JP4643183B2 - Translation apparatus and translation program - Google Patents

Description

  The present invention relates to a translation apparatus and a translation program, and more specifically, a translation that translates from a first language to a second language using a first language example sentence and a second language example sentence corresponding to the first language example sentence. The present invention relates to an apparatus and a translation program.

  A technique that has been widely used as a translation device in the past is that a rule that constitutes a sentence of a source language (first language) is made to correspond to a rule of a translated language (second language), and a translated sentence (second language) is based on the rule. ) Is a so-called rule-driven translation device. When this method is applied to a sentence with a standard syntax, it can be translated without any sense of incongruity, but it tends to be an unnatural expression for a sentence with a somewhat unusual syntax.

  In order to remedy this drawback, there is known a translation support device called an example sentence-driven type or an accumulated example sentence use type, which is used to generate a translation by creating a database of examples of human translation results and actual second language sentences.

As an example of such a translation device, use of a dictionary database in which examples are stored in an example database and words of a language are hierarchized into a tree structure based on similarity in meaning, the similarity between the original sentence (input sentence) and the example 2. Description of the Related Art An example-driven translation device is known in which a numerical value is calculated, an example having the smallest value is searched, and translation is performed according to the search result (see, for example, Patent Document 1). Furthermore, a technique is known in which matching is performed after an input sentence is divided into syntax units (see, for example, Patent Document 2). Alternatively, a translation support device is known in which a translation of an example sentence similar to the original sentence is displayed in an editable manner, and the user translates the original sentence by editing the similar translated sentence displayed (see, for example, Patent Document 3). ). In addition, a translation memory device that performs translation by searching a bilingual database that stores bilingual sentences in the first language and the second language based on the translation information, the bilingual information translated by the machine translation device and the bilingual information translated by the translation memory device are provided. There is known a translation support device that captures and cooperates with each other to improve automation of translation work (see, for example, Patent Document 4).
Japanese Patent No. 2609173 JP-A-10-21243 JP 09-2445040 A JP 2002-278964 A

  However, a method using a dictionary database hierarchized in a tree structure together with an example database needs to include both the example and the dictionary as a database, and it is necessary to create a huge amount of data. For this reason, the burden of the work man-hour and cost which are required for development is large. The method of displaying similar example sentences and editing by the user places a heavy burden on the user. Moreover, the method of dividing the input sentence into syntax units requires syntax analysis means, and the burden of working man-hours and cost for preparing the example database and the syntax analysis dictionary is large. A technique including a translation memory device and a machine translation device also has a large work man-hour and cost burden for preparing both of them.

  The problem to be solved by the present invention is to provide a translation apparatus that does not require a large-scale database and can automatically perform natural and high-quality translation based only on the example database.

  The present invention is a translation device that translates from a first language to a second language using a first language example sentence and a second language example sentence corresponding to the first language example sentence, the first language example sentence, Information related to the composition of phrases in one language example sentence, second language example sentences corresponding to the first language example sentences, information related to the composition of words in the second language example sentences, phrases in the first language example sentences and phrases in the second language example sentences A bilingual example sentence storage unit for storing a plurality of sets of bilingual example sentences including information on correspondence with the text, and a bilingual example sentence search unit for retrieving a first language example sentence similar to the original sentence of the first language to be translated from a plurality of sets of bilingual example sentences And a difference part determination unit for determining a difference part between the original sentence and the first-language example sentence of the searched bilingual example sentence, and a phrase of the second language example sentence corresponding to the difference part in the first language example sentence is extracted. And the second language phrase is the first word of the difference in the original text. A synthesizing unit that creates a synthesized sentence that is replaced with the phrase of the above; a translation generating unit that creates a translated sentence in which a phrase in the original sentence included in the synthesized sentence is replaced with a translated word of the second language corresponding to the phrase of the first language; There is provided a translation apparatus including an original sentence, a first language example sentence similar to the original sentence, a second language example sentence corresponding to the first language example sentence, an output part that outputs at least one of a synthesized sentence and a translated sentence.

  The present invention is also a translation device for translating from a first language to a second language using a first language example sentence and a second language example sentence corresponding to the first language example sentence, wherein the original text of the first language A first language example sentence, information about the configuration of the phrase of the first language example sentence, a second language example sentence corresponding to the first language example sentence, information about the configuration of the phrase of the second language example sentence, A bilingual example sentence storage unit for storing a plurality of sets of bilingual example sentences including information on correspondence between the phrases of the first language example sentences and the phrases of the second language example sentences, and a plurality of sets of bilingual example sentences of first language example sentences similar to the original sentence A bilingual example sentence search unit for searching from, a difference part determining unit for determining a difference between the original sentence and the first language example sentence of the searched bilingual example sentence, and a bilingual example sentence related to the phrase of the difference part in the original sentence Translation is determined based on information obtained from And a translation generation unit that creates a translation by substituting a phrase of the second language example sentence corresponding to the difference in the first language example sentence with the determined translation, and an output unit that outputs the generated translation A translation device is provided.

  The translation apparatus of the present invention includes a parallel example sentence search unit that searches a plurality of pairs of parallel example sentences for a first language example sentence similar to the original sentence of the first language to be translated, a first language example sentence of the parallel sentence example searched for the original sentence, A difference portion determination unit that determines a difference portion of each word and a second language example sentence corresponding to the difference portion in the first language example sentence, and extracts the second language phrase of the difference portion in the original sentence A synthesis unit for creating a synthesized sentence replaced with words in the first language, and a translation generation for creating a translated sentence in which words in the original sentence included in the synthesized sentence are replaced with translated words in the second language corresponding to the words in the first language Since a composite sentence or a translated sentence is created with a part, a large-scale database such as a parsing dictionary is required in addition to a plurality of pairs of parallel example sentences stored in the parallel example sentence storage section when generating a composite sentence, that is, an example database. In addition, the example Can output a good translation support quality information based on the database.

  Furthermore, the translation apparatus of the present invention includes an output unit that outputs at least one of an original sentence, a first language example sentence similar to the original sentence, a second language example sentence corresponding to the first language example sentence, a synthesized sentence, and a translated sentence. Therefore, it is possible to present the original text, reference information on the process of creating the final translation result, and the translation in a manner according to the user's intention, so that even if the translation is not appropriate, the user can And synthesized text can be obtained as translation support information.

  In addition, the translation device of the present invention includes a translation determination unit that determines a translation based on information obtained from a parallel translation example sentence related to the phrase of the difference in the original sentence, and the translation of the first language example sentence with the determined translation word A translation generation unit that creates a translation by replacing words of the second language example sentence corresponding to the difference portion, so that in addition to a plurality of pairs of translation example sentences stored in the translation example sentence storage unit when generating the translation, that is, the example database A large-scale database such as a parsing dictionary is not required, and a natural expression can be translated based on an example database. Therefore, it is possible to provide a translation apparatus that can reduce the man-hours and costs required for database development, reduce the storage capacity for storing the database, and perform high-quality translation at low cost.

  A translation device according to the present invention is a translation device that translates from a first language to a second language using a first language example sentence and a second language example sentence corresponding to the first language example sentence, wherein the first language example sentence Information about the composition of the phrases of the first language example sentences, second language example sentences corresponding to the first language example sentences, information about the composition of the phrases of the second language example sentences, phrases and second languages of the first language example sentences A bilingual example sentence storage unit for storing a plurality of sets of bilingual example sentences including information on correspondence with phrases of example sentences, and a bilingual search for a first language example sentence similar to the original sentence of the first language to be translated from a plurality of sets of bilingual example sentences An example sentence search unit; a difference part determination part that determines a difference between the original sentence and the first-language example sentence of the searched parallel example sentence; and a second-language example sentence corresponding to the difference part in the first-language example sentence Extract the phrase and replace the phrase in the second language with the difference A synthesis unit that creates a synthetic sentence that is replaced with a phrase in the first language, and a translation that creates a translation in which a phrase in the original sentence included in the synthetic sentence is replaced with a translation in the second language corresponding to the phrase in the first language A generating unit; and an output unit that outputs at least one of an original sentence, a first language example sentence similar to the original sentence, a second language example sentence corresponding to the first language example sentence, a synthesized sentence, and a translated sentence, To do.

Here, the language is interpreted in a broad sense, and as long as it is a rule for transmitting information, it is not limited to foreign languages but may be dialects or the same language. Alternatively, the language is not limited to the language in which humans talk, and may be a language for computer processing.
The form of the example sentence is not limited to a single sentence, and may include a word, a phrase, or a plurality of sentences.
The output from the output unit may be displayed as characters, but is not limited thereto, and may be, for example, an aspect stored in the storage device as information, an aspect such as sound.

