JPH05334348A - Translation system - Google Patents

Abstract

PURPOSE:To provide the translation system which can be improved in translation throughput by improving the efficiency of a process for editing an original sentence before its translation to assist the translating process. CONSTITUTION:On this translation system, whether or not the pre-editing which is performed before the translation is learnt in word units or word string units is optionally selected through a keyboard 4. A translation module 5 learns a word or word string containing a pre-editing symbol in the original sentence whose learning is specified when the translation is indicated, and the learnt pre-editing symbol is stored in a file; when translation is performed next, the past learning contents stored in the file are utilized to perform the translation.

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 accuracy by pre-editing an original sentence during translation by a user, and particularly to an improvement thereof. ..

[0002]

2. Description of the Related Art Conventionally, the following machine translation devices are available. This translation apparatus includes a keyboard, a CPU (central processing unit), a translation module for translation processing, a CRT (cathode ray tube), and the like. The original text in the source language input from the keyboard is given to the translation module under the control of the CPU. The translation module has dictionary data, grammatical rules, and tree structure conversion rules. The given source language is first morphologically analyzed using a dictionary and then analyzed into a series of words with parts of speech A syntactic analysis was performed from the part-of-speech of the target language using a dictionary and grammatical rules to obtain a translated sentence in the target language. The translated sentence in the target language obtained by the translation is displayed on the CRT or the like in contrast to the original sentence. If the user does not get the correct translated sentence by looking at this screen and determines that the part of speech of the word in the original sentence is incorrect in the translation module, the correct translated sentence is obtained from the original sentence via the keyboard. As described above, the pre-editing process may be added and the translation process may be performed again. This pre-editing process is performed for the purpose of supporting the translation by the translation module by directly specifying the part of speech of a word in the original sentence.

An example of the pre-editing process will be described below. Was
For example, “The invention uses the said components for the d
When the original text "evice." is entered, the conventional translation device
Interprets the word "said" in the original sentence as a verb, such as "the invention uses the components said for the device."
translate. To fix this, for example, “The invention uses the a said components for th
e device. ”, the source text is preedited, that is, the word“ sa
Pre-edit symbol "a" in "id" ", To describe the part of speech of the word
It is possible to specify the words (“sai”) and translation
It The word "said" is used in patent texts.
Frequently, and not "verbs" ("said")
May also be used as an "adjective" ("above")
High For example, in the text of the patent field,
The word "means" is not a "verb" ("means").
Often used as a "noun" ("means").
In addition, the word "close" in a financial text is "moving".
"Noun" ("close price"), not "word" ("close")
Often used.

As described above, with respect to sentences in a specific field, there is a tendency that a word for which repeated preediting may be performed is considerably specified. Therefore, in the conventional translation device, when performing the pre-editing process such as specifying the part of speech of the specific word in the original sentence as described above in order to obtain a correct translated sentence, it is necessary to repeat the pre-editing. There was a problem that the translation work was inefficient in that there was.

Further, in the pre-editing process, in addition to the above-mentioned part-of-speech designation, the dependency between the words in the original sentence is designated, the group at the time of translation of the words in the original sentence is designated, the word / phrase is deleted, and There are processes such as adding words / phrases. In order to prevent a decrease in translation work efficiency caused by repeatedly performing such pre-editing, a method of collectively performing search / replacement of a character string in the original sentence for which pre-editing processing is desired can be considered. By doing this, it is not necessary to repeat the same pre-editing process, which improves the efficiency of translation work.
In an extreme case, the search / replacement of a batch of character strings may significantly rewrite the original text, and it may not be possible to know what the original text was like. For example, if the process of deleting words / phrases in the original sentence is performed collectively, the original sentence will be rewritten to a large extent. There was a problem that it was impossible to know if there was any, and it was difficult to judge whether or not the translation was correct.

Further, in the case where the same pre-editing process is collectively applied to the original sentence as described above, there is a problem that the effect of the pre-editing process cannot be interactively confirmed.

