JP2875100B2 - Translation equipment - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a translation apparatus capable of improving translation efficiency and translation accuracy by pre-editing an original sentence at the time of translation.
It is about the improvement.

[0002]

2. Description of the Related Art Conventionally, there are the following machine translation apparatuses. This machine translator has a keyboard, CP
U, a memory, a CRT (cathode ray tube), and the like, and a source language input from a keyboard is translated into a translation module for translation processing prepared in advance as a program.
It is input as data under the control of U. The translation module translates the input source language into a target language using dictionary data, grammar rules, tree structure conversion rules, and the like stored in a memory in advance. The target language obtained by translation is CR
The translation result is output to T or the like, and the translation result is displayed to the user. If this machine translation device is a device that translates English into Japanese, the source language is English,
The target language will be Japanese. Also, an input sentence to the translation device composed of a source language is called an original sentence, and an output sentence composed of a target language, which is an output result of the translation device, is called a translated sentence.

In the above-described translation module, the original sentence is morphologically analyzed. By morphological analysis, each word contained in the original sentence is extracted by specifying its part of speech. When performing a syntactic analysis using a dictionary and grammatical rules from the parts of speech of words obtained as a result of morphological analysis, the translation module outputs a plurality of translation results by assembling all structures of syntax. Therefore, it is necessary for the user to appropriately select which of the plurality of translation results is the correct translation. Further, since the translation module is required to assemble all the structures of the syntax, it takes a long time for the syntax analysis in the translation module, resulting in a decrease in translation efficiency. In order to alleviate this decrease in translation efficiency, the user performs pre-editing on the original text before performing the translation, so that the candidates of the syntax structure in the translation module can be limited or the syntax analysis can be performed. Translation devices that can increase the speed have been proposed. The pre-editing process referred to here is to prepare a plurality of special character strings and symbol strings that are not usually used in the original text, define them in advance as "pre-edit symbols", This means that the translation processing is supported by specifically specifying the part of speech of a specific word or by specifically specifying the syntax structure during the translation processing in the translation module by appropriately inserting it into the original sentence before inputting it into the translation module.

[0004]

As described above, conventionally, there has been a translator for improving translation efficiency by using pre-edited symbols. However, this translator has the following two problems. Had a point.

(1) There is a problem that the "pre-edited symbol" may coincide with a character string or a symbol string used in the original text.

More specifically, the pre-editing symbol is predetermined so as to be a character string / symbol string having a very low possibility of appearing in a normal original text. It is not unlikely that a sequence will appear in ordinary text. In particular, if the original text contains mathematical formulas,
It is very likely that a part of the formula will match the preedit symbol. If the original sentence whose character string or symbol string matches the pre-editing symbol is translated using a translation device, the translation module will use the symbol string or character that matches the pre-editing symbol in the original sentence. The translation process proceeds while recognizing the column as if it were a pre-editing symbol, causing a problem that a correct translation is not performed.

(2) The work of inserting the "pre-editing symbol" in the original text in advance is troublesome.

More specifically, in a machine translator capable of processing pre-edited symbols, frequently-used pre-edited symbols are translated by ignoring a specified range in a sentence by dividing a sentence or by translating a specified range in a sentence. This symbol specifies a function to output the source language without translating the range of. These functions are often used because they are functions that are easier to specify intuitively than word part of speech or sentence pattern specification. When such a function is designated, the place where the pre-editing symbol is inserted in the original text often coincides with the place where the description symbol based on the notation rule of the source language already exists. For example, assuming that the source language is English, a conventional machine translator performs a parsing in units of "one sentence" at the time of translation. The term “one sentence” here usually consists of a set of words between a period (.) And a period.
Even if symbols such as commas (,), colons (:), and semicolons (;) exist in this "one sentence", they are regarded as descriptive symbols inside the "one sentence" during parsing, and "one sentence" Are not considered to be boundaries. However, in the original text, these symbols are used to express some semantic unit. Therefore, if "one sentence" is delimited at these symbols, that is, inserting a pre-editing symbol that means dividing the sentence and translating it with the translation module, it is better than translating the original text without inserting the pre-editing symbol. Parsing is performed accurately or at high speed, and a more accurate translation can be efficiently obtained. This operation of sequentially inserting pre-editing symbols that indicate sentence division into the original sentence is nothing more than intentionally inserting a pre-editing symbol where a descriptive symbol that indicates sentence division already exists. Therefore, the operation of inserting the pre-editing symbol causes a problem of giving the user a sense of labor.