In addition, the translation device of the present invention includes an input unit for inputting a source text in a first language, a parallel example sentence storage unit for storing a plurality of sets of parallel translation example sentences, a parallel example sentence search unit, a difference part determination unit, A translated word determining unit that determines a translated word based on information obtained from a parallel translated example sentence related to the phrase of the different part, and a phrase of the second language example sentence corresponding to the different part in the first language example sentence with the determined translated word It is characterized by comprising a translation generation unit that creates a translation by replacement, and an output unit that outputs the generated translation.
Here, the input from the input unit is not limited to characters as long as it is converted into information that can be processed by the translation apparatus of the present invention, and may be, for example, speech.

  The translation device described above may be configured as a single device, but is not limited thereto. For example, the translation device is provided as software installed in a general-purpose processing device such as a personal computer, and functions as a translation device by executing processing of the software. You may do.

The translated word determining unit may be configured to determine the translated word in consideration of the degree of co-occurrence between a phrase at a different location in the original sentence and another phrase in the original sentence. In this way, when there are a plurality of translation candidates, an appropriate translation can be selected from those candidates.
In addition, a machine translation unit that searches and translates a storage unit that is different from the bilingual example sentence storage unit may be configured to select whether or not to output a translation by machine translation.
Furthermore, you may further provide the edit part for a user to edit the output sentence. In this way, the user can freely edit the content while referring to the sentence output from the translation device, and can support the creation of a final translated sentence that meets the user's intention. .
The phrase may be a word and a collocation.

  Further, phrase information relating to the composition of the phrase of the example sentence is created from the first language example sentence, phrase information relating to the composition of the phrase of the example sentence is created from the second language example sentence corresponding to each first language example sentence, and the created phrase The information processing apparatus may further include a word / phrase association unit that creates correspondence information on correspondence between the first language word / phrase and the second language word / phrase from the information and stores the correspondence information in the bilingual example sentence storage unit. In this way, the user can create a bilingual example sentence without inputting additional information just by inputting corresponding example sentences in the first language and the second language, and can add the registered example sentence in addition to the previously registered bilingual example sentence. Translations can be performed using translated example sentences. Therefore, it is easy to create and register bilingual example sentences for frequently used fields by yourself, or to obtain and register additional bilingual example sentences for specialized fields. This makes it easier to translate more appropriately.

  Furthermore, the bilingual example sentence search unit is configured to search for a plurality of sets of bilingual example sentences for the words constituting the original sentence when the corresponding bilingual example sentence does not exist as a result of the search, and the output unit outputs the search result for the word or phrase. May be. In this way, even when a sentence close to the input sentence does not exist in the bilingual example sentence database, it is possible to support the creation of the translated sentence by the user by combining machine translation and partial wording.

The present invention also relates to an input process for inputting an original sentence in a first language, and a plurality of sets of parallel example sentences stored in a parallel example sentence storage unit, the first language example sentence and the configuration of the phrases of the first language example sentence Information, a second language example sentence corresponding to the first language example sentence, information about a configuration of a phrase of the second language example sentence, and information about a correspondence between the phrase of the first language example sentence and the phrase of the second language example sentence A bilingual example sentence search process for searching a first language example sentence similar to the original sentence from the bilingual example sentence; a difference part determining process for determining a difference part between the original sentence and the first language example sentence of the searched bilingual example sentence by phrase unit; A translation determination process for determining a translation word based on a result of searching a plurality of sets of parallel translation example sentences for the phrase of the difference part in the original sentence, and a second language corresponding to the difference part in the first language example sentence with the determined translation word Replace words in example sentences and translate Providing a clean up process for forming a translation program for executing an output process on the computer to output the generated translation.
Hereinafter, the present invention will be described in detail based on embodiments shown in the drawings.

(Example of device configuration)
FIG. 1 is a block diagram showing an example of the configuration of the translation apparatus of the present invention.
The input unit 300 in FIG. 1 is a means for inputting an original sentence to be translated, instructions relating to processing translation, and contents relating to translation editing. For example, the input part 300 includes a keyboard and a mouse button. There may be.

  The storage unit 900 is a unit that stores an example database including sentences in a plurality of languages having a common meaning and correspondence between languages of words constituting the sentence. FIG. 11 is an explanatory diagram showing an example of a data format of each bilingual example sentence stored in the bilingual example sentence database. Details of the bilingual example sentence database will be described later. The storage unit 900 can be configured using a hard disk, but may be a rewritable storage unit such as a flash memory or a magneto-optical disk, or a non-rewritable storage such as a mask ROM or CD. It may be a means. Further, the storage unit may be configured to be detachable from the apparatus main body and exchangeable, or may be fixed to the main body and not exchangeable.

  The CPU 100 executes processing of input data or instructions from the input unit 300 and data read from the storage unit 900. The storage unit 900, the input unit 300, or the CPU 100 is connected to the transfer unit 200 such as a bus. Alternatively, the transfer unit 200 may include a network at least in part.

  The translation memory program module 700 is a storage unit that stores a program for processing executed by the CPU 100. The translation memory program module 700 can be configured using a flash memory, but is not limited thereto, and may be a rewritable storage means such as a hard disk or a magneto-optical disk, or a mask ROM, CD, or the like is not rewritable. It may be a storage means.

  The control program module 500 stores a program related to processing other than the processing of the translation memory to be executed by the translation apparatus. For example, a process for editing a translation candidate character string, and a process related to input and output control. The control program module 500 can be configured using a flash memory, but is not limited thereto, and may be a rewritable storage unit such as a hard disk or a magneto-optical disk, or a mask ROM, a CD, or the like that cannot be rewritten. It may be storage means.

  The main operation memory 800 is a high-speed memory that stores data temporarily used by the CPU. The main memory 800 includes, for example, an original sentence buffer 810 that temporarily holds the contents of the original sentence and a translation memory related buffer 820 that holds data related to the translation memory.

The present invention can be used in combination with a conventional rule-driven translation apparatus. In this case, a rule-driven translation processing program is stored in the machine translation program module 600. The dotted line is drawn to show that the module is not an essential component of the present invention. When the machine translation program module 600 is included in the apparatus, a buffer related to rule-driven translation processing (not shown) is secured in the operation main memory 800.
The output unit 400 is a means for outputting a translation, translation candidate, and editing result. The output unit 400 may be composed of a display device using an LCD or the like, but is not limited to this, and is another display device such as a CRT. Alternatively, an output I / F unit that outputs data to a printer such as an ink jet method or an external device may be included.

(Configuration from the viewpoint of data processing)
FIG. 2 is a block diagram showing an example of the configuration of the translation apparatus according to the present invention from the viewpoint of data processed in the translation process. In the figure, the original text for translation is input by the user via the input section 300 or read from the original text storage section 10 in which the original text data is stored. Instructions regarding these inputs are managed by the user interface 15. A screen related to the input operation is displayed on the output unit 400 by the user interface 15, and the user instructs a desired input method from the contents of the screen display using the input unit 300, and the user interface unit 15 indicates the instructed content. Predetermined processing is executed according to the determination and the user's selection. Then, the operation performed by the user and the result of translation are displayed on the output unit 400, or the translation result is stored in the translation result storage unit 11.

In addition, the user can edit the translation result displayed on the output unit 400. The editing unit 40 performs processing related to editing.
The input original text is input to the bilingual example sentence search processing unit 31. The parallel translation example sentence search processing unit 31 extracts an example sentence similar to the input original sentence with reference to the parallel translation example sentence database stored in the parallel translation example sentence storage unit 51. The bilingual example sentence database may include bilingual example sentences registered by the user in addition to the examples prepared in advance in the translation apparatus. The bilingual example sentence search processing unit 31 includes data on correspondence between the first language of example sentences having a certain degree of similarity or more, example sentences in the second language corresponding thereto, and words constituting the sentences, and the original sentence and example sentence. And the data related to the difference between them are sent to the translation composition / generation processing unit 30.

Based on the sent data, the translation composition / generation processing unit 30 is a sentence that shows the difference between the extracted example sentence in the second language and the extracted example sentence in the second language that is different from the original sentence. Only the sentence substituted with the first language is generated. In addition, the translated example sentence database 31 is referred to via the parallel example sentence search processing unit 31 to generate a translated sentence that complements the difference.
The translated sentence generated in this way and the sentence of the translation process are output to the output unit 400 via the user interface.

When the user registers a bilingual example sentence, the bilingual example sentence input by the user is sent to the word / phrase association processing unit 50. The word / phrase association processing unit 50 refers to the bilingual dictionary 53 via the bilingual dictionary lookup processing unit 52 in order to analyze the example sentence and add the word / phrase configuration and correspondence information. In this way, the registered bilingual example sentences having the same format as the bilingual example sentence database with the addition of the information on the word / phrase structure and association are stored in the bilingual example sentence storage unit 51, and are stored in advance in the subsequent translation processing. It is referenced in the same way as the parallel translation example.
When a rule-driven translation device is used in combination, the translation dictionary 21 of the device may be shared with the bilingual dictionary 53. Further, a part of the machine translation processing unit 20 may be shared with the bilingual dictionary lookup processing unit 52.

