means for giving priority to the update instruction. 2. The parallel translation database management means stores a search result for activating execution of translation in response to a search instruction,
When replying to the machine translation device that issued the search instruction, the translation in the translation database related to the search instruction is added with "translation in progress", and the translation for "search in progress" is obtained in the search for the search instruction. If the search instruction is given, the search instruction is reinserted at the back of the instruction queue, and if the update instruction including the translation corresponding to the translation to which “Translation is in progress” is given, the above translation database is searched. The machine translation system according to claim 1, wherein "translating" is deleted while updating. 3. As an operation mode of the machine translation device,
An example translation mode in which the parallel translation database device is searched and updated, and a normal translation mode for simply translating and inputting the input original sentence are provided, and when the normal translation mode is instructed and input, the input means 3. The machine translation system according to claim 1, wherein the translation means translates the original sentence taken in by the translation means, and the obtained translated text is output by the output means. 4. The operation mode of the machine translation device includes:
A parallel translation input mode for registering parallel translations in the parallel translation database is provided, and when the parallel translation input mode is instructed and input is performed, the parallel translation update request means issues an update instruction including the parallel translation information including the original text and the translated text captured by the input means. The machine translation system according to any one of claims 1 to 3, wherein the machine translation system is provided to the parallel translation database device to update the parallel translation database. 5. Including user dictionary information in the input,
The bilingual translation stored in the bilingual database also includes the user dictionary information, and when the bilingual database management means searches according to the search instruction, the original sentence or bilingual translation taken in by the input means matches the combination of the original sentence and the translated sentence. Even if the corresponding translation exists in the parallel translation database, if the user dictionary information is different, a reply that the corresponding parallel translation does not exist in the parallel translation database is returned.
4. The machine translation system according to any one of 4.