Therefore, an object of the present invention is to improve the efficiency of the pre-editing process when the original sentence is edited using the pre-editing symbols for supporting the translation process before the translation process. An object of the present invention is to provide a translation device capable of improving the efficiency of processing itself.

[0008]

A translation device according to the present invention is a device for translating a given original sentence based on a predetermined rule including dictionary data, and outputting a translated sentence for the original sentence. , Learning means, editing means, and control means.

Prior to the translation process, the learning designating means is configured to designate whether or not to learn, for each character string to which the preedit symbol included in the original sentence and supporting the translation process is attached. ..

The learning means is configured to learn by associating the preedit symbol designated by the learning designating means with the character string to which it is attached.

The editing means selects, from among the character strings in the original sentence, a character string corresponding to the learned character string learned by the learning means.
The learning character string is configured to be edited with an associated preedit symbol.

The control means inputs the original text edited by the editing means at the time of translation processing, applies the information for translation processing support defined in advance corresponding to the preedit symbol and the above-mentioned rule, and performs the translation processing. Is configured to perform.

Further, the above-mentioned learning means has a plurality of files in advance, and stores the pre-edit symbol designated by the learning designating means in the selected file in association with the learning character string to which it is attached. The pre-editing means further has a storage state, and among the character strings in the original sentence, a character string corresponding to the learned character string stored in the selected file,
Editing by adding a related preedit symbol to the learning character string
It may be configured to further include an editing unit.

The above-mentioned plurality of files are prepared for each field to which the description content of the original sentence belongs, or for each user of the translation device, and the selection of files is performed by the field to which the description content of the original sentence to be translated belongs or the translation device of the translation device. It may be selected according to the user.

Further, the learning content by the learning means or the content stored in the file selected by the file storage means may be configured to be arbitrarily changeable.

[0016]

The present invention is configured as described above, and the learning means specifies the learning of the preedit symbol in the original sentence, and the learning means attaches the preedit symbol to the original sentence. The editing means edits the original text by adding the relevant pre-editing symbol to the learned text string in the original text according to the content learned by the learning means Since it is redone, even if there are a plurality of relevant character strings in the learning character string in the original sentence, it is possible to save the trouble of repeatedly adding the same pre-editing symbol to edit.

Further, the learning means has a file storage means having a plurality of files, and a pre-editing symbol designated by the learning designating means is attached to a file selected from the plurality of files. Since it is stored in association with the character string, the pre-editing symbol designated for learning and the character string corresponding thereto are accumulated and stored for each file, and the file is selected for each field to which the description content of the original text belongs, or the translation. Since it is performed for each user of the apparatus, information for supporting the above-mentioned field-specific or user-specific translation processing is stored in each file.

Further, the content learned by the learning means or the content stored in the file by the file storage means can be easily changed without the intervention of the learning designating means and the learning means, and the editing content in the editing means can be varied. It is possible to bring easily.

[0019]

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

1 (a) and 1 (b) are block diagrams of a translation apparatus according to an embodiment of the present invention.

In FIG. 1A, a translation device is a CPU (central processing unit) 1 for centrally managing and controlling the device, and a main unit for storing data including programs required for translation processing. Memory 2, CRT or LC
A display unit 3 including a D (liquid crystal display), a keyboard 4 for inputting data from the outside, a translation module 5 for translating a given source language into a target language, and a memory accessed by the data stored therein by the translation module 5. 6 and a printer 7 for printing out translation results and the like. Each of these units is communicatively connected to each other via a bus.

The memory 6 includes areas M1 to M4. Area M
1 is used as a work area during translation processing. A dictionary D1 is stored in the area M2, a grammar rule D2 is stored in the area M3, and a tree structure conversion rule D3 is stored in the area M4.