(Outline processing procedure)
FIG. 3 is a flowchart showing an example of a schematic processing procedure of the translation apparatus of the present invention. As shown in FIG. 3, the user inputs the original sentence to be translated (Step s11) and stores it in the original sentence buffer 810 of FIG. 1, and then inputs an instruction for generating a translated sentence (Step s12). When the rule-driven translation process is used in combination, the process is executed (Step s13). The rule-driven translation device is not an essential component of the present invention, and s13 is drawn with a dotted line to indicate that. When rule-driven translation is not performed, the routine proceeds to s14 after executing s12. Further, since the processing procedure is known, no further explanation will be given in this specification. Next, the routine searches the bilingual example sentence database to perform translation composition / generation processing (Step s14). The details of this process are described in FIG. 5 and will be described later as an explanation of FIG. Then, the routine outputs and displays the result of the above processing (Step s15). Details of this processing will be described later. The user edits the displayed translated text or the data in the middle of translation as necessary, or converts the data into a form that is easy for the user to use (Step s16). Details of this processing will be described later.
Here, the process of s12-s14 does not necessarily need to be performed in the procedure as shown in the figure. For example, s13 and s14 may be performed simultaneously, or s14 may be performed before s13.

  FIG. 4 is a flowchart illustrating an example of a schematic processing procedure when the user registers a bilingual example sentence. As shown in FIG. 4, the user inputs a bilingual example sentence to be registered including the first and second languages (Step s31), and further inputs a registration instruction (Step s32). When the instruction is input, information on the composition and correspondence of the phrase for the bilingual example sentence is added (Step s33). By this processing, the translated example sentences are registered in the database after being arranged in the same format as the already translated example sentences (Step s34).

(Bilingual example sentence search and composition / generation processing-overview)
In the following description, detailed processing of each unit will be described using specific sentences as examples. Here, it is assumed that the “first language” is Japanese and the “second language” is English. Further, “Δ” is a symbol representing a blank character.

FIG. 5 is a flowchart showing details of the bilingual example sentence search and composition / generation processing shown in Step s14 of FIG. The original text input by the user in Step 11 of FIG.
"I can't find my glasses."
Is a Japanese sentence. This is stored in the original text buffer 810 of FIG. FIG. 12 is an explanatory diagram schematically showing how the above-described original text is stored in the original text buffer 810. As shown in FIG. 12, the original sentence is stored as a sentence composed of 13 characters in an original sentence buffer 810 configured so that the character position of each character can be identified. Here, the length of the original sentence buffer is not fixed, and a sentence longer or shorter than the example sentence can be stored. The same applies to other buffers described later.

FIG. 11 is an explanatory diagram showing an example of the data structure of the bilingual example sentence database. FIG. 11 shows the following.
First, FIG. 11A shows that the bilingual example sentence database includes “first language sentence data” and an example of the data. In other words, the Japanese sentence (first language sentence) of the bilingual example sentence is “My suitcase is not found” consisting of 17 characters.
Next, FIG. 11B shows that the bilingual example sentence database includes “word information data of the first language sentence” and an example of the data. As shown in FIG. 11 (b), the word information data of the first language sentence includes words “I”, “no”, “suitcase”, and “ga” in the Japanese sentence of the above-described parallel translation example sentence included in the parallel translation example sentence database. It shows that there are 7 places between “Find”, “Not”, “N”, “.”, That the sentence consists of 8 words, and the number of characters of each word.
Further, FIG. 11C shows that the bilingual example sentence database includes “second language sentence data” and an example of the data. That is, the English sentence (second language sentence) of the parallel translation example sentence is an English sentence including “I can't find my suitcase” in total of 25 characters.
FIG. 11D shows that the bilingual example sentence database includes “word information data of the second language sentence” and an example of the data. As shown in FIG. 11 (d), the second language sentence has 6 word breaks between “I”, “can”, “'t”, “find”, “my”, “suitcase” and “.”. , Consisting of 7 words in total, and the number of characters in each word.
Finally, FIG. 11E shows that the bilingual example sentence database includes “correspondence data” and an example of the data. As shown in FIG. 11 (e), the correspondence data consists of four pairs of word correspondences between the first language sentence and the second language sentence, and the correspondence of each pair is as follows. This is shown using each word number.
"I can't find"-"I can't find my"
(Here, "-" in a Japanese sentence is not an ellipsis, but indicates that a word or phrase is inserted in that part, so to speak, it is a perforated part)
"My"-"my"
"Suitcase"-"suitcase"
"Cannot find"-"I can't find"
The similar sentence search process shown in Step s141 of FIG. 5 is performed according to the following procedure.

(Parallel translation example sentence search and composition / generation process-similar sentence search process)
First, using the "My glasses are not found" key, a parallel example sentence similar in Japanese is searched from the parallel example sentence database 51 of FIG. 1 (Step s141). The bilingual example sentence database 51 stores the following example sentences as example sentences using the phrase of the first language “not found”, for example. That is,
Example 1 of bilingual example sentences
First language example sentence: “My suitcase is not found”
Second language example sentence: "I can't find my suitcase."
Example 2 of bilingual example sentences
First language example sentence: “The product you are looking for is not found.”
Second language example sentence: “I can't find the thing I want.”
Example 3 of bilingual example sentences
First language example sentence: “I called in New York, but I couldn't find the book.”
Second language example: “Although we have telephoned all over New York, we have been unable to find copies.”
Example 4 of parallel translation example sentences
In the first language example sentence, "Can you manage it until it is found?"
There are five pairs of second language example sentences “What will you do for me until you find it?”.

In addition, the bilingual example sentence database 51 includes example sentences using the phrase of the second language “atmosphere”.
Example 5 of bilingual example sentences
Second language example sentence "I like the homey atmosphere of that restaurant."
First language example sentence “I like the restaurant's homey atmosphere”,
Example 6 of bilingual example sentences
Second language example sentence "bad atmosphere engulfed the city"
A set of example sentences in the first language, “The town is in a terrible atmosphere”
The two sets of example sentences are memorized.

FIG. 29 is an explanatory diagram showing the data structure of Examples 1 and 2 of these parallel translation example sentences. The data structure of Example 1 of the bilingual example sentence shown in FIG. 29 is that the word breaks of the “word information data of the first language sentence” are “I”, “NO”, “Suitcase”, “GA”, “Find”, “Not”, “N” ”“. ”, And“ word information data in the second language ”has English word breaks of“ I ”,“ can ”“ 't ”“ find ”“ my ”“ suitcase ”“. ” Indicates 7 words. “Correspondence relationship data” indicates that the word correspondence information is as follows.
Action 1: “I ca n’t find my”-“I can't find my”
Here, as mentioned above, “—” in the Japanese sentence indicates not a ellipsis but a so-called perforated part.
Response 2: "My"-"my"
Action 3: "suitcase"-"suitcase"
Action 4: “Cannot find”-“I can't find”

Also, the data structure of Example 2 of the bilingual example sentence is “searching” “having” “product” “ga” “finding” “not” “n” “in the word information data of the first language sentence”. "Word information data in the second language" has English word breaks of "I""can""'t""find""the""thing""I""want"" . ”Indicates 9 words. “Correspondence relationship data” indicates that the word correspondence information is as follows.
Response 1: "Product you are looking for"-"the thing I want"
Response 2: “I want” – “I want”
Action 3: "Product"-"thing"
Action 4: “Cannot find”-“I can't find”

Further, FIG. 30 is an explanatory diagram showing the data structure of Example 3 and 4 of the parallel translation example sentence. The data structure of Example 3 of the bilingual example sentence is that the word breaks of “word information data of the first language sentence” are “New York”, “Medium”, “Telephone”, “Shi”, “Ga”, “That”, “Book”, “ “Is”, “not found”, “not”, “n”, “was”, “.”, And “word information data in the second language” have word separators of English sentences “Although” “we” “have” “Telephoned” “all” “over” “New” “York” “,” “we” “have” “been” “unable” “to” “find” “copies” “.” . “Correspondence relationship data” indicates that the word correspondence information is as follows.
Response 1: "New York"-"New York"
Action 2: “New York”-“all over New York”
Action 3: "Call"-"telephoned"
Response 4: “We have telephoned”-“we have telephoned”
Action 5: “Ga”-“Although”
Action 6: “The book”-“copies”
Response 7: “We could not find”-“we have been unable to find”
Action 8: “We didn't”-“we have been unable to”

In addition, the data structure of the example sentence 4 of the parallel translation example is that the word breaks of the word information data of the first language sentence are “it”, “ga”, “found”, “until”, “no”, “between”, “how” ”“ Can you get it ”or“. ”, And the“ word information data in the second language ”has the word delimiters in English sentences“ What ”“ will ”“ you ”“ do ”“ for ”“ me ” “Until” “you” “find” “it” “?” Indicates 11 words. “Correspondence relationship data” indicates that the word correspondence information is as follows.
Action 1: “It”-“it”
Action 2: “Until you find it”-“until you find”
Action 3: “Can you please”-“will you”
Response 4: “Can you help me?”-“What will you do for me”

