JP2007086867A - Translation device, translation method and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a technology for acquiring as highly precise translation result as possible in a short time. <P>SOLUTION: This translation device 1 translates a text by using a plurality of different translation systems in the ascending order of the translation accuracy, and each time translation by each of those translation systems is completed, the translation device 1 stores a translation sentence completed by the translation system. Then, when the output of the translation sentence is instructed, the latest translation sentence is extracted from among the stored translation sentences, and data showing the translation sentence are output. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a technique for translating sentences into sentences in different languages.

A machine translation technique for translating a sentence (original text) described in a certain natural language into another natural language sentence (translated sentence) is known. There are various methods for machine translation. For example, as a method with the highest translation accuracy, a so-called analysis-based translation method combining known morphological analysis, syntax analysis, and semantic analysis is known. Further, as the simplest translation method, there is a method of dividing an original sentence into morphemes and sequentially outputting translated words corresponding to the morphemes.
Each translation method has advantages and disadvantages. For example, in general, in a high-accuracy translation system, it takes a long time to obtain a translation result after inputting an original sentence. On the other hand, in the case of the method that only outputs the translation of each morpheme, the time required for translation is relatively short, but the translation result is not structured as a sentence, so there is a problem that it is difficult to grasp the meaning as it is. .

In view of this, a technique has been proposed that aims to achieve both processing time required for translation and translation accuracy at a certain level. For example, the translation device disclosed in Patent Document 1 stores a table in which a specific keyword is registered, and either a fast reading mode for translating only a sentence including the keyword or a full translation mode for translating the entire sentence. Is selectable. When the user selects the fast reading mode, only the sentence including the specific keyword is translated, so that the outline of the document can be quickly grasped.
JP-A-8-6950

However, the above prior art has the following problems. That is, first, it takes time for the user to select a translation method and give an instruction to the translation apparatus. Secondly, when a user translates with a simple method and is dissatisfied with the translation result, it takes time to redo the translation with a more advanced method.
The present invention has been made under the above-described background, and an object of the present invention is to provide a technique capable of obtaining a translation result with the highest possible accuracy in a short time.

  In order to solve the above-described problems, the present invention provides an input unit that inputs data representing original text and a plurality of different translation methods, which are sequentially used in descending order of translation accuracy, and is input by the input unit. A translation means for translating the original sentence represented by the above, a translation means for generating a translation sentence, a storage means for storing the translation sentence according to the translation system that has been translated as soon as translation by each of the plurality of translation methods is completed, and a translation sentence An instruction means for instructing the output of a translation sentence, and a translation sentence stored last in the translation sentence stored in the storage means when the output of the translation sentence is instructed by the instruction means, and the translation sentence And an output means for outputting data representing the translation device.

  In order to solve the above-described problem, the present invention uses an input step for inputting data representing the original text and a plurality of different translation methods in order from the one with the lowest translation accuracy, and is input in the input step. A translation step for translating the original sentence represented by the data, and generating a translation sentence; and a storage step for storing the translation sentence by the translation system that has been translated as soon as the translation by each of the plurality of translation systems is completed; An output step of extracting the last stored translation sentence from the stored translation sentences and outputting data representing the translation sentence when the output of the translation sentence is instructed Provide translation methods.

  Further, in order to solve the above-described problem, the present invention provides a computer apparatus that uses the input means for inputting data representing the original text and a plurality of different translation methods in order from the one with the lowest translation accuracy, and performs the input. The translation means for translating the original sentence represented by the data input by the means to generate a translation sentence, and storing the translation sentence according to the translation system that has been translated as soon as the translation by each of the plurality of translation methods is completed A storage means, an instruction means for instructing output of a translated sentence, and a translation sentence stored last in a translation sentence stored in the storage means at the time when the output of the translation sentence is instructed by the instruction means. A program for extracting and functioning as output means for outputting data representing the translated sentence is provided.

  According to the above translation apparatus, translation method, or program, the original text is translated using a plurality of different translation methods in descending order of translation accuracy, and as soon as translation by each of the plurality of translation methods is completed, A translated sentence according to the completed translation method is stored. Then, when the output of the translated sentence is instructed, the last stored translated sentence is extracted from the stored translated sentences, and data representing the translated sentence is output.