The structure of the translation module 5 is shown in FIG. The translation module 5 is a dictionary lookup morpheme analysis unit 1
1, syntax analysis unit 12, conversion unit 13 and translation sentence generation unit 1
Including 4. The translation module 5 uses the given source language in the dictionary lookup morpheme analysis unit 11 to extract the original sentence in the dictionary D.
A morpheme analysis process is performed with reference to 1 to divide into a series of morphemes. A morpheme is the smallest linguistic unit that constitutes the original sentence, and consists of words, prefixes, and suffixes. Analysis unit 11
For each of the words included in the morpheme string obtained in (3), grammatical information such as a part of speech and translated words are obtained, and the tense / person / number is also analyzed. The syntactic analysis unit 12 inputs the morpheme string obtained by the analysis unit 11 of the previous stage, and determines a structural analysis tree indicating the dependency relation between each word according to the dictionary D1 and the grammatical rule D2 of the memory 6. The converting unit 13 converts the structure analysis tree for the original sentence into the structure analysis tree for the translated sentence using the tree structure conversion rule D3 for the structure analysis tree obtained by the analysis unit 12 in the previous stage. Then, the translated sentence generation unit 14 adds appropriate particles and auxiliary verbs as translated sentences in the target language, and the obtained translated sentences are output from the translation module 5. The translated sentence output from the module 5 is output to the printer 7 or the display unit 3 in pairs with the original sentence.

FIG. 2 is a flow chart of processing in the dictionary lookup morpheme analysis unit 11 of FIG. 1 (b). FIG. 3 is a diagram showing an example of a pre-edit symbol used in an embodiment of the present invention.

FIG. 4 is a diagram showing an example of a translation result display screen according to an embodiment of the present invention.

FIG. 5 is a diagram showing an example of the contents of the buffer A referred to when the processing flow of FIG. 2 is executed.

FIG. 6 is a diagram showing an example of contents of the buffer B referred to when the processing flow of FIG. 2 is executed.

FIG. 7 is a diagram showing an example of the contents of the buffer C referred to when the processing flow of FIG. 2 is executed.

FIG. 8 is a diagram showing an example of the contents of the buffer D referred to when the processing flow of FIG. 2 is executed.

FIG. 9 is a diagram showing an example of the contents of the buffer E referred to when the processing flow of FIG. 2 is executed.

FIG. 10 is a diagram showing an example of the contents of the buffer F referred to when the processing flow of FIG. 2 is executed.

FIG. 11 is a diagram showing an example of the contents of the buffer F of FIG. 10 after the data is updated.

FIGS. 12A to 12C are views showing examples of the contents of the preedit learning file according to the embodiment of the present invention.

The preedit symbol used in this embodiment and the meaning of the symbol will be described with reference to FIG.

Pre-edit symbol g [] Is the symbol g in the original text
This means learning the pre-editing of the character string enclosed by [and symbols]. Details of learning the pre-edit symbol will be described later. Previous edit symbol a Means that the part of speech of the word to which the symbol is attached in the original sentence is designated as "adjective". Previous edit symbol n Means that the part of speech of the word to which the symbol is attached in the original sentence is designated as "noun". Previous edit symbol v
Means that the part of speech of the word to which the symbol is attached in the original sentence is designated as "verb".

Next, referring to FIGS. 3 to 12, the process of learning the pre-editing symbol attached to the original sentence in the translation apparatus of FIG. 1 and the translation using the information of the pre-editing symbol learned in the past. The processing will be described in order.

Processing for Learning Pre-Editing Symbols Attached to Original Text In the following description, since a pre-editing symbol learning function is described, a pre-editing learning file desired by a user to be described later is selected in advance. Suppose. The pre-editing learning file desired by the user is prepared and prepared individually in advance in the patent field, legal field, economic field, current affairs field, etc. according to the purpose of the user. It is assumed that the learning file is selected according to the field of the source text to be translated.

As described above, it is assumed that the memory 6 stores a plurality of learning files which the user individually creates and prepares according to the purpose, buffers A to F described later, and translated sentences.