Further, FIG. 31 is an explanatory diagram showing a data structure of examples 5 and 6 of the parallel translation example sentences. The data structure of Example 5 of the bilingual example sentence is “I” “like” “the” “homey” “atmosphere” “of” “that” “restaurant” “. "Word information data in the first language" has word breaks of "that""restaurant""no""homely""atmosphere""ga""like""da""." Indicates that it is 9 words. “Correspondence relationship data” indicates that the word correspondence information is as follows.
Response 1: "I like"-"I like"
Response 2: "homey"-"homely"
Action 3: "atmosphere"-"atmosphere"
Response 4: “of that restaurant”-“that restaurant”
Action 5: "restaurant"-"restaurant"

Moreover, the data structure of Example 6 of the bilingual example sentence is 6 words of “Bad”, “atmosphere”, “engulfed”, “the”, “city”, “.” As word separators of “word information data of the first language sentence”, “Second-word word information data” indicates that the word break is 7 words of “town”, “ha”, “bad”, “atmosphere”, “in”, “wrapped”, and “.”. “Correspondence relationship data” indicates that the word correspondence information is as follows.
Response 1: "Bad"-"Terrible"
Action 2: "atmosphere"-"atmosphere"
Action 3: "the city"-"town"

  FIG. 6 is a flowchart showing an example of a detailed procedure of the similar sentence search process (Step s141). As shown in FIG. 6, the routine reads a pair of parallel example sentences from the parallel example sentence database 51 (Step s1411), and obtains the similarity between the parallel example sentence and the original sentence (Step s1412). The retrieval of similar sentences can be realized using a known technique. Further, since the similarity between the original text and the example sentence is also obtained by a known method, further explanation will not be given here. Then, it is checked whether or not the obtained similarity is equal to or greater than a predetermined value (threshold) (Step s1413), and if it is equal to or greater than the threshold, the read bilingual example sentence is stored in the bilingual example sentence search result buffer 822. Thereafter, the routine proceeds to Step s1415. On the other hand, if the obtained similarity is less than the predetermined value, the process proceeds to Step s1415 without performing the above-described processing. In step s1415, it is checked whether or not there is a bilingual example sentence that has not been read. If there is a bilingual example sentence that has not been read, the routine selects a pair of parallel example sentences to be read next, and then returns to the first step s1411 to select it. The above processing is repeated for the translated example sentence. In this way, all the parallel translation example sentences are checked, and when there are no remaining parallel translation example sentences, the routine ends the similar sentence search process and proceeds to Step s142 in FIG.

  FIG. 14 is an explanatory diagram illustrating an example of the parallel translation example sentence search result buffer 822 that stores the retrieved parallel translation example sentences. In FIG. 14, two search results are stored in the bilingual example sentence search result buffer 822. The first language sentence of the first is “My suitcase is not found.” The second first language sentence is “The product you are looking for is not found.” In addition to these data, the first language word information, the second language sentence, and the second language word information are copied and stored from the bilingual example sentence database, and further, the original sentence and the bilingual example sentence obtained by the similar sentence retrieval process described above. Is stored.

  Here, when the threshold for determining the similarity between the original text and the bilingual example sentence to be searched is lowered, the above-mentioned bilingual example sentence with the phrase “not found” is called “In New York, but the book was not found. Or “translation example sentence example 4“ can you help me until it is found? ”. Here, the similarity threshold is set to “50%”, and “My suitcase is not found” in Example 1 of the parallel translation example with similarity “87%” and similarity “58%”. As a result of the search, two sentences “Example of the product you are looking for.

(Bilingual example sentence search and composition / generation process-first language difference determination process)
Next, the routine determines the difference in the first language (Step s142 in FIG. 5). FIG. 7 is a flowchart showing an example of a detailed procedure of the process for determining the first language difference portion. As shown in FIG. 7, the routine initializes a counter for sequentially processing all similar parallel translations searched from the parallel translation example sentence database 51 (Step s1421). Then, a difference between the first language sentence of the similar parallel translation to be processed and the original sentence is determined (Step s1422). The determination of the difference portion can be realized using a known technique.

  After the difference is determined, the routine proceeds to step s1423, where it is determined whether or not the difference is determined for all the similar parallel translations searched (step s1423). Selects the next parallel translation (Step s1424) and repeats the determination of the difference. When the difference portion is determined for all the similar parallel translations, the routine ends the first language difference portion determination process and proceeds to Step s143 in FIG.

  FIG. 15 is an explanatory diagram showing an example of the data structure and data of the different location data buffer 825 that stores the result of the different location in the first language determined by the above-described processing. The difference portion data buffer 825 may define a buffer structure as a part of the structure of the bilingual example sentence search result buffer 822 when storing data for each searched similar parallel translation. FIG. 15 shows the following.

Regarding the first difference between the input sentence “My glasses are not found” and the Japanese sentence “My suitcase is not found.”
The “difference part number” indicates that there is one difference between the original text and the similar translation.
The difference between the “input sentence character position” and the “input sentence character number” between the original sentence “My glasses are not found” and the similar parallel translation is two characters from the third letter of the original sentence. Then, by referring to the original text buffer 810 in FIG. 12, it can be seen that the difference is “glasses”.
The difference between the "first language sentence word number" and the "first language sentence word number" is one word from the third word in the Japanese sentence "My suitcase is not found." It is shown that. Then, by referring to the “word information buffer of the first language sentence” and the “first language sentence buffer” of “search result 1” in FIG. 14, it is understood that the difference is the “suitcase”.
In the example of FIG. 15, the number of differences between the input sentence “My glasses are not found” and the Japanese sentence “My suitcase is not found” in the parallel translation candidate of the search result 1 is 1 Since there are only two locations, the number of data columns in the difference data buffer 825 is one column. In general, when there are a plurality of different parts, the number of data strings is plural, and the “difference part number” takes an integer value in ascending order from 1 to 1.

(Search for bilingual example sentences and synthesis / generation processing-second language difference determination processing)
Subsequently, the routine determines a difference portion in the second language of the similar parallel translation with respect to the original sentence (Step s143 in FIG. 5).
FIG. 8 is a flowchart showing an example of a detailed procedure of the process for determining the second language difference portion. As shown in FIG. 8, the routine initializes a counter for sequentially processing all similar parallel translations retrieved from the parallel translation example sentence database 51 (Step s1431). Then, the difference portion of the second language sentence corresponding to the difference portion of the first language sentence of the similar parallel translation to be processed is determined (Step s1432). The determination of the difference part is performed by determining the contents of the difference part data buffer 825 indicating the difference part of the first language, the contents of the correspondence buffer indicating the correspondence relation between the first language sentence and the second language sentence, and the second language sentence. Processing can be performed using the contents of the word information buffer and the contents of the second language sentence buffer.

  After the determination of the difference portion, the routine proceeds to step s1433 to check whether the difference portion determination has been performed for all the similar parallel translations that have been searched. A similar parallel translation is selected (Step s1434), and the determination of the difference is repeated. When the difference portion is determined for all the similar parallel translations, the routine ends the second language difference portion determination process and proceeds to Step s144 in FIG.

FIG. 16 shows a data structure and data of a synthesized sentence word information data buffer for synthesizing a difference portion of the second language determined by the above processing with the first language, and a synthesized sentence buffer 823 for storing the synthesized sentence. It is explanatory drawing which shows an example. These data buffers may define a buffer structure as a part of the structure of the parallel translation example sentence search result buffer 822 when storing data for each similar parallel translation searched. FIG. 16 shows the following.
The column of “translation example sentence word number” indicates that the difference in the English sentence “I can't find my suitcase.” Of the translation candidate is one word “suitcase” from the sixth word. The column of “difference part number” indicates that the difference part relates to the first difference part between the original sentence and the Japanese sentence of the translation candidate of the search result 1.
The column “Language flag” is information indicating whether the word with the word number is in the first language or the second language, but the word with a certain word number is in the first language (original language). Since a valid value is stored in the column of “difference part number” of the word number, that is, the first language is synthesized with the second language sentence of the parallel translation, ”And“ difference part number ”may be used together.

(Bilingual example sentence search and composition / generation process-compound sentence creation process)
Next, the routine creates a composite sentence in which the difference in the second language of the similar parallel translation to the original sentence is replaced with the original sentence (first language) (Step s144 in FIG. 5).
FIG. 9 is a flowchart showing an example of a detailed procedure of a process for creating a synthesized sentence in which the second language difference portion is replaced with an original sentence. As shown in FIG. 9, the routine initializes a counter for sequentially processing all similar parallel translations retrieved from the parallel translation example sentence database 51 (Step s1441). Then, a synthesized sentence in which the original word is applied to a different part of the second language sentence of the similar parallel translation to be processed is created (Step s1442). The replacement of the difference portion is processed using the contents of the synthesized word information buffer indicating the difference portion in the second language of the similar parallel translation to the original sentence, the contents of the word information buffer of the second language sentence, and the contents of the second language sentence buffer. Can do.