  According to the present invention, a translation result that is as accurate as possible can be obtained in a short time.

Embodiments of the present invention will be described below with reference to the drawings.
<Configuration>
The translation apparatus 1 has a translation function for translating an input original sentence to obtain a target language translation sentence, and a copying function for reading a manuscript and creating a duplicate.
First, the hardware configuration of the translation apparatus 1 will be described. FIG. 1 is a diagram illustrating a hardware configuration of a translation apparatus 1 according to an embodiment of the present invention.
The translation apparatus 1 includes a control unit 4 including a CPU (Central Processing Unit) 44, a ROM (Read Only Memory) 45, a RAM (Random Access Memory) 46, and the like (all not shown), and is stored in the ROM 45. Each part of the translation apparatus 1 is controlled by the CPU 44 executing an OS (Operating System) program.

The storage unit 5 is a nonvolatile memory such as a hard disk device. The storage unit 5 stores a program describing procedures such as reading, translating, and outputting a document.
The instruction input unit 41 includes a keyboard 40 including a numeric keypad and a start button, and a display unit 39 including a liquid crystal panel having a touch panel function. The user can input an instruction to the translation apparatus 1. The instruction input by the user is, for example, an instruction to start document reading and translation processing, an instruction to output a translation result, or the like. Further, the display unit 39 can display information related to the state of the translation apparatus 1 and displays information indicating that the translation process is in progress, information indicating that the translation process has been completed, and the like.
A sheet 10 for forming an image is accommodated in the sheet feed tray 9. When the user inputs an image formation instruction using the instruction input unit 41, the paper feed roller 33 is driven to rotate, and the sheets 10 are sent out from the paper feed tray 9 one by one. The sheet 10 sent out from the paper feed tray 9 is conveyed along a conveyance path 36 by roller pairs 34, 35, and 37.

  The image input unit 12 is a scanner device that optically reads a document and generates image data. The original placed on the platen glass 2 is irradiated with light from the light source 13, and the reflected light is processed by the optical system 3. The reflected light is received by the light receiving unit 17 via the mirrors 14, 15 and 16. Then, the image processing unit 18 converts the reflected light into an electrical signal, and generates image data composed of each color of yellow, magenta, cyan, and black.

The image forming unit 6 includes image forming engines 7Y, 7M, 7C, and 7K, a transfer belt 8, and the like.
The image forming engines 7Y, 7M, 7C, and 7K form toner images of yellow (Y), magenta (M), cyan (C), and black (K), respectively. Since the configuration of each image forming engine is common, only the image forming engine 7Y will be described here.
The image forming engine 7Y includes a charging device 21Y, an exposure device 19Y, a developing device 22Y, a cleaner 24Y, and the like around a photosensitive drum 20Y as an image carrier on which an electrostatic latent image is formed.

The charging device 21Y charges the surface of the photosensitive drum 20Y that is rotationally driven in the direction of arrow A to a predetermined potential.
The exposure device 19Y is a ROS (Raster Output Scanner) that irradiates a photosensitive drum 20Y charged to a predetermined potential with an exposure beam LB based on image data. The exposure device 19Y emits a laser beam based on image data from a semiconductor laser (not shown), and forms an electrostatic latent image on the surface of the photoreceptor drum 20Y by deflecting and scanning the laser beam. On the surface of the photoreceptor drum 20Y, the potential of the portion irradiated with the laser light is reduced to a predetermined level due to the photoconductivity of the photoreceptor drum 20Y. As described above, when the surface potential of the photosensitive drum 20Y changes, an electrostatic latent image based on the image data is formed on the surface of the photosensitive drum 20Y.

The developing device 22Y is a device that visualizes the electrostatic latent image formed on the surface of the photosensitive drum 20Y. Toner (charged color material) is supplied from the toner tank 23Y, and a toner image is obtained by reversal development of the electrostatic latent image with toner charged to the same polarity as that of the photosensitive drum 20Y.
The transfer belt 8 is stretched around rollers 26, 27, 28, and 29, and is circulated and driven in the direction of arrow B. The photosensitive drum 20Y is in pressure contact with the transfer belt 8 positioned below the photosensitive drum 20Y, and the toner image formed as described above is transferred to the transfer belt 8.
The cleaner 24Y is a device that removes the toner remaining on the photosensitive drum 20Y.