Further, in the translation device of FIG. 1 (a),
The original text that the user wants to translate is the keyboard 4 to 1
It may be input for each sentence, or it may be stored in the memory 6 in advance and read from the memory 6 by the CPU 1 as needed.

FIG. 4 shows an example of a screen display on the display unit 3 of the translation result obtained by the translation apparatus of FIG. 1. The English sentence (corresponding to the original sentence) shown in the figure is translated. A process for obtaining the illustrated translated text (corresponding to the translated text) will be described.

First, the No. 1 English sentence "The inventio" in FIG.
"uses the said components for the device." is input from the keyboard 4, for example, it is given to the translation module 5 by the control of the CPU 1. The translation module 5 translates the English sentence of No1 and the translated sentence of No5 of FIG. In this translation result, the part of speech of the word "said" should have been an adjective ("said"), but it was interpreted as a verb ("said"), resulting in a mistranslation. In contrast, No. 2 English sentence “The inventio” in FIG.
n uses g [The a said] components for the devi
It is assumed that the original text is pre-edited and input to the translation device in advance, such as “ce.”. When the pre-edited No. 2 English text is translated, the correct translation result No. 6 is obtained. The details will be described below.

Referring to the processing flow of FIG. 2, first, in step S1 (abbreviated as S1 in the figure), the original text (No in FIG. 4) is entered.
2) is set in buffer A (Fig. 5).
reference).

Next, in the processing of step S2,
In the character string of file A, the pre-editing symbol "g There is []
It is detected whether or not. If g [] Is in the string
If g Word / word string after converting the character string enclosed in [and]
Set it in buffer B as B2. In parallel with that,
From the character string of the word string B2, the pre-edit symbol (in this case, the symbol a
) Is used as the pre-conversion word / word string B1
Set in buffer B. As a result, the word string B1 is "th
e said "and the word string B2 becomes" the a became said
Are set in the buffer B (see FIG. 6). In addition, this
In the example, the original sentence stored in buffer A is desired to be learned.
Since there is only one pre-edit symbol, the word string is stored in buffer B.
Only one pair of B1 and B2 will be stored
However, learning is desired in the original text stored in buffer A.
If different preedit symbols are inserted in multiple places, the word
Buffer B contains a plurality of pairs of rows B1 and B2.
Will be.

In the processing of the next step S3, the pre-editing symbol g of the learning instruction is extracted from the character string stored in the buffer A. The character string from which [] is removed is set in the buffer C (see FIG. 7).

Next, the process proceeds to step S4, as shown in FIG.
Character string stored in the buffer C and the character string “th stored in the buffer B as the pre-conversion word / word string B1
e said ”is performed. If there is a character string stored in the buffer C that matches the word string B1, the matched character string in the original sentence is converted to the corresponding word / word in the buffer B. Character string of word string B2 "the a said ”
To buffer D. This is because even if the original sentence stored in the buffer A has a plurality of places where the same pre-editing symbol needs to be added, the pre-editing symbol g indicating the learning instruction is present only at one of the places. This is because if the pre-editing is performed using [], the same pre-editing symbol can be added to other portions. Therefore, as a result of the processing in step S4, the character string as shown in FIG. 8 is stored in the buffer D.

Here, the pre-editing learning file that can be selected for each user will be described with reference to FIG. The original text to be translated is a text peculiar to the patent field as shown in FIG. Therefore, the user selects the learning file 20 in the patent field accordingly. The preedit learning file 20 includes items I1 and I2 as shown.
including. The item I1 stores a plurality of pre-conversion words / word strings B1, and the item I2 stores the word string B2 corresponding to each of the word strings B1 stored in the item I1.

Returning to the processing flow of FIG. Next, in the process of step S5, it is checked whether or not the user has selected the pre-editing learning file. If not selected, the process proceeds to step S7 and subsequent steps, which will be described later, but since the pre-editing learning file 20 is currently selected, the process proceeds to step S6 according to the determination in step S5.