[0009] Therefore, an object of the present invention is to provide a translation apparatus which inputs an original sentence and outputs the translated sentence, when a user adds a symbol string for assisting translation to the original sentence and translates the original sentence. It is an object of the present invention to provide a translation apparatus which is always performed, and which can reduce the troublesome operation of adding the symbol string to the original sentence itself.

[0010]

SUMMARY OF THE INVENTION A translation device according to the present invention is a device for translating a given original sentence into a predetermined language and outputting a translated sentence for the original sentence. It comprises a column inserting means, a storage means, and a translation control means.

The symbol string inserting means is configured to insert, via the input means, a symbol string for supporting the translation process into the original text. And the storage means at least
One or more translation support information, and each of the translation support information
A predetermined symbol string and at least one predetermined
It is configured to store an arbitrary symbol string to be written .

The above-mentioned translation control means inputs the original sentence into which the symbol string has been inserted by the symbol string inserting means and converts the symbol string in the original sentence into the corresponding symbol string in the storage means which is made valid.
The translation processing is performed while interpreting based on the translation support information.

Then, in the storage means, the translation support information
One or more arbitrary strings were written for each of
In the case, the corresponding predetermined symbol string is invalid and written
One or more optional symbol strings are valid and one or more optional
If the symbol string is not written, the corresponding predetermined symbol string is
It is configured to be defined as valid .

[0014]

The translation apparatus according to the present invention is configured as described above, and the translation control means inputs the original sentence into which the symbol string is inserted by the symbol string insertion means, and stores the symbol string in the original sentence.
This original text is translated while being interpreted based on the translation support information corresponding to the corresponding symbol string that is regarded as valid . Oite in the storage means to be referenced at this time, its translation support information
When one or more arbitrary character strings are written for each
The corresponding predetermined symbol string is invalidated and written
One or more optional strings are valid and one or more optional
If no string is written, the corresponding symbol string is valid
Since that may be defined to be a, a predetermined symbol sequence in the storage unit that be present already as a string in the original text string is mistaken for symbol string to assist the aforementioned translation process Can be avoided. Conversely, a character string that already exists in the original sentence can be processed as a symbol string to support the above-described translation processing by setting it as an arbitrary symbol string in the storage unit .

[0015]

Embodiments of the present invention will be described below in detail with reference to the drawings.

The machine translator applied to the present embodiment inputs English as a source language, translates it,
Although the translation device outputs Japanese as a target language, the source language and the target language are not limited to English and Japanese described in the present embodiment.

FIGS. 1A and 1B are block diagrams of a machine translation apparatus and a translation module applied to an embodiment of the present invention.

FIG. 1A shows a block configuration of a machine translation apparatus. The machine translation apparatus includes a CPU (central processing unit) 1 for centrally managing and controlling the apparatus, a main memory 2 for storing data including a program required for translation processing, a CRT (cathode ray tube). Or L
A display unit 3 including a CD (liquid crystal display), a keyboard 4 for external data input, a memory 6 and a translation module 5 for translating a source language input into a target language while accessing the memory 6 are included. These units are mutually connected for communication via a bus.

The memory 6 includes areas M1 to M4. Area M
1 is a dictionary D1, an area M2 is a grammar rule D2, an area M3.
Stores a tree structure conversion rule D3 in advance, and the area M4 is used as a work area at the time of translation processing.

The structure of the translation module 5 is shown in FIG. The translation module 5 is a dictionary lookup morphological analyzer 1
1. Syntax analyzer 12, converter 13, and translated sentence generator 1.
4 inclusive.

Next, the translation operation of the machine translation apparatus will be described with reference to FIG.