  The above is the configuration of the image forming engine 7Y. In the image forming engines 7M, 7C, and 7K, toner images corresponding to the respective colors are formed and transferred onto the transfer belt 8 in an overlapping manner. Hereinafter, when it is not necessary to distinguish the image forming engines 7Y, 7M, 7C, and 7K, they are simply referred to as the image forming engine 7. Similarly, for other components, the Y, M, C, and K notations are omitted when it is not necessary to distinguish between Y, M, C, and K.

The sheet 10 sent out from the paper feed tray 9 onto the conveyance path 36 enters a nip portion formed by the transfer belt 8 and the transfer roller 30 and is pressed against the transfer belt 8. The toner image is transferred to the surface of the sheet 10 by this pressure contact force and electrostatic attraction force.
The sheet 10 on which the toner image is transferred is guided to the fixing device 11 by the roller pair 31. In the fixing device 11, the sheet 10 is pressed and heated to fix the toner image on the sheet 10. The sheet 10 on which the image is formed in this manner is discharged to the paper discharge tray 32.

Next, the functional configuration of the translation apparatus 1 will be described. FIG. 2 is a diagram illustrating a functional configuration of the translation apparatus 1. These functions are realized by the CPU 44 executing the program stored in the storage unit 5.
The dictionary 201 stores words that are synonymous between different languages in association with each other, and is stored in the storage unit 5. The dictionary 201 stores various languages such as Japanese, English, and Chinese. Further, the dictionary 201 stores information indicating the type of part of speech of the word in association with each word.

The input means 101 inputs data representing the original text. Specifically, the input unit 101 reads an original document in which an original is written using the image input unit 12 and generates original image data representing an image of the original document.
The character recognition unit 102 recognizes a character string included in the data input by the input unit 101. The storage unit 5 stores a known OCR (Optical Character Recognition) program, and character recognition is performed by the CPU 44 executing the OCR program.
The morpheme analysis unit 103 divides the character string recognized by the character recognition unit 102 into morphemes (words or the like). The morpheme analyzing unit 103 extracts a word corresponding to the character string from the dictionary 201.

  The translation unit 104 translates the character string recognized by the character recognition unit 103 to generate a translated sentence. A plurality of translation methods can be used in the generation of the translation, and all are known methods. In the present embodiment, the following three types of translation methods can be used, but this is only an example, and translation methods other than those shown here can also be used.

<A method>
The A method is a method of dividing an original sentence into morphemes and sequentially outputting translated words corresponding to the morphemes.
<B method>
The B method is a fast reading mode disclosed in the above-mentioned Patent Document 1. That is, a table in which a specific keyword is registered is stored in the storage unit 5, and only a sentence including this keyword is translated and output by an analysis-based translation method (C method described later).

<C method>
The C method is a so-called analysis-based translation method that combines morphological analysis, syntax analysis, and semantic analysis. In this method, first, the original text is divided into morphemes using a morphological analysis technique, the syntax of the original text is analyzed using a syntax analysis technique and a semantic analysis technique, and a syntax tree is created and converted to a syntax tree in the target language. To do. Then, the translation corresponding to each morpheme is extracted from the dictionary 201 and applied to the syntax tree to generate a translated sentence.

Here, when each method is compared, it has the following characteristics.
The A method has the lowest translation accuracy because the translation result is not configured as a sentence, but the processing time is the shortest. The C method has the highest translation accuracy, but takes the longest time for translation processing. The B method is an intermediate method between the A method and the C method in both translation accuracy and processing time.

Furthermore, the translation means 104 in the present invention has the following features.
The translation unit 104 performs translation by sequentially using the above three translation methods in the following order. The order of use of the translation methods is arbitrary, but in this embodiment, the order of use is determined by one of the following two types of order determination methods.
<Order determination method 1>
A plurality of different translation methods are used sequentially from the one with the lowest translation accuracy. According to the order determination method 1, the above three methods are used in the order of the A method, the B method, and the C method.
<Order determination method 2>
A plurality of different translation methods are used sequentially from the one with the shortest time required for translation processing. According to the order determination method 2, the usage order of the above three methods is the order of the A method, the B method, and the C method.