  After the compound sentence is created, the routine proceeds to step s1443, where it is checked whether or not a difference portion has been determined for all the similar parallel translations that have been searched, and if there is a similar parallel translation that has not been determined, the routine The parallel translation is selected (Step s1444), and the composite sentence is repeatedly created. When it is determined that there are different parts for all the similar translations, the routine ends the process of creating the second language compound sentence, and proceeds to Step s145 in FIG.

FIG. 16 shows an example of the data structure and data of the synthesized sentence buffer 823 that stores the synthesized sentence. The composite statement buffer 823 in FIG. 16 indicates the following.
The column of “translation example sentence character number” is the 17th and 18th differences in the English (second language) sentence “I can't find my suitcase.”, And the 19th character is the translation This indicates the 25th character of the second language sentence of the example sentence.
The column of “difference part number” indicates that the difference part of the 17th and 18th characters relates to the first difference part between the input sentence and the Japanese sentence of the translation candidate of the search result 1.
The column of “Language flag” is information indicating whether the word of the word number is in the first language or the second language, the 17th and 18th characters are the first language, and the other characters are the first language. It shows that it is 2 languages.

(Bilingual example sentence search and composition / generation process-second language sentence generation process)
Finally, the routine generates a translated sentence in which the first language of the synthesized sentence is replaced with an appropriate second language (Step S145 in FIG. 5).
FIG. 10 is a flowchart illustrating an example of a detailed procedure of second language sentence generation processing for replacing the first language of the synthesized sentence with an appropriate second language. As shown in FIG. 10, the routine initializes a counter for sequentially processing all similar parallel translations searched from the parallel translation example sentence database 51 (Step s1451). Then, a second language sentence in which an appropriate second language word is applied to the synthesized portion of the first language of the synthesized sentence to be processed is created (Step s1452). Here, in order to apply an appropriate second language word to the synthesized part of the first sentence of the synthesized sentence, a bilingual example sentence is searched for the first language of the different part, that is, the word “glasses”. The most appropriate word is determined from the search result, and the corresponding second language word is applied to the synthesized sentence. In this example, the word “glasses” is used, but a phrase composed of a plurality of words may be used. Details of this processing will be described later.

  After creating the synthesized sentence, the routine proceeds to step s1453, where it is checked whether the second language sentence has been generated for all the retrieved synthesized sentences. If there is a synthesized sentence that has not been generated, the routine The similar parallel translation is selected (Step s1454), and the creation of the second language sentence is repeated. When the second language sentence is generated for all the similar parallel translations, the routine ends the second language sentence generation process, ends the process of Step s145 of FIG. 5, that is, the process of Step s14 of FIG. 1, and performs the process of Step s15. Proceed to

  FIG. 17 shows the data structure of a different sentence first language word buffer, a different part language conversion result buffer, and a generated sentence buffer 824 that stores the generated second language sentence used to generate the second language sentence by the above-described processing. It is explanatory drawing which shows an example of data. These data buffers may define a buffer structure as a part of the structure of the parallel translation example sentence search result buffer 822 when storing data for each similar parallel translation searched. FIG. 17 shows the following.

First, the “character” column of the difference portion first language word buffer indicates that the first difference portion between the input sentence and the search result 1 is “glasses”. The column of “difference part number” indicates that this difference part is the first difference part. Further, the “synthetic sentence word number” field obtains that the word number “6” of the synthetic sentence is a different part from the different part number field of the synthetic sentence word information buffer, and stores the result. . In addition, as a result of searching the bilingual example sentence database for the word of the first language “glasses” at the different location as described above, when the appropriate expression of the corresponding second language is “glasses”, The result is stored in the difference language conversion result buffer. Then, a second language sentence is obtained from the contents of the synthesized sentence word information buffer of FIG. 16 and the contents of the above-described difference language conversion result buffer, and the result is stored in the generated sentence buffer 824 of FIG. Or you may make it produce | generate a 2nd language sentence not only with the content of a different location language conversion result buffer but with a utilization form. Such a utilization form can be derived using an analysis technique related to language processing.

The contents of the generated sentence buffer 824 indicate the following.
The “character” column indicates that the second language generation sentence is “I can't find my glasses.”.
The “Language Flag” column indicates that the “glasses” of the 17th to 25th characters of the second language synthesized sentence is derived from a difference from the original sentence.
The column of “difference part number” indicates that “glasses” of the 17th to 25th characters of the second language compound sentence is derived from the first difference part shown in the difference part data buffer 825 of FIG. .
According to the column of “translation example sentence character number”, in the second language generation sentence, the first to sixteenth characters “I can't find my” and the twenty-fourth character “.” Are the search results of FIG. It is derived from the first to sixteenth characters and the twenty-fifth character in one second language sentence buffer.
The “language flag” field and the “difference part number” field can be used as information for indicating the different part in different modes when displaying the second language generation sentence.
The above is the detailed processing procedure of the bilingual example sentence search / combination / generation processing shown in s14 of the flowchart of FIG.

(Result display processing)
Next, the routine executes a result display process indicated by s15 in the flowchart of FIG.
FIG. 18 is an explanatory diagram illustrating an example of a result display mode. The parallel translation search result display section of g110 is an area for displaying the original text of the translation apparatus of the present invention, the progress of translation, and the results. In the bilingual search result display part g110, the display “gear my glasses is not found” in g111 is the one in which the contents of the original text buffer 810 in FIG. 12 are displayed. In addition, the content of the first language sentence buffer of the search result 1 shown in FIG. 14 is displayed in g112 “My suitcase is not found.” “I can't find my suitcase.” of g113 displays the search result 1 of the second language sentence buffer of the search result 1 shown in FIG. Also, “I can't find my glasses.” Of g114 is the one in which the contents of the synthesized sentence buffer 823 of the first search result shown in FIG. 16 are displayed. Then, “I can't find my glasses.” In g115 indicates the contents of the generated sentence buffer 824 in FIG.
Also, the following contents are displayed below these. My glasses are not found.
The product you are looking for is not found.
I can't find the thing I want.
I can't find my glasses.
I can't find my glasses.
These displays are the contents related to the bilingual example sentences of the search result 2 shown in FIG. 14, and each display corresponds to the above-described displays g111 to g115 related to the parallel example sentences of the search result 1.

The user can individually select whether or not to display the above five types of sentences for each result for the input sentence. The example of the output screen in FIG. 18 is an example when all the above five sentences are displayed. On the other hand, FIG. 21 is an explanatory view showing an example of an output screen in which only the first to third sentences are displayed and the fourth and fifth sentences are not displayed. The display / non-display of each sentence is determined by the value in each column of the processing result display flag buffer 850. An example is shown in FIG. In FIG. 19, the values in the columns of “original text”, “first language sentence”, and “second language sentence” are displayed, and the values in the fields of “second language synthesized sentence” and “second language generation sentence” are displayed. Since “non-display” is selected, only the first to third items are displayed and a screen as shown in FIG. 21 is displayed.
The initial value of the processing result display flag buffer 850 may be set to a predetermined value when the operation of the apparatus starts, or may be configured so that the user can change the value.

FIG. 20 is a flowchart showing details of the result display process (the process of Step s15 in FIG. 3). 18 shows processing when the user operates the display selection instruction column (g121, g122, g123, g124, g125) of each sentence in the output screen example of FIG. Note that “click” in the following description and flowcharts is an example of the design of the apparatus, and may be another operation method.
First, display / non-display of each sentence of g111 to g115 in each search result is determined based on the value of each column of the processing result display flag buffer at that time (s1502 to s1511), and the screen is displayed according to the determined content Is drawn (s1514).
When the drawing process is completed, the routine waits for a user operation (s1515). When any of the display selection instruction fields (g121, g122, g123, g124, g125) in FIG. 18 is clicked by the user, the contents of the processing result display flag buffer 850 are changed accordingly, and the processing result display flag buffer 850 is changed. Is rechecked, the display / non-display of each sentence is determined, and the output screen is redrawn. This state continues from the state waiting for the user's operation until the user performs an operation to exit this state (s1516) (for example, the window is closed). By such processing, the original text, reference information on the process of creating the final translation result, and the translated text can be presented in a manner according to the user's intention.
The above is the detailed processing procedure of the result display process in Step s15 of FIG.

(Edit display contents)
Next, the routine executes a display content editing process shown in Step s16 of FIG. FIG. 23 is an explanatory diagram illustrating an example of a display mode in the editing process of the display content. This processing is a function that allows the user to edit the display contents as desired from the state in which bilingual example sentence search results are displayed as shown in FIG. 18 or FIG. 21, or the results of machine translation to be used together are displayed. I will provide a.