In the processing in step S6, the character string in the original sentence stored in the buffer D is matched with each of the pre-conversion word / word string B1 stored in the item I1 of the preediting learning file 20. .. If there is a match, the corresponding character string in buffer D is the learning file 20.
The corresponding converted word / word string B2 is replaced and the result is set in the buffer E. On the other hand, if no match is found, the contents of buffer D are unchanged.
(See FIG. 9). This is the preedit symbol g of the input source text. [] And the batch replacement work of the character string using the pre-editing learning file 20 are completed.

Thereafter, the process proceeds to step S8, and if it is determined that some value is set in the buffer B, the content of the buffer B is added to the preedit learning file 20 in the process of step S9, The data of the edit learning file 20 is updated. The state of the pre-editing learning file 20 at the time when this operation is completed is updated from the state of FIG. 12 (a) to the state of FIG. 12 (b).

Thereafter, the processing shifts to the dictionary lookup processing of step S10, and the character string set in the buffer E is converted into the dictionary D.
It is divided into words while subtracting 1, and the preedit symbol a
Is separated and set in the buffer F (see FIG. 10).
As shown in FIG. 10, the buffer F includes a word position, a word character string, three parts of speech, and a preedit symbol. The word character string is a character string of each word obtained by dividing the character string set in the buffer E into word units. The word position represents the position information of the corresponding word character string in the character string of the buffer E. The three part-of-speech stores the part-of-speech information obtained by subtracting the corresponding word character string from the dictionary D1 and the part-of-speech information specified by the corresponding preedit symbol. The pre-edit symbol is the pre-edit symbol added to the corresponding word character string. The character string set in the buffer E is divided into word units, and the pre-edit symbol is cut off.
After setting in the buffer F of 0, the dictionary D1 is drawn to obtain information such as the part of speech and the form attribute of each word character string, and the information is stored in the "part of speech" column of the buffer F. Here, for ease of explanation, description of the morphological attributes other than the part of speech is omitted. FIG. 11 shows a state in which the information obtained by drawing the dictionary D1 is set in the buffer F.

Thereafter, the processing shifts to step S11,
For the information stored in the buffer F of FIG. 11, the pre-edit symbol (in this example, the pre-edit symbol "a )) Instruction is performed. That is, referring to FIG. 11, the pre-edit symbol “a From among the candidates for the part of speech of the word character string "said" at the word position "4" Part of speech other than "adjective" specified by "" (in this case, "verb")
Is deleted from the buffer F. Although the state of the buffer F obtained by this deletion is not shown, the part of speech “verb” corresponding to the word position “4” is deleted from the state of the buffer F shown in FIG. As a result, the morpheme analysis using the dictionary D1 and the learning file 20 in the dictionary lookup morpheme analysis unit 11 in FIG. 1 is completed, and the morpheme string obtained by the analysis is given to the syntax analysis unit 12 in the next stage.

Returning to the processing of step S5, if it is determined that the user does not select the pre-editing learning file, the contents of the buffer D shown in FIG. It is divided and stored in the buffer F as shown in FIG. Then, the above-mentioned step S1
The processing from 0 onward is executed in the same manner.

As described above, the pre-editing process of the original sentence using the dictionary D1 and the learning file 20 is performed according to the process flow of FIG. 2, and the morpheme analysis is performed on the pre-edited original sentence. Then, the obtained morpheme string is used by the syntax analysis unit 1
Given to 2. Since the information of the part-of-speech “verb” at the word position “4” in the buffer F of FIG. 11 has been deleted, the syntactic analysis unit 12 interprets the corresponding word character string “said” as an “adjective” and parses the syntactic analysis tree. To create. The obtained syntactic analysis tree passes through the conversion unit 13 and the generation unit 14, and the translated text No. 6 of FIG. 4 is displayed together with the English text No. 2 on the display unit 3 and the like.
This is the translation processing of this original text and the preedit symbol g Learning using [] (update of the learning file 20) is completed.