As described above, in the case of the above-described three methods, in general, the order of use of the three methods is the same in both the order determination method 1 and the order determination method 2. However, there may be exceptions in the order of use. For example, if the source language and the target language have a very similar grammatical structure, a simple word-by-word translation that only sequentially outputs translations corresponding to the morphemes of the source text may have sufficient accuracy. . In that case, the translation accuracy of the entire translated sentence may be higher in the A method than in the B method. Then, the order of use by the order determination method 1 is the order of the B method, the A method, and the C method.
In addition, when the source language and the target language have a very similar grammatical structure, the time required for the analysis-based (B method, C method) translation processing is relatively short compared to other language combinations. Can be. In that case, the processing time may be shorter in the B method than in the A method. Then, the order of use by the order determination method 2 is the order of the B method, the A method, and the C method.

  As described above, the order in which the translation methods are used may differ depending on the combination of the source language and the target language. In the translation apparatus 1 of the present invention, the order of use of a plurality of translation methods is determined in advance according to the combination of the source language and the translation target language, and the combination of the languages corresponds to the order of use of the translation methods. Attached and stored in the storage unit 5. The order determination unit 108 reads out the use order corresponding to the combination of the source language and the translation target language from the storage unit 5 and supplies information representing the use order to the translation unit 104. The translation unit 104 performs translation according to the order of use represented by the information supplied from the order determination unit 108.

The storage unit 105 causes the storage unit 5 to store data representing a translated sentence that has been translated as soon as translation by each of the plurality of translation methods is completed. When storing the translated text, the translated text data is stored in association with data representing the original text. Here, the data representing the original text may be the original text itself or identification data that can identify the original text.
The instruction means 106 instructs the output of the translated sentence. Specifically, when the user instructs the output of the translation using the keyboard 40 provided in the instruction input unit 41 or the touch panel function of the display unit 39, information representing the instruction is supplied to the CPU 44.

The output means 107 extracts the translation sentence stored lastly from the translation sentences stored in the storage unit 5 at the time when the instruction means 106 instructs the output of the translation sentence, and outputs the data representing the translation sentence. Output. Here, for example, it is assumed that the usage order determined by the order determination unit 108 is the order of the A method, the B method, and the C method. In this case, if only the translation by the A method is completed when the instruction means 106 instructs the output of the translated text, and the translated text is stored in the storage unit 5, the output means 107 outputs this translated text. To do. On the other hand, if the translation of the A method and the b method is completed when the output of the translated sentence is instructed and those translated sentences are stored in the storage unit 5, the output means 107 outputs the translated sentence of the B method. To do. Similarly, if the translation by the A method, the b method, and the C method is completed when the output of the translated sentence is instructed, and these translation sentences are stored in the storage unit 5, the output means 107 uses the C method. Output translated text.
The output data is supplied to the image forming unit 6, and the image forming unit 6 forms an image of the translated sentence on the sheet surface and discharges the sheet to the paper discharge tray 32.

<Operation>
Next, the operation of the translation apparatus 1 will be described. Here, it is assumed that the translation apparatus 1 is powered on and the CPU 44 is executing a program. FIG. 3 is a diagram illustrating an operation flow of the translation apparatus 1. Note that the operation shown below is realized by the CPU 44 executing a program, and therefore the main subject of the operation is the CPU 44.

  First, in step A01, the CPU 44 performs input of original text, character recognition processing, and morphological analysis. Specifically, first, the CPU 44 uses the input unit 101 to read a document on which the original text is written. The user places the document so that the reading surface of the document is in contact with the platen glass 2, and presses the start button of the instruction input unit 41. Then, the image input unit 12 of the translation apparatus 1 starts reading a document. The image input unit 12 generates image data of the read original and stores it in the storage unit 5. When the image data is stored, the CPU 44 recognizes a character string included in the image data using the character recognition means 102. Then, the CPU 44 divides the character string into morphemes using the morphological analysis unit 103. When the character string is divided into morphemes, the CPU 44 uses the order determination unit 108 to determine the order of use of the translation method in accordance with the combination of the source language and the target language. In this example, it is assumed that the order of use of the translation methods is the order of the A method, the B method, and the C method.