First, an original sentence input via the keyboard 4 is sent to the translation module 5 for each sentence under the control of the CPU 1. The translation module 5 uses the dictionary D1, grammar rules D2, and tree structure conversion rules D3 stored in the memory 6 to translate the given source sentence of one sentence into the target language. To explain in detail, FIG.
As shown in (1), the source language of one sentence is first subjected to morphological processing by the dictionary lookup morphological analysis unit 11 using the dictionary D1 of the memory 6 and divided into a series of morphemes. A morpheme is
It is the smallest linguistic unit that composes a sentence, and is a word, prefix, and suffix. At the time of this morphological processing, grammatical information such as a part of speech for each of the divided words and a translated word are obtained,
Further, tense / person / number etc. are analyzed. When the morphological processing of the sentence is completed, the syntactic processing is performed on the assumption that the morphemes constituting the sentence are correctly extracted in the syntactic analysis unit 12. The syntax analysis unit 12 clarifies the grammatical structure of the sentence according to the dictionary D1 and the grammatical rule D2 of the memory 6, and determines a structure analysis tree indicating the dependency relationship between the words. The conversion unit 13 converts the structure analysis tree obtained by the preceding syntax analysis unit 12 into a structure analysis tree for the translated sentence using the tree structure conversion rule D3 of the memory 6.
Then, in the translated sentence generation unit 14, an appropriate particle or auxiliary verb is added as a translated sentence in the target language,
The obtained translations are sequentially output from the translation module 5 as a target language, and output to the display unit 3 and the like under the control of the CPU 1. At the time of outputting the translated sentence, the corresponding original sentence is also displayed. The user can correct the translation result via the keyboard 4 while viewing the displayed two sentences.

FIG. 2 is a processing flowchart for explaining the operation of translating an original sentence including a pre-edited symbol in the machine translation apparatus according to the embodiment of the present invention.

FIGS. 3A and 3B are diagrams showing an example of preedit symbols applied to the embodiment of the present invention.

Referring to FIG. 3, these preedit symbols are roughly classified into two types according to their form. The first type is shown in FIG. 3A, and is used to modify the word by attaching it to the beginning of the word. For example, as shown in FIG. When the word is added, the word is designated as a part of speech and the original text is edited so as to be translated as a noun. Also, a pre-editing symbol v1 , The sentence is designated as a sentence pattern, and the original sentence is edited so that it is translated as the first sentence pattern in English during translation. Further, a pre-editing symbol uj is added at the beginning of a word. , The original text is edited so that the word is translated using a dictionary specially prepared by the user.

The second type of pre-editing symbol is used for inserting a pre-editing symbol between words and separating the sign before and after the symbol with some meaning. For example, referring to FIG. 3 (b), a portion enclosed in a pre-editing symbol (and (pre-editing symbol)) in the original text is pre-edited so as to specify that it is excluded from translation. You.

In the pre-editing process of the original sentence using the above-mentioned pre-edited symbol, the original sentence is displayed on the display unit 3 and the user inserts the pre-edited symbol at an arbitrary position of the original sentence displayed via the keyboard 4. It is done by doing. The pre-edited original text is provided to the translation module 5.

In this embodiment, the pre-editing symbols have the pre-editing symbols shown in FIGS. 3A and 3B as the prescribed pre-editing symbols. Is defined as Therefore, the user can cause the original text including the newly defined preedit symbol to be translated by inserting the newly defined preedit symbol into the original text.

FIG. 4 is a diagram showing an example of the contents of a buffer in which only prescribed preedit symbols are stored.

FIG. 5 is a diagram showing an example of the contents of a buffer in which pre-edit symbols defined and defined by the user are stored.

The prescribed preedit symbols shown in FIGS. 3A and 3B are stored in the buffer D as shown in FIG. The buffer D stores preedit symbols in a table format, and the table includes items I1 to I7. The item I1 has an identification number ID for identifying the pre-editing symbol, the item I2 has a prescription symbol K0 representing the pre-editing symbol, and the item I3 has a translational meaning MA represented by the corresponding prescription symbol K0. Stored. The buffer D includes a user-defined flag F in the item I4 and a user-defined symbol K in each of the items I5 to I7.
1 to K3 are stored. User-defined symbol K1
To K3 are pre-editing symbols that the user arbitrarily defines by replacing the corresponding meaning MA with the corresponding prescribed symbol K0. When the pre-editing symbol defined by the user is stored in any of the symbols K1 to K3, the corresponding user-defined flag F is set to "1". Conversely, the flag F is represented by the symbol K1.
Is reset to 0 when none of .about.K3 is stored. The buffer D is stored in the main memory 2 or the memory 6 and displayed on the display unit 3 in response to a user's desired input from the keyboard 4. The user can operate the keyboard 4 to arbitrarily set each of the user-defined symbols K1 to K3 in the buffer D displayed on the display unit 3. When the user does not define the preedit symbol, the buffer D is as shown in FIG. At this time, the user-defined flags F have been reset to 0 for all the prescribed symbols K0, and all of the user-defined symbols K1 to K3 are also blank.