Process of translating using information of pre-edited symbols learned in the past After the learning process of pre-edited symbols is completed as described above, the English sentence No. 3 in FIG. 4 is input to the translation module 5 as the original sentence. Then, again in the dictionary lookup morpheme analysis unit 11, FIG.
The processing flow of is executed.

It is assumed that the user has selected the pre-editing learning file 20 (see FIG. 12B) at this time. It is also assumed that the buffers A to F are cleared and used each time an original sentence for each sentence is input.

When the No. 3 English sentence in FIG. 4 is input, the analyzing unit 11 performs steps S1 to S in FIG. 2 in the same manner as described above.
Process 5 is executed. At this time, a pre-edit symbol g for instructing learning of the pre-edit symbol is input to the input original text. Since [] is not included, the English sentence of No3 is stored in the buffer D as it is. Further, no value is stored in the buffer B.

Since the pre-edit learning file 20 is selected as a result of the determination in step S5, the process proceeds to the next step S6.

In the process of step S6, the buffer D
The character string stored in is searched / replaced based on the pre-editing learning file 20 shown in FIG. That is,
Unconverted word / word string B1 in the file 20 of FIG. 12 (b)
Is detected in the No. 3 English sentence stored in the buffer D, the detected character string is converted to the corresponding converted word / word string B2 “the a of the file 20. After being replaced with "said", the string in buffer D is set in buffer E. In this case, the state of the buffer E is the same as that of FIG. 9 described above.

Next, in the process of step S8, it is determined whether or not there is a value in the buffer B. As mentioned above, N
Since no value is set in the buffer B for the English sentence of o3, the process proceeds to step S10. Step S10
Process, the contents of the buffer E (see FIG. 9) are stored in the buffer F (FIG. 10), and the buffer F (FIG. 10) is stored.
The dictionary lookup processing is executed for the contents of the above, and the result is set again in the buffer F (see FIG. 11). As a result, the pre-editing process using the dictionary D1 and the past pre-editing symbol learning information for the No3 English sentence ends.

Similarly, the No. 7 translated sentence is obtained for the No. 3 English sentence through the syntax analysis unit 12, the conversion unit 13 and the translated sentence generation unit 14 as shown in FIG. 4, and the result is converted into the original sentence and the translated sentence. The images are displayed on the display unit 3 or the like as a pair.

Further, when the No. 4 English sentence shown in FIG. 4 is next input to the translation device, the dictionary lookup morphological analysis unit 1
1 executes the processing of steps S1 to S4 of FIG. 2 in the same manner as described above, the content of the buffer D is "It
v means that the component is good. ”Also, No
In the English sentence of 4, the symbol g that indicates the learning of preediting Since [] is not included, the value is not stored in the buffer B.

In the process of step S5, it is determined that the user has selected the pre-editing learning file 20 (see FIG. 12B), and accordingly the process proceeds to the next step S6.

In the process of step S6, the character string stored in the buffer D and the pre-conversion word / word string B1 of the pre-editing learning file 20 are matched, but the character string "v" in the buffer D is matched. means "and" means "of the word string B1
Does not match, so the converted word / word string B2 "n
String replacement using "means" is not performed. Therefore, "It v means that the component i
s good. ”will be set.

Next, in the processing of step S8, since it is determined that there is no value in the buffer B, the next step S10
And through the process of S11, the translated sentence of No8 shown in FIG. 4 is obtained. That is, the character string "v means "is a pre-editing symbol" v "as shown in FIG. The word "means" with "" will be translated to specify it as a "verb". In this way, the prioritized processing order of the pre-editing symbols is (1) the pre-editing symbol in the character string, (2) the pre-editing symbol designated to be learned in another part of the same original sentence (that is, the buffer B (Pre-edit symbol stored in), (3) pre-edit symbol learned in advance in the pre-edit learning file 20.