  Next, in step A02, the CPU 44 translates the original text using the A method. In step A03, the CPU 44 determines whether or not a translation output instruction has been input. When the output instruction is input (step A03: YES), the process proceeds to step A04, and it is determined whether or not the translation processing by the A method is completed. If the translation process has been completed (step A04: YES), the process proceeds to step A17. In step A17, the CPU 44 stores data representing the translated sentence in the storage unit 5, proceeds to step A18, and outputs the translated sentence by the output means 107. If the translation process has not been completed (step A04: NO), the process returns to step A02.

On the other hand, if a translation output instruction has not been input (step A03: NO), the process proceeds to step A05, where it is determined whether or not the translation processing by the A method has been completed. If the translation process has been completed (step A05: YES), the process proceeds to step A06. In step A06, the CPU 44 causes the storage unit 5 to store data representing the translated sentence. If the translation process has not been completed (step A05: NO), the process returns to step A02.
The CPU 44 constantly monitors whether or not a translation output instruction has been input by performing the determination in step A03 at regular time intervals (for example, at intervals of 1 second).

If the processing of step A06 is completed, the CPU 44 proceeds to step A07 and starts translation processing by the B method. Since the operation from step A07 to step A11 is the same as the operation from step A02 to step A06 except that the translation method is the B method, the description thereof will be omitted. The operation from step A12 to step A16 is the same as the operation from step A02 to step A06 except that the translation method is the C method.
The above is the operation of the translation apparatus 1.
As described above, according to the translation apparatus 1 of the present invention, it is possible to obtain a translation result that is as accurate as possible in a short time.

<Modification>
The present invention is not limited to the form described above, and can be implemented in various forms. For example, the embodiment described above can be modified as follows.
The image input unit 12 may be a separate scanner device connected to the translation device 1 via a network. The image input unit 12 may be a digital still camera that captures an image of a subject and generates image data representing the image of the subject. In this case, the image data is recorded on a recording medium such as a semiconductor memory incorporated in the digital still camera, the image data is read from the recording medium, and the read image data is input to the translation apparatus 1.
The image forming unit 6 may be a separate printer device connected to the translation device 1 via a network.

The instruction unit 106 may receive an instruction to output a translated sentence from a computer device or the like connected to the translation apparatus 1 via a network.
The output unit 107 may output data representing a translated sentence to an image forming apparatus or the like connected to the translation apparatus 1 via a network.
The data output by the output means 107 may be text data representing a translated sentence or image data representing an image of a translated sentence.
Moreover, it is also preferable to comprise the translation apparatus 1 as shown below. That is, the translation apparatus 1 has a known layout analysis function, and analyzes the layout of the original document. The layout analysis identifies the text area, design area, chart area, etc. included in the original text, and character font, character size, character modification (bold, italic, color, halftone, etc.) and character spacing in the text area. Information about line spacing, etc. is obtained. Then, the translated sentence generated by the translation unit 104 is arranged in the same or approximate layout as the original sentence, data representing the translated sentence is generated in a page description language, bitmap data format, etc., and output by the output unit 107 . If the original document includes a pattern or a diagram, data representing an image in which these patterns and diagrams are arranged in the same or approximate layout as the original document is generated together with the translation. In addition, a translation, a picture, a chart, etc. may be generated and output as separate data files.

  If the translated text output instruction is not input by a predetermined time after the input of the original text, the translated text data stored in the storage unit 5 at that time may be deleted. If the storage capacity of the storage unit 5 is insufficient, the translated text data corresponding to the original text with the oldest input time may be deleted from the storage unit 5. Further, when the storage capacity of the storage unit 5 is insufficient, the translated sentence data with the oldest generation time may be deleted from the storage unit 5. Further, when the storage capacity of the storage unit 5 is insufficient, the translated sentence data having the lowest translation accuracy may be deleted from the storage unit 5. In this way, it is possible to prevent the storage area of the storage unit 5 from being occupied unnecessarily.