Now, it is assumed that a new form of the preedit symbol is defined by the user. The pre-editing symbol whose shape is newly defined is the meaning MA of the number ID = 1 in the buffer D of FIG.
Is a “noun” and a “split translation designation symbol” with a number ID = 21, and the form of a preediting symbol of a “noun” is “no”
un , And the form of the "split translation designating symbol" is defined as ":" and ",". As a result,
The buffer D in FIG. 4 is updated like the buffer D in FIG.

Here, the user-defined symbol K1 of the "noun" having the number ID = 1 is "noun" Is stored as the number ID
= 21, the user-defined symbols K1 and K2 of the "split translation designation symbol" are defined as ":" and ",", respectively, and stored. The CPU 1 sets the corresponding user-defined flag F to 1 in response to the user newly specifying any of the user-defined symbols K1 to K3. Therefore, as shown in the buffer D of FIG.
The user definition flag F of 1 and number ID = 21 is set to 1. Based on the set state of the user-defined flag F, it is specified that a user-defined symbol exists for these two types of preedit symbols during the translation processing, and the corresponding prescribed symbol K0 becomes invalid. In this case, the buffer D in FIG.
, The user-defined flag F of the number ID = 1 is “1”
, The stipulation symbol K0 = “n” corresponding to the number ID = 1 No longer has the meaning as a preediting symbol. Similarly, the prescribed symbol K0 = "//" corresponding to the number ID = 21 also has no meaning as a pre-edited symbol. Note that the definition symbol K0 corresponding to each of the number IDs 2 to 20 for which the user definition flag F is not set to "1" still has the meaning as a pre-editing symbol.

The translation processing in the case where the pre-edited symbol is defined by the user in advance according to the processing flow of FIG. 2 will be described in the following first and second embodiments.

First, a first embodiment will be described.

FIG. 6 is a diagram showing an example of the contents of a buffer for storing original text according to the first embodiment of the present invention.

FIG. 7 is a diagram showing an example of a screen display of the result of translating the original sentence of FIG.

FIGS. 8 (a) and 8 (b) show examples of the contents of a buffer in which information on the types of preedit symbols included therein and their locations are stored in the original text of FIG. FIG. 5 is a diagram showing a comparison between a case in which is defined and a case in which it is not.

In the first embodiment, a character string that matches a prescribed pre-editing symbol is present in the original text to be translated, and translation processing is performed without recognizing the matching portion as a pre-editing symbol. This is an example in the case of the purpose.

In the first embodiment, an original sentence as shown in FIG. 6 is stored in the buffer S1 of the memory 2 or 6. "N" in the original text of FIG. The symbol “symbol” is replaced with “n” at the time of translation. It should be translated as "symbol".
That is, when trying to translate this original sentence by the translation device, the character string “n” in this original sentence is used. Is not recognized as a pre-editing symbol. In this translation device, the user-defined symbol "noun ], The character string “n” in the original sentence Can be translated without recognizing preedit symbols. This will be described below.

First, before translating the original sentence of FIG. 6, the user-defined symbol K1 =
"Noun Is defined. As a result, the buffer D
The number I corresponding to the pre-editing symbol representing "noun"
The data of D (= 1) is updated as shown in FIG. 5 and the reference symbol K0 (“n ") Is invalid. Further, the CPU 1 sets the corresponding user-defined flag F to “1” in response to the user-defined symbol K1 having the number ID = 1 being set.

Next, referring to the processing flow of FIG. 2 , in step N1 (abbreviated as N1 in the figure), the original text to be translated is stored in the buffer S1 as shown in FIG. The original text is set by the user through a keyboard 4 through key input, or is stored in advance in the memory 6 or the like.
PU1 reads from the memory 6 and sets it in the buffer S1. Next, the translation module 5
The original sentence set to 1 is input and translated into the target language by the processing of the following step N2.