As is clear from the above, in the present embodiment, by selectively learning the contents of the pre-editing processing performed by the user for each pre-editing learning file arbitrarily selected by the user, a specific field can be obtained. The need for frequent pre-editing processing can be minimized. This is done more effectively as the learning progresses. Also, if you do not want to apply the pre-editing information learned in the past for a specific word in the original sentence when the pre-editing learning file is selected, perform the original pre-editing process for the word in the original sentence. By applying it, it is possible to avoid applying the pre-editing learning file. In this way, the efficiency of the pre-editing work by the user, and thus the work efficiency of the entire translation work using the translation device can be improved.

Further, in order to cancel the pre-editing learning information registered in the pre-editing learning file 20, for example, the pre-editing symbol "c" is used. A pre-edit symbol, such as [], which indicates to cancel learning information is provided in advance, and if there is a character string enclosed by this symbol in the original sentence, the pre-edit symbol is removed from the character string. Character string (word before conversion / word string B1)
This can be realized by deleting the data related to the pre-editing learning file 20.

Further, the result obtained by applying the contents of the buffer B or the contents of the preediting learning file 20 to be selected to the original sentence (contents of the buffer E) is displayed on the display unit 3 or the like, and is displayed on the device. It is also possible to display, as desired, how the input original text was preedited inside the device.

Further, in order to know what kind of pre-editing process has been learned in the past, the latest pre-editing learning file 20 (see FIG. 12C) is displayed on the display unit 3 as desired, or It is also possible to print on the printer 7 for confirmation.

Further, in this embodiment, FIG.
Although there is no mention of improving the efficiency of the search of the pre-editing learning file 20 shown in (c), the inside of the pre-editing learning file 20 is converted to the pre-conversion word / word string B1 in order to increase the efficiency of the whole translation work. Store while sorting by using as a key, create an index for the sorted key,
At the time of searching the pre-editing learning file 20, it is possible to use this index for high-speed searching.

13 (a) to 13 (c) and FIG. 14 are diagrams showing other examples of pre-edit symbols used in one embodiment of the present invention. FIG. 13A shows a pre-editing symbol for designating the part of speech of a word. Reference numeral 13 (b) is a pre-editing symbol for designating any of the first sentence pattern to the fifth sentence pattern in an English sentence or the like. 13 (c) is the translation module 5
It is a pre-editing symbol for specifically designating any one of the user dictionary, the technical term dictionary, and the basic word dictionary for the dictionary D1 referred to in the translation process in. The pre-editing symbol shown in FIG. 14 is a pre-editing symbol used for dividing the original sentence. For example, some requirements are summarized as 1
If the original sentence is one long sentence, efficient translation can be executed by preediting the original sentence by dividing the original sentence by using a symbol as shown in FIG. 14 at the position of comma or colon.

Further, in the translation apparatus according to the present embodiment, it is possible to arbitrarily select whether or not to learn a specific preedit symbol, so that the preedit processing with less side effects than unconditionally learning the preedit symbol is possible. Is obtained. The side effect mentioned here means that the learning content of the preedit symbol is adversely affected when it is applied to another original text. In detail, assuming that there is no selection operation for learning or not, the word “mean
"s" is always the string "n" with a preedit symbol means ”
If the original text to be translated is an original text applicable to a field other than patent-related, it will be misinterpreted frequently. For example, if the original sentence to be translated is a patent-related original sentence, in most cases, the word "means" can be specified as a "means" and a noun and translated, but there is no problem, but in other fields such as economic relations. If it is a source sentence related to, it means that the word "means" should be translated as a verb of "meaning". Therefore, in the latter text of economic relations, one may choose not to learn the word "means". On the other hand, when translating original texts related to patents, the word "means" is frequently used as a noun of "means" as is well known, so it can be said that it is better to choose to learn.

[0072]

As described above, according to the present invention, it is only necessary to specify, through the learning designating means, that the preediting symbol in the original sentence is to be learned. Learned by associating with the character string. Further, the editing means re-edits the original text by adding the relevant pre-editing symbol learned to the character string in the original text corresponding to the learning character string according to the content learned by the learning means. Even if there are a plurality of corresponding character strings in a column, it is possible to save the trouble of adding the pre-edit symbol and editing it. As a result, there is an effect that the work of adding a pre-editing symbol to support the translation process and performing the editing process is labor-saving, and the processing efficiency of the translation process itself is improved.