FIG. 22 is a flowchart showing details of the processing related to editing / use by the user (Step s16 in FIG. 3).
When entering this mode, first, as an example of the translation result, the second language generation sentence “I can't find my glasses.” Of the search result 1 described above is transferred to g102 (output result display / editing unit). (S1601) and enters the user operation waiting state (Step s1602). This initial operation is an embodiment of the translation apparatus of the present invention, and is not limited to this.
In the user operation waiting state, the following operation is performed according to the operation.
When the user clicks in g110 (translation search result display section) shown in FIG. 18, the routine proceeds to Step s1603, and highlights the sentence at the clicked position (Step s1604). If a sentence other than the sentence at the clicked position is already highlighted at that time, the sentence is returned to the normal display. That is, the sentence highlighted in the g110 region is controlled to only one sentence. And it returns to a user's operation waiting state.

If the user clicks on g132 (translation instruction button for bilingual search result) in FIG. 18, the routine proceeds to s1605, and the sentence highlighted in the g110 area is displayed in the area of g102 (output result display / editing unit). In addition, transcription is performed (Step s1606). And it returns to a user's operation waiting state.
Alternatively, when g131 (machine translation result transfer instruction button) in FIG. 18 is clicked, the routine proceeds to s1607, and the character string of the machine translation result displayed in the g103 area (stored in the machine translation sentence buffer 830 in FIG. 13). Content) is additionally transferred to the region of g102 (Step s1608). And it returns to a user's operation waiting state.
Alternatively, when the area in g102 (output result display / editing unit) in FIG. 18 is clicked, the routine proceeds to Step s1609, and the area of g102 operates as a text editor (Step s1610). The user can freely edit the content while referring to the existing sentence. Details of the operation of the text editor are well-known techniques, and thus detailed description thereof is omitted. If the user clicks anywhere outside the area of g102, the text editor mode is exited and the user is returned to the waiting state.
When the g133 (output result confirmation instruction button) in FIG. 18 is clicked, the routine proceeds to s1611, and the character string existing in the g102 (output result display / editing unit) area at that time is confirmed. For example, it is copied to the clipboard or input to another application program that moves in conjunction with the clipboard (Step s1612). Then, the editing operation ends.

FIG. 23 is an explanatory diagram showing a state after clicking g132 (transfer instruction button for machine translation result) shown in FIG. 20 from the state shown in FIG. As a result, the machine translation sentence “My glasses aren't found.” Displayed in the area of g103 (machine translation result display section) is generated in the second language originally in g102 (output result display / edit section). It shows the state added after the sentence "I can't find my glasses."
With this function, the user can also create a final translation by performing editing while using information presented by the machine. You don't need to use this feature if the translation is appropriate and you don't need to edit it.

(Automatic transition to word / phrase search mode)
By performing the processing as described above, translation using the bilingual example sentence database can be performed. However, when a bilingual example sentence similar to the input original sentence is searched from the bilingual example sentence database, a similarity degree equal to or higher than a predetermined threshold is obtained. There may be a case where there is no example sentence. In such a case, the input sentence may be divided into words / phrases, the bilingual example sentence database is searched for each divided word / phrase, and the result may be output. As a result, even when the optimum translation is not found, the reference information can be output to assist the user's translation.

  FIG. 24 is a flowchart showing the overall processing of the translation apparatus including such a function. 24, the flow of processing on the left side of the branch of Step s205, that is, the flow from START → s201 → s202 → s203 → s204 → s205 → s211 → s212 → s213 → s214 → s215 → s216 is described above. In the same manner as described above, a sentence similar to the input original sentence is searched from the parallel translation example sentence database, and a translated sentence is generated based on the result. That is, s201 corresponds to s11 in FIG. 3, s202 corresponds to s12 in FIG. 3, s203 corresponds to s13 in FIG. 3, and s204 corresponds to s141 in FIG. The determination at s205 is the same as the similarity determination at Step s1412 of FIG. 6, and the search processing is sequentially performed for all the parallel translations as in the processing shown in FIG. If there is no pair of translated example sentences whose similarity to the input sentence is equal to or greater than a threshold, the routine proceeds to s221, but if there are one or more pairs, the routine proceeds to s211. 5 corresponds to s142 in FIG. 5, s212 corresponds to s143 in FIG. 5, s213 corresponds to s144 in FIG. 5, s214 corresponds to s145 in FIG. 5, s215 corresponds to s15 in FIG. 3, and s216 corresponds to s16 in FIG.

  In the following, a process when there is no pair of parallel example sentences having a similarity with the input sentence equal to or higher than a threshold in the process shown in FIG. Now, assuming that the original text entered is “We recently introduced a new electronic dictionary product on the market.”, There is no pair of translated example sentences that have a similarity above the threshold. And In that case, the routine proceeds to Step s221 in the determination at Step s205, and the original text is divided into words and phrases. As a specific determination method for division, for example, there is a method using a morphological analysis means (a known technique) and a word dictionary. In this case, a dictionary is required, but the dictionary may be a monolingual dictionary of the first language and does not need to be a bilingual dictionary. For example, if the input sentence is Japanese, it is sufficient that Japanese words are collected and organized, and it is not necessary that the Japanese translation has an English translation. Alternatively, as another method, there is a method of dividing by a space position. In this case, a dictionary is not necessary, and it can be used in the case of a language such as English where there is a space between words. Further, as a method different from the above two methods, there is a method of dividing at a position where the character type is switched. In this case, a dictionary is not necessary. For example, in the case of Japanese, when character types are classified into kanji, hiragana, katakana, alphabetic characters, numbers, symbols, etc., they are separated by character boundaries with different character types. It can be used for languages such as Japanese, where the characters used are roughly divided into several types and each character type has a certain tendency.

In this way, the original sentence is divided into words / phrases, and the bilingual example sentence database is searched using those Japanese (first language) words / phrases that are meaningful by searching (Step s222). ), A bilingual example sentence including the searched word / phrase in the first language is obtained as a search result. Here, for example, a method of dividing a Japanese sentence into words / phrases, a method of searching from those words / phrases and selecting a meaningful one, and a database storing Japanese sentences by Japanese words / phrases The searching method can be realized using a known technique.
Then, the search result is output (Step s223). 25 and 26 are explanatory diagrams illustrating an example of an output mode of search results. FIG. 25 is a window similar to FIG. 18 described above, but shows a state in which there is no parallel translation example sentence whose similarity to the input sentence exceeds a threshold value, and that fact is displayed on the parallel translation search result display unit. ing. In addition, the output result display / editing section will display the translated text “Recently, we introduced the new model of the electronic dictionary in the market” instead of “Second Language Generated Text” in “Search Result 1”. Is automatically transcribed. However, this is only an example of the design of this apparatus, and various other processing policies can be considered. At the same time, a screen as shown in FIG. 26 is displayed using, for example, a multi-window system. In the screen of FIG. 26, words / phrases divided from the original sentence and parallel example sentences searched using the words / phrases are displayed. In the output screen example of the word / phrase search mode of FIG. 26, the input original sentence “We recently introduced a new electronic dictionary product at the market” is automatically displayed in the g201 (input sentence display part) area. Is being transcribed. In the search result display part of g204 and the search word list display part of g207, words / phrases (g205 and g207) divided from the input sentence are displayed. When these words / phrases of g205 or g207 are clicked, bilingual example sentences searched by these words / phrases are displayed below the words / phrases in the region of g204, as indicated by g206.

Further, the input sentence display section of g201 can be edited by the user, and it is also possible to search for parallel translated example sentences by dividing into words / phrases based on the sentence rewritten by the user. In that case, by clicking the search instruction button of g202, a search instruction by word / phrase is explicitly performed again.
In the area of g204, when a bilingual example sentence is displayed, if a part where the bilingual example sentence is displayed is clicked, the bilingual example sentence at that part is highlighted. There is at most one pair of translated example sentences that are highlighted, and when one translated example sentence is highlighted, if a part of another translated example sentence is clicked, the translated example sentence at the newly clicked part is highlighted. The bilingual example sentences that have been highlighted until then return to normal display.

  After the search results are displayed as described above, a process for causing the user to edit these search results or to use part or all of them without editing (Step s224) is executed. That is, when a bilingual example sentence in the region of g204 is highlighted on the output screen of the word / phrase search mode of FIG. The second language sentence of the translated bilingual example sentence is additionally transferred to the output result display / editing section of the screen of FIG. 25, and at the same time, the output screen of the word / phrase search mode of FIG. 26 is closed. As a result, the original output screen is as shown in FIG. In the screen of FIG. 27, the output result display / editing section includes the sentence “I ′” retrieved by the word / phrase in addition to the machine translation sentence “Recently, we introduced the new model of the electronic dictionary in the market.” d like to show you our new line of electronic goods. " In the screen of FIG. 27, the user can edit based on the sentence displayed in the output result display / editing unit.