In the next step N2, the buffer P of the memory 2 or 6 shown in FIG.
) . The buffer P is a buffer for storing, as a result of detecting a pre-edited symbol from the original sentence of FIG. 6, the type of the pre-edited symbol and its location in the original sentence, and a pre-edited symbol for translation processing. It is used as information. In this buffer P, the preedit symbol is not in its form but in FIG.
Alternatively, it is stored as the numerical value of the corresponding number ID shown in FIG. The preedit symbols are of the type "modify word" shown in FIG. 3 (a) and those of the type "inserted between words" shown in FIG. 3 (b). There is. Therefore, the buffer P is configured so that both types of edit symbols can be stored in a distinguishable manner. That is, FIG.
In the buffer P of FIG. 3A, the preedit symbol of the type shown in FIG. 3A is stored in the column of the preedit symbol whose position number LN is an odd number, and the preedit symbol of the type whose position number is an even number. Stores a preedit symbol of the type shown in FIG.

In the next step N3, the buffer S
1 is divided into words and stored in buffer P
Is stored in the column of the word W. At this time, the pre-edited symbol added to the word in the form of a suffix or prefix (not separated by a space or the like) is stored in the word W column as it is added to the word. At this point, buffer P is
The word W is stored as shown in FIG.

In the next step N4, the buffer D in FIG.
In order to sequentially search for the pre-edit symbol in step (i), an initial value (=
1) is set.

The loop processing of the following steps N5 to N13 is repeatedly executed while incrementing the variable i by one until the variable i reaches the final value of the number ID of the buffer D. Then, in this loop processing, it is determined whether or not any of the user-defined symbols K1 to K3 is set in the buffer D for each number ID (variable i). If set, the buffer P is processed so that it is not regarded as a pre-edited symbol even if the corresponding prescriptive symbol K0 is present in the original text. Conversely, any of K1 to K3 is set in the buffer D. If not, the corresponding regulatory symbol K
If a 0 is found in the original text, it processes the buffer P to consider it as a pre-editing symbol. After completion of the loop processing of steps N5 to N13, step N
At 14, the buffer S using the information of the buffer P
1 is translated and the result is shown in FIG.
Is displayed on the display unit 3 as follows.

The loop processing including the above-described steps N5 to N13 will be described. From step N5, the buffer D is searched in the following order.

When the variable i = 1, the user-defined flag F corresponding to the number ID = 1 is set to "1" (see FIG. 5), and the process shifts to the next step N7. In step N7, the user-defined symbol K1 is detected from the original sentence stored in the buffer S1 for the pre-edit symbol with the number ID = 1. Here, as shown in FIG.
The user-defined symbol K1 corresponding to “1” is “noun ”, The character string of the original text in the buffer S1 is changed to“ noun "
Search for the string As a result, the character string “noun” is included in the original text of the buffer S1. Does not exist, the result of the determination in the next step N8 is that the contents of the buffer P in FIG. 8A are not updated, and the process proceeds to step N12. In the process of step N12, since the variable i has not reached the final value of the number ID (21 in FIG. 5), the variable i is incremented by 1 in the next step N13, and the process returns to step N5 again.

When the variable i = 2, the user-defined flag F corresponding to the number ID = 2 of the buffer D is "0" (reset) according to FIG. Proceed to processing. In the process at step N6, the regulation symbol K0 corresponding to the number ID = 2, “v Is searched for in the original text of the buffer S1. As a result, the character string “v” appears in the original text of the buffer S1. Does not exist, the result of determination in the next step N8 is
The process moves to the process of step 2.

Hereinafter, when the variable i is from i = 3 to 21 (the final value of the number ID), the steps N5 → N6 → N
8 → N12 → N13 → N5 or a loop process including steps N5 → N7 → N8 → N12 → N13 → N5 is executed. As a result, in the original text of the buffer S1, the number ID = 3 to 21 Since there is no pre-editing symbol corresponding to each of the number IDs of FIG.
The state becomes (a), which does not change from the state immediately after the processing of step N3. Therefore, in the processing of the next step S14 after the end of the loop processing, the information of the buffer P as shown in FIG. The original text is translated.

As described above, according to the first embodiment, the pre-editing symbol meaning "noun" is redefined by the user using the user-defined symbol K1 to something other than the stipulated symbol K0. Thus, the same character string as the prescription symbol K0 meaning “noun” (“n )) Can be subjected to a translation process that does not interpret the character string portion as a pre-editing symbol for an original sentence that does not have the meaning as a pre-editing symbol.