Further, the learning means has a file storage state having a plurality of files, and each file has information for supporting the translation process for each field to which the content described in the original text belongs or for each user. Since the information is stored, the editing efficiency in the first editing means for editing the original text using the file in which the information is stored improves with the amount of information stored in the selected file.

Further, since the learning contents by the learning means or the contents stored in the file by the file storing means can be arbitrarily changed, the processing for editing the original sentence using these learning contents is diversified. The work relating to pre-editing also has the effect of diversifying its contents.

[Brief description of drawings]

1A and 1B are configuration diagrams of a translation device according to an embodiment of the present invention.

FIG. 2 is a flow chart of processing in a dictionary lookup morpheme analysis unit of FIG. 1 (b).

FIG. 3 is a diagram showing an example of a pre-editing symbol used in an embodiment of the present invention.

FIG. 4 is a diagram showing an example of a translation result display screen according to an embodiment of the present invention.

5 is a diagram showing an example of contents of a buffer A referred to when the processing flow of FIG. 2 is executed.

FIG. 6 is a diagram showing an example of contents of a buffer B referred to when the processing flow of FIG. 2 is executed.

7 is a diagram showing an example of contents of a buffer C that is referred to when the processing flow of FIG. 2 is executed.

FIG. 8 is a diagram showing an example of contents of a buffer D referred to when the processing flow of FIG. 2 is executed.

9 is a diagram showing an example of contents of a buffer E referred to when the processing flow of FIG. 2 is executed.

10 is a diagram showing an example of contents of a buffer F referred to when the processing flow of FIG. 2 is executed.

11 is a diagram showing an example of the contents of the buffer F of FIG. 10 after the data is updated.

12A to 12C are diagrams showing examples of contents of a preediting learning file according to an embodiment of the present invention.

13A to 13C are diagrams showing other examples of pre-edit symbols used in an embodiment of the present invention.

FIG. 14 is a diagram showing another example of the pre-editing symbol used in the embodiment of the present invention.

[Explanation of symbols]

 5 Translation Module D1 Dictionary D2 Grammar Rule D3 Tree Structure Conversion Rule In each figure, the same reference numerals indicate the same or corresponding parts.

Claims (5)

[Claims] 1. A translation device for translating a given original sentence based on a predetermined rule including dictionary data, and outputting a translated sentence for the original sentence, which is included in the original sentence before the translation process, And a learning designating means for designating whether or not to learn the pre-editing symbol for supporting the translation processing for each character string to which it is attached, and the pre-editing symbol designated by the learning designating means,
A learning unit for learning by associating with the character string to which it is attached; a character string corresponding to the learning character string learned by the learning unit among the character strings of the original sentence; Editing means for adding a pre-editing symbol, and inputting the original sentence edited by the editing means at the time of the translation processing, and supporting the translation processing defined in advance corresponding to the pre-editing symbol. And control means for executing the translation process by applying the information and the rules.
Translation device. 2. The learning means has a plurality of files in advance, and associates the preedit symbol designated by the learning designating means with the learning character string to which the preedit symbol designated by the learning designating means is associated with the selected file. The editing means further includes a file storage means for storing, of the character strings in the original sentence, a character string corresponding to the learned character string stored in the selected file, The translation device according to claim 1, further comprising first editing means for editing with the pre-edit symbol. 3. The plurality of files are prepared for each field to which the description content of the original sentence belongs, or for each user of the translation device, and the selection of the file is performed in the field to which the description content of the original sentence to be translated belongs or The translation device according to claim 2, wherein the translation device is selected according to a user of the translation device. 4. The translation device according to claim 1, wherein the contents learned by the learning means can be arbitrarily changed. 5. The translation device according to claim 2, wherein contents stored in the selected file can be arbitrarily changed.