With the above-described function, the translation apparatus of the present invention supports the creation of a translation (second language) by the user by combining machine translation and partial wording even when a sentence close to the input sentence does not exist in the parallel translation example sentence database. be able to.
Note that the processing described above is merely one embodiment of the translation apparatus of the present invention. For example, the search result list, search word list, and one display are not provided separately from the g204 search word list display unit. Various different embodiments are possible, such as displaying on the part. And those aspects are included in the category of this invention.

(Registration of bilingual example sentences by the user)
The translation apparatus of the present invention allows a user to register a bilingual example sentence to be used for translation, and can perform translation using a registered bilingual example sentence in addition to a pre-registered bilingual example sentence. This allows the user to create and register bilingual example sentences for frequently used fields, or to obtain and register bilingual example sentences for specialized fields that have been packaged. It can be enriched and translated more appropriately.

  Further, according to the present invention, when a new bilingual example sentence is registered as a user, not only a sentence set of a plurality of languages registered by the user is registered as a sentence but also a sentence in each language is assigned to each word / phrase. Are divided into words, and the divided words / phrases are associated with each other, and information regarding the division and association of the words / phrases is added and registered. Therefore, the bilingual example sentence registered by the user has the same information as the pre-registered bilingual example sentence, and is used for translation without any inferiority to the pre-registered bilingual example sentence, and a high-quality translation is provided. The

  FIG. 4 is a flowchart showing the flow of the process of registering the bilingual example sentences by the user. As shown in FIG. 4, first, a new bilingual example sentence is input by the user (Step s31). In FIG. 2, a bilingual example sentence is input from the input unit 300 or the original text file 10 under the control of the user interface 15.

The pair of example sentences to be registered now consists of two languages.
First language sentence (Japanese): The product you are looking for is not found.
Second language sentence (English): I can't find the thing I want.
Suppose that However, the input bilingual example sentences are not limited to the character strings (so-called plain texts) of the sentences in a plurality of languages, and are used in other “translation memory” devices (example sentence search type translation support devices). It may include a function of directly importing the data format of the translated example sentence.

  After inputting the parallel translation example sentence, the user instructs “registration” (Step s32). In FIG. 2, a registration instruction is input from the input unit 300 under the control of the user interface 15.

  Then, in response to the user's registration instruction, the input bilingual example sentence is divided into words / phrases constituting the sentence, and the process of associating the languages with the divided words / phrases is performed (Step s33). In the present embodiment, the bilingual example sentences are divided into words / phrases and the word / phrase association processing between the bilingual example sentences is performed by the word / phrase association processing unit 50, as shown in FIG. This is performed by using the parallel dictionary lookup processing unit 52 for searching for it.

Since the bilingual example sentences to be registered are the above-described sentence pairs, the word / phrase association processing unit 50 first divides these sentences into the following words.
First language sentence (Japanese): "Looking""Iam""Product""Ga""Find""Not""N""."
Second language sentence (English): "I""can""'t""find""the""thing""I""want""."
Then, the correspondence information of words / phrases between languages is determined as follows.
Response 1: "Product you are looking for"-"the thing I want"
Response 2: “I want” – “I want”
Action 3: "Product"-"thing"
Action 4: “Cannot find”-“I can't find”
Action 5: "I'm looking for-I can't find"-"I can't find-I want"
Here, the part indicated by “-” in the sentence is not an ellipsis, but indicates that a word or phrase is inserted into the part, which is a perforated part.

  Based on the input bilingual example sentence, the word / phrase information obtained by the above-described processing, and the information on the correspondence between the words / phrases, among the buffers shown in the search result 2 of FIG. All the information stored in the buffer except "" is obtained.

  After the above-described word / phrase association processing, a bilingual example sentence is registered (Step s34). Referring to FIG. 2, the above-described bilingual example sentence, word / phrase information to be added thereto, and word / phrase correspondence information are registered in the bilingual example sentence database 51. The buffer for storing these pieces of information and the contents thereof are as shown in the search result 2 of FIG. However, “similarity buffer with input sentence” in search result 2 of FIG. 14 is excluded.

  In the description of this embodiment, the bilingual dictionary 53 and the bilingual dictionary lookup processing unit 52 that searches the bilingual dictionary 53 as means for dividing the bilingual sentence / phrase inputted by the user and associating the divided word / phrase between languages. However, the means for associating words / phrases in the bilingual sentence is not limited to this. For example, a method is also conceivable in which information is acquired from a bilingual example sentence already registered in a bilingual example sentence database and word / phrase association of a newly input bilingual example sentence is performed. When machine translation is used together, a part of the word / phrase association processing unit 50, the bilingual dictionary lookup processing unit 52, and the bilingual dictionary 53 may be shared with the machine translation processing.

(Translation generation by selecting appropriate translations)
In the example sentence-driven translation device of the present invention, the translated sentence synthesis / generation processing unit 30 in FIG. 2 replaces words / phrases in the first language, which are different parts of the synthesized sentence, with the second language. This process is indicated by the process of Step s145 in the flowchart of FIG. The replacement to the second language, that is, the translation generation process is an advanced one that selects an appropriate translation to be replaced. This process will be further described.

In the following description, when the original language (first language) is English and the second language is Japanese, the input original sentence is
That restaurant offers a good atmosphere.
This case will be described as an example. This is an example for explanation, even if it is another sentence, even if the first language is Japanese and the second original language is English, or one or both of them are Japanese and A language other than English may be used.

Assume that the following pairs of parallel example sentences are listed as search results for parallel example sentences similar to the original sentence.
First language (English): That restaurant offers a good service.
Second language (Japanese): The restaurant has good service.
In this case, as a result of the determination process of the first language difference part (Step s142 in FIG. 5), the difference part in the first language between the original sentence and the translated example sentence is “atmosphere” in the original sentence, and in the first language sentence of the translated example sentence. Determined to be “service”.

Then, in the subsequent determination process of the second language difference portion (Step s143 in FIG. 5), it is determined that the difference portion in the second language sentence of the bilingual example sentence is “service”. In the state where the above processing has been performed, the synthesis process of the second language compound sentence (Step S144 in FIG. 5) is performed, and the compound sentence “The restaurant is good at atmosphere.”
Get.

In the generation process of the second language generation sentence (Step s145 in FIG. 5), the difference portion “atmosphere” in which the first language (English) is inserted in the synthesized sentence is converted into the second language (Japanese). As described above, the translation composition / generation processing unit 30 in FIG. 2 searches the bilingual example sentence database 51 using the keyword “atmosphere” at the difference. As a result of the search, it is assumed that a plurality of pairs of translated example sentences including the following examples are found as a pair of translated example sentences including “atmosphere”.

Example 5 of the above parallel translation example sentence
English example sentence “I like the homey atmosphere of that restaurant.” And Japanese example sentence “I like the restaurant's homey atmosphere.”

Example 6 of the above parallel translation example sentence
English example "bad atmosphere engulfed the city"
A set of Japanese example sentences, “The town is in a terrible atmosphere”

And by the correspondence data contained in the data of these parallel translation example sentences shown in FIG.
In Example 5 of the parallel translation example, it can be seen that "atmosphere"-"atmosphere" are associated,
In the example 6 of the parallel translation example, it is understood that “atmosphere” − “atmosphere” is associated.
Assume that “atmosphere” and “atmosphere” are found as Japanese corresponding to “atmosphere” in this way. The translation composition / generation processing unit 30 needs to select the most appropriate word from the Japanese expressions of these search results and determine it as one word. In order to realize this, the translation composition / generation processing unit 30 performs the following processing.
As described above, the degree of co-occurrence with the other words (for example, “restaurant”, “offers”, and “good”) in the input sentence is obtained for the first language (English) sentence of the retrieved bilingual example sentence. Here, the words to be co-occurrence may be all the words included in the original text, but the words to be targeted may be narrowed down according to a predetermined rule. Details thereof will be described later.

As a result of obtaining the degree of co-occurrence as described above, it is assumed that the bilingual example sentence in which “atmosphere” − “atmosphere” is associated with the word in the input sentence has a higher degree of co-occurrence. In this case, the translation that replaces the difference “atmosphere” in the synthesized sentence is determined to be “atmosphere”, and the translation of the second language generated as a result is
“The restaurant has a good atmosphere.”
Is obtained.

  With this function, the translation apparatus of the present invention can select an appropriate translated word from among the candidates of the second language to be replaced with the difference between the original sentence and the parallel translation example sentence.

  Here, the rules for narrowing down the words to be co-occurrence will be described below. For example, words to be excluded from the target when examining the degree of co-occurrence can be defined in advance. As in this embodiment, when the first language is English, parts of speech such as articles, pronouns, auxiliary verbs, be verbs and prepositions may be excluded. This is because such words rarely have independent meanings. FIG. 28 is an explanatory diagram of an example of a list of exclusion target words. If a list of words to be excluded as shown in FIG. 28 is created and these are included in the original text, the degree of co-occurrence is checked by excluding those words, and the process for generating the translated text The amount can be reduced.