In order to compare with the case where the user-defined symbol in the first embodiment is defined in advance, the user edits the pre-edited symbol "noun". Next, a case where the original text stored in the buffer S1 in FIG. 6 is translated in the same manner as in the first embodiment in a state in which "."

When the user has not defined the preedit symbol, the buffer D is set as shown in FIG.

In the processing flow of FIG.
In the processing from Step N4 to Step N4, the processing proceeds in the same manner as described above. When the processing up to step N4 is completed, the buffer P is set as shown in FIG.

In the loop processing including steps N5 to N13, the processing proceeds as follows.

First, when the variable i = 1, the process proceeds to step N6 because the user definition flag F corresponding to the number ID = 1 is 0. In step N6, a prescription symbol K0 is detected from the original text of the buffer S1 with respect to the pre-edit symbol having the number ID = 1. Here, as shown in FIG. 4, the defining symbol K0 corresponding to the number ID = 1 is “n”. "
The character string "n" from the original sentence of the buffer S1 shown in FIG.
Is detected. As a result, the character string “n” appears in the original text of the buffer S1. Exists, so that the processing of step N9 is executed via the determination processing of the next step N8.

In the process of step N9, the character string "n" in the original sentence in the process of step N6 described above. Is detected as the character string "n" ID number corresponding to "
(The value of the current variable i) "1" is set to the previous edit symbol value KF1 corresponding to the position number LN = 3 of the buffer P. Then, the process proceeds to step N10.

In the process of step N10, the character string "n" is added to the word W corresponding to the position number LN = 3 in the buffer P.
Is included. Here, word W (= n) corresponding to position number LN = 3 symbol)
Contains the string "n Is included, so step N
The process proceeds from the process of step 10 to the process of the next step N11, and the character string Is rewritten to the word W that does not include the word "." That is, "n symbol "→" symbo
The word W corresponding to the position number LN shown in FIG. 8A is updated. As a result, the column of the number LN = 3 of the buffer P is set as shown in FIG. 8B, including the processing in step N9.

Hereinafter, for each of the cases where the variables i = 2 to 21 (the final value of the number ID), steps N5 → N6 → N8
The loop processing of → N12 → N13 → N5 is repeated,
As a result, in the original text of the buffer S1, the number ID = 2 to 2
Since none of the pre-edit symbols up to 1 is detected, the buffer P is finally set to the state shown in FIG. In the process of the next step N14, the information of the buffer P shown in FIG.
The translation process is performed (assuming that a pre-editing symbol for specifying “noun” is added to “mbol”).

As described above, when the user does not define the pre-editing symbol, if a character string that matches the specified pre-editing symbol (specifying symbol K0) exists in the original text, the portion is replaced with the pre-editing symbol. Is translated.

Note that the user-defined symbol K1 = “noun” in the original text of the buffer S1. Is present, the processing of steps N9 to N11 in FIG. 2 is executed in the same manner as described above, and the character string "noun" Is pre-edited so as to be designated as “noun” and part of speech, and is translated.

Next, a second embodiment will be described.

The second embodiment is a case where the main purpose is to translate a character string or a symbol string which is not originally a pre-edited symbol as a pre-edited symbol.

FIG. 9 is a diagram showing an example of the contents of a buffer for storing original text according to the second embodiment of the present invention.

FIGS. 10 (a) and 10 (b) are diagrams showing examples of the contents of the buffer in which the type of the pre-edited symbol included therein and the information on its location are stored in the original sentence of FIG.

FIG. 11 is a diagram showing an example of a screen display of the result of translating the original sentence of FIG.

In the second embodiment, an original sentence as shown in FIG. 9 is translated into Japanese. The original sentence of FIG. 9 is remarkably found in patent-related sentences, and is a style in which a plurality of sentences constituting requirements are connected by commas (,) and colons (:) to form one long sentence as a whole. Generally, in order to translate such a sentence by the translation device, it has been considered desirable to divide the original sentence at a position of a comma or a colon and then to proceed to a translation process.

In the second embodiment, the pre-editing process is performed as described below in order to reduce the labor of manual division by the user in the above-described sentence dividing process.

First, before translating the original sentence of FIG. 9, the user defines characters “,” and “:” as pre-editing symbols meaning “divided translation designation” via the keyboard 4. . As a result, the user-defined symbols K1 and K2 whose meaning MA corresponds to "split translation designation" in the buffer D are set as shown in FIG. 5, and the stipulation symbol K0 () whose meaning MA indicates "split translation designation". “//”) is set to be invalid.

In the processing flow of FIG. 2, first, in step N1, the original text is stored in the buffer S1 of FIG.

In the process of step N2, the buffer P of FIG. 5 is cleared. In the next step N3, the original sentence stored in the buffer S1 of FIG. 9 is divided into words and stored in the word W column of the buffer P. At this point, the buffer P is set as shown in FIG. The buffer P in FIG. 10A stores a plurality of words W constituting the original sentence. At this time, the position number LN of the corresponding word is also stored at the same time.

In the processing of the next step N4, FIG.
The value 1 is initialized to the variable i in order to search the buffer D of.

In the loop processing including the following steps N5 to N13, the variable i is repeatedly executed while incrementing the variable i by one until the variable i reaches the final value of the number ID of the buffer D. At this time, the variable i is i =
In the case of 1 to 20, since the preedit symbol corresponding to the number ID matching the variable i in the buffer D does not exist in the original text of the buffer S1, the contents of the buffer P in FIG.

Further, the variable i is incremented and i =
When it becomes 21, it is determined that the user definition flag F corresponding to the number ID = 21 of the buffer D is set to "1" in the process of step N5, and the process proceeds to step N7. In step N7, the user-defined symbols K1 and K2 are detected from the original text of the buffer S1 with respect to the pre-edit symbol with the number ID = 21. Here, as shown in FIG. 5, the user-defined symbols K1 and K2 with the number ID = 21 are ":" and ",".
The character ":" or "," is searched from the original text of the buffer S1 shown in FIG. As a result, since the characters “:” and “,” exist in the original text of the buffer S1, the process proceeds to step N9 according to the determination in the next step N8.

In the process of step N9, in order to specify the position in the original text where the character ":" or "," is detected in the process of step N7, the position number LN = 17, 41, 61 in the buffer P is added. Corresponding pre-edit symbol value KF
1 stores the current value of the variable i (= 21). After that, the processing shifts to the processing of Step N10.

At step N10, it is determined whether or not the word W corresponding to the position numbers LN = 17, 41 and 61 in the buffer P contains the character string ":" or ",". Here, the position number LN = 1
The word W of No. 7 is a character string ",", the character string of the word W corresponding to the position number LN = 41 is ":", and the character string of the word W corresponding to the position number LN = 61 is ",". Therefore, in the buffer P, since the word W corresponding to the detected position number LN contains the user-defined symbol K1 (“:”) or the user-defined symbol K2 (“,”), the step N
The process moves from the process of step 10 to the process of step N11, and the word W corresponding to the position number LN is replaced with the symbol K1 or the symbol K2.
Rewrite not to include That is, in this case, since the character string of the word W corresponding to the position numbers LN = 17, 41, and 61 is composed of only “:” or “,”, the corresponding word W is deleted. As a result, the buffer P changes from the state of FIG.
The state is set as shown in FIG.

At step N12, since the variable i has reached the final value of the number ID of the buffer D, the processing shifts to the next step N14, according to which the buffer P finally returns to the state shown in FIG. Is set. In the process of step N14, the translation process of the original sentence stored in the buffer S1 of FIG. 9 is performed using the information of the buffer P set as shown in FIG. 10B. That is, a preedit symbol “//” representing “split translation designation” is inserted between the words “1” and “a” in the buffer S1, and “split” is inserted between the word “with” and the word “a”. A preedit symbol "//" representing "translation designation" is inserted, and a preedit symbol "//" representing "split translation designation" is inserted between the word "process" and the word "a". The translation processing is performed.

As described above, according to the second embodiment, the default pre-editing symbol, such as "designation of divided translation", is replaced with ",".
The user manually defines a symbol K0 representing "designation of divided translation" etc. in the original text by manually redefining the symbol into "appropriate descriptive symbols already existing in the original text, such as": "and": ". By setting the user-defined symbols K1 to K3 only once without performing pre-edit processing, it is possible to obtain a result equivalent to that obtained by pre-editing the original text manually by the user. Become. In the second embodiment, the result of translating the original sentence of the buffer S1 of FIG. 9 according to the information of the buffer P of FIG. 10B is displayed on the display unit 3 by comparing the original sentence and its translated sentence as shown in FIG. Is displayed.