It is a block diagram which shows an example of a structure of the translation apparatus of this invention. It is the block diagram which showed an example of the structure of the translation apparatus of this invention from the viewpoint of the data processed in the process of translation. It is a flowchart which shows an example of the schematic process sequence of the translation apparatus of this invention. It is a flowchart which shows an example of the rough process sequence in case a user registers a bilingual example sentence. It is a flowchart which shows the detail of the search of a translation example sentence shown to Step s14 of FIG. 3, and a synthesis | combination / production | generation process. In this embodiment of this invention, it is a flowchart which shows an example of the procedure of a similar sentence search process. In this embodiment of this invention, it is a flowchart which shows an example of the detailed procedure of the process which determines a 1st language difference location. In this embodiment of this invention, it is a flowchart which shows an example of the detailed procedure of the process which determines a 2nd language difference location. In this embodiment of this invention, it is a flowchart which shows an example of the detailed procedure of the process which produces the synthetic sentence which replaced the 2nd language difference part with the original sentence. In this embodiment of this invention, it is a flowchart which shows an example of the detailed procedure of the 2nd language sentence production | generation process which replaces the 1st language of a synthetic sentence with the appropriate 2nd language. In the translation apparatus of this invention, it is explanatory drawing which shows an example of the data format of each bilingual example sentence stored in a bilingual example sentence database. It is explanatory drawing which showed typically the structure of the original sentence buffer which stores an original sentence. It is explanatory drawing which showed typically the structure of the machine translation sentence buffer in the translation apparatus of this invention. It is explanatory drawing which shows the structure of the parallel translation example sentence search result buffer which stores the parallel translation example sentence searched in the translation apparatus of this invention. In the translation apparatus of this invention, it is explanatory drawing which shows the structure of the different location data buffer which stores the result of the different location of a 1st language. It is explanatory drawing which shows the structure of the synthetic sentence word information data buffer for synthesize | combining the difference part of a 2nd language with a 1st language, and the synthetic sentence buffer which stores the synthesized sentence. It is explanatory drawing which shows the structure of the generation sentence buffer which stores the difference part 1st language word buffer used in order to produce | generate a 2nd language sentence, a difference part language conversion result buffer, and the produced | generated 2nd language sentence. It is explanatory drawing which shows an example of the aspect of a result display in the translation apparatus of this invention. It is explanatory drawing which shows an example from which the aspect of a result display differs in the translation apparatus of this invention. It is a flowchart which shows the detail of a result display process in the translation apparatus of this invention. It is explanatory drawing which shows another example of the aspect of a result display in the translation apparatus of this invention. It is a flowchart which shows the detail of the process regarding the edit and utilization by a user in the translation apparatus of this invention. It is explanatory drawing which shows an example of the display mode in the case of the edit process of a display content in the translation apparatus of this invention. It is a flowchart which shows the whole process including the automatic transfer function to word and phrase search mode in the translation apparatus of this invention. It is explanatory drawing which shows an example of the output mode of a search result in the translation apparatus of this invention. It is explanatory drawing which shows an example from which the output mode of a search result differs in the translation apparatus of this invention. It is explanatory drawing which shows another example of the output aspect of a search result in the translation apparatus of this invention. It is explanatory drawing which shows an example of the list of exclusion object words in the translation apparatus of this invention. It is explanatory drawing which shows the data structure of the examples 1 and 2 of a bilingual example sentence. It is explanatory drawing which shows the data structure of the examples 3 and 4 of a parallel translation example sentence. It is explanatory drawing which shows the data structure of the examples 5 and 6 of a parallel translation example sentence.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Original sentence memory | storage part 11 Translation result memory | storage part 15 User interface 20 Machine translation process part 21 Translation dictionary 22 Grammar rule 30 Translation sentence synthesis | combination production | generation part 31 Bilingual example sentence search process part 40 Editing part 50 Word / phrase matching process part 51 Example sentence database (including user-registered bilingual example sentences)
52 Bilingual Dictionary Lookup Processing Unit 53 Bilingual Dictionary 100 CPU
200 Bus line or network 300 Input unit 400 Output unit 500 Control program module 510 Character string editing unit 521 Means for inputting machine translation sentence to character string editing means 522 Means for inputting second source language generated sentence to character string editing means 530 Display Control unit 531 Original sentence display selection part 532 Parallel translation example sentence first language sentence display selection part 533 Parallel translation example sentence second language sentence display selection part 534 Second language synthesized sentence display selection part 535 Second language generation sentence display selection part 600 Machine translation program module 610 First-second language dictionary search processing unit 700 Translation memory program module 710 Bilingual example sentence search processing program 711 Similarity determination processing program 712 Difference portion determination processing program 720 Second language difference portion determination processing program 730 Second language sentence synthesis processing Program 740 Second language Sentence generation processing program 741 Grammar analysis processing for generation 750 Bilingual example sentence registration processing program 751 Bilingual phrase association processing 800 Operation main memory 810 Original sentence buffer 820 Translation memory related buffer 821 Work buffer 822 Bilingual example sentence search result buffer 823 Synthetic sentence buffer 824 Generated sentence buffer 825 Difference part data buffer 830 Machine translation sentence buffer 840 Parallel translation example sentence registration-related buffer 841 Work buffer 842 Parallel translation example sentence buffer 850 Processing result display flag buffer 900 Storage unit 910 Phrase associated parallel translation example sentence data 920 First to second Language bilingual dictionary data 930 First → second language translation dictionary data, first → second language translation grammar rule data, other first → second language translation rule data 940 Second language generation rule data g102 Force result display / editing unit g103 Machine translation result unit g110 Bilingual search result display unit g111 Original buffer display unit g112 Search result first language sentence buffer display unit g113 Search result second language sentence buffer display unit g114 Composite sentence buffer display unit g115 Sentence buffer display parts g121 to 125 Display selection instruction box g131 Machine translation result transfer instruction button g132 Bilingual search result transfer instruction button g133 Output result confirmation instruction button

Claims (4)

A translation device for translating from a first language to a second language using a first language example sentence and a second language example sentence corresponding to the first language example sentence,
(1) First language example sentence,
(2) Separation in units of words of the first language example sentence and a place in the sentence of each word,
(3) a second language example sentence corresponding to the first language example sentence;
(4) Separation in units of words of the second language example sentence and locations in the sentence of each word; and (5) location of a phrase composed of at least one word of the first language example sentence and at least one word of the second language example sentence. A bilingual example sentence storage unit that stores a plurality of sets of bilingual example sentences that form a set from association information that associates the location of the phrase consisting of:
A bilingual example sentence search unit that retrieves from the bilingual example sentence storage unit a bilingual example sentence having a predetermined similarity with the original sentence, wherein at least a part of the phrase of the first language original sentence to be translated is included in the first language example sentence;
A difference part determination unit that determines a difference part between the original sentence and the searched bilingual example sentence in the first language example sentence;
A synthesizing unit that obtains a phrase of a second language example sentence corresponding to a different portion in the first language example sentence based on the association information and creates a synthesized sentence in which the phrase is replaced with the original phrase;
A bilingual example sentence that includes the same phrase as the phrase of the different part found by the different part judgment unit is searched from each of the bilingual example sentences stored in the bilingual example sentence storage unit, and based on the correspondence information of the bilingual example sentence A translation generation unit for creating a translation by replacing the original words included in the synthesized sentence with the second language;
A translation apparatus comprising: an output unit that outputs at least the synthesized sentence and the translated sentence.   The translation apparatus according to claim 1, further comprising an editing unit for the user to edit the output sentence.   The translation apparatus according to claim 1, wherein the phrase is a word and a collocation. An input process for inputting the original text in the first language;
A plurality of pairs of example sentences stored in the example sentence storage unit,
(1) First language example sentence,
(2) Separation in units of words of the first language example sentence and a place in the sentence of each word,
(3) a second language example sentence corresponding to the first language example sentence;
(4) Separation in units of words of the second language example sentence and locations in the sentence of each word; and (5) location of a phrase composed of at least one word of the first language example sentence and at least one word of the second language example sentence. A bilingual example sentence search process consisting of a set of correspondence information for associating a portion of a phrase consisting of:
A difference part determination process for determining a difference part between the original sentence and the searched bilingual example sentence in the first language example sentence;
A synthesizing process for acquiring a phrase of a second language example sentence corresponding to a different portion in the first language example sentence based on the association information and creating a synthesized sentence in which the phrase is replaced with the original phrase;
From each parallel example sentence stored in the parallel example sentence storage unit, a bilingual example sentence including the same phrase as the phrase of the difference part found by the difference part determination process is searched, and based on the correspondence information of the parallel example sentence A translation generation process for creating a translation by replacing the original phrase included in the synthesized sentence with the second language;
An example sentence-driven translation program that causes a computer to execute output processing for outputting at least the synthesized sentence and the translated sentence.

Images