As is clear from the first and second embodiments as described above, the translation efficiency and accuracy can be improved by the user pre-editing the original text at the time of translation. In such a machine translation apparatus, a function is provided for a user to freely define a character string or a symbol string used only for pre-editing processing. Even if it exists, the translation process can proceed arbitrarily assuming that those character strings and symbol strings are not pre-editing symbols. In addition, since a description symbol or the like already existing in the original text can be arbitrarily redefined as a pre-editing symbol and the translation process can be performed, there is an advantage that the trouble of inputting the pre-editing symbol in the original text can be omitted.

[0080]

As described above, according to the present invention, at least one translation support information is stored in the storage means.
If an arbitrary symbol string is written, the corresponding predetermined symbol string
Is invalid, and one or more written arbitrary character strings
Valid, one or more arbitrary strings are not written
Because the case can be defined changed to the corresponding predetermined symbol sequence is valid, predetermined which is prestored in the storage means
Even if the Symbol column is already present as a character string in the original, one or more of any Symbol operator in response to the translation support information
No. These strings only definition change write the column is performed will not be mistaken, the predetermined Symbol column, it is possible to highly translation processing accuracy. Further, like description symbols already present in the original text, as urchin any one or more symbol strings by using as the information for supporting the above-mentioned translation process
Since the definition can be easily changed by writing, there is an effect that the trouble of inserting the above-described symbol strings one by one into the original text to support the translation processing is reduced, and the translation efficiency is improved.

[Brief description of the drawings]

FIGS. 1A and 1B are block diagrams of a machine translation apparatus and a translation module applied to an embodiment of the present invention.

FIG. 2 is a processing flowchart for explaining an operation of translating an original sentence including a pre-edited symbol of the machine translation apparatus according to the embodiment of the present invention.

FIGS. 3A and 3B are diagrams illustrating examples of preedit symbols applied to the embodiment of the present invention; FIGS.

FIG. 4 is a diagram showing an example of the contents of a buffer in which only prescribed pre-edit symbols are stored.

FIG. 5 is a diagram showing an example of the contents of a buffer in which pre-edit symbols defined and defined by a user are stored.

FIG. 6 is a diagram showing an example of the contents of a buffer in which an original sentence according to the first embodiment of the present invention is stored.

FIG. 7 is a diagram showing an example of a screen display of a result of translating the original sentence of FIG. 6;

FIGS. 8 (a) and (b) show the original text of FIG.
It is a figure which shows the example of the content of the buffer which stores the information about the kind of the pre-editing symbol contained in it, and its existence position with the case where the user defines the pre-editing symbol, and the case where there is no pre-editing symbol. .

FIG. 9 is a diagram illustrating an example of the contents of a buffer in which an original sentence according to the second embodiment of the present invention is stored.

FIGS. 10A and 10B are diagrams showing examples of the contents of a buffer in which information on the types of preedit symbols included therein and their locations are stored in the original sentence of FIG. 9;

FIG. 11 is a diagram showing an example of a screen display of a result of translating the original sentence of FIG. 9;

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 CPU 2 Main memory 3 CRT 4 Keyboard 5 Translation module 6 Memory S1, D, P buffer In each figure, the same code | symbol shows the same or equivalent part.

Claims (1)

(57) [Claims] 1. A translation device for translating a given original sentence into a predetermined language and outputting a translated sentence for the original sentence, comprising: input means operable externally; Symbol string insertion means for inserting a symbol string for supporting translation processing, at least one or more pieces of translation support information, and the translation support information
And a predetermined symbol string set in advance for each
And stores one or more arbitrarily written arbitrary strings
And inputting the original sentence into which the symbol string is inserted by the symbol string inserting means, and storing the symbol string in the original sentence in the storage means .
Translation control means for performing the translation process while interpreting based on the translation support information corresponding to the corresponding symbol string that is made effective , wherein in the storage means, each of the translation support information To
When one or more of the arbitrary symbol strings are written
The corresponding predetermined symbol string has been invalidated and written
The one or more arbitrary symbol strings are set to be valid.
And if one or more of the arbitrary strings is not written
Wherein the corresponding predetermined symbol string is defined to be valid .