At the time when the output instruction of the translated sentence is input, the translation processing by the translation method that has not been completed yet may be stopped. In this way, useless use of computing resources of the control unit 4 can be prevented.
Further, when a translation output instruction is input, the translation process using a translation method that has not been completed may be continued. In this way, when the user determines that the translation accuracy of the output translation is insufficient, a translation with a higher translation accuracy can be obtained quickly.

  Alternatively, the translated sentence may be output as soon as the translation process by a certain translation method is completed. Further, as soon as translation processing by a certain translation method is completed, a message indicating that translation is completed may be displayed on the display unit 39 to notify the user of completion of translation.

2 is a diagram illustrating a hardware configuration of a translation apparatus 1. FIG. 2 is a diagram illustrating a functional configuration of a translation apparatus 1. FIG. It is a figure which shows the flow of operation | movement of the translation apparatus.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 ... Translation apparatus, 4 ... Control part, 5 ... Memory | storage part, 51 ... Dictionary, 41 ... Instruction input part, 9 ... Paper feed tray, 10 ... Sheet, 12 ... Image input part, 6 ... Image formation part, 7Y, 7M , 7C, 7K ... image forming engine, 8 ... transfer belt, 201 ... dictionary, 101 ... input means, 102 ... character recognition means, 103 ... morphological analysis means, 104 ... translation means, 105 ... storage means, 106 ... instruction means, 107: output means, 108: order determining means.

Claims (7)

An input means for inputting data representing the original text;
Translating means for translating the original text represented by the data input by the input means by using a plurality of different translation methods from the one with the lowest translation accuracy, and generating a translated text;
As soon as the translation by each of the plurality of translation schemes is completed, storage means for storing a translation sentence by the translated translation scheme;
An instruction means for instructing output of the translation,
An output means for extracting a translation sentence stored last from the translation sentences stored in the storage means at the time when the output of the translation sentence is instructed by the instruction means, and outputting data representing the translation sentence; A translation apparatus comprising: An input means for inputting data representing the original text;
Translating means for translating the original text represented by the data input by the input means by using a plurality of different translation methods from the shortest time required for translation processing, and generating a translated text;
As soon as the translation by each of the plurality of translation schemes is completed, storage means for storing a translation sentence by the translated translation scheme;
An instruction means for instructing output of the translation,
An output means for extracting a translation sentence stored last from the translation sentences stored in the storage means at the time when the output of the translation sentence is instructed by the instruction means, and outputting data representing the translation sentence; A translation apparatus comprising: Order determining means for determining the order of use of the plurality of translation methods according to the combination of the source language and the target language,
The translation apparatus according to claim 1, wherein the translation unit performs translation according to a use order determined by the order determination unit. Reading means for reading a document in which the original is described and generating image data representing an image of the document;
Character recognition means for recognizing a character string included in the image data generated by the reading means,
The translation apparatus according to claim 1, wherein the input unit inputs a character string recognized by the character recognition unit. A reading unit that captures an image of a subject displaying the original text and generates image data representing the image of the subject;
Character recognition means for recognizing a character string included in the image data generated by the reading means,
The translation apparatus according to claim 1, wherein the input unit inputs a character string recognized by the character recognition unit. An input step for inputting data representing the original text;
A translation step of translating the original text represented by the data input in the input step by using a plurality of different translation methods from the one with the lowest translation accuracy, and generating a translated text,
As soon as translation by each of the plurality of translation systems is completed, a storage step of storing a translation sentence by the translated translation system,
An output step of extracting the last stored translation sentence from the stored translation sentences and outputting data representing the translation sentence when the output of the translation sentence is instructed Translation method. Computer equipment,
An input means for inputting data representing the original text;
Translating means for translating the original text represented by the data input by the input means by using a plurality of different translation methods from the one with the lowest translation accuracy, and generating a translated text;
As soon as the translation by each of the plurality of translation schemes is completed, storage means for storing a translation sentence by the translated translation scheme;
An instruction means for instructing output of the translation,
As an output means for extracting the translation sentence stored last from the translation sentences stored in the storage means at the time when the output of the translation sentence is instructed by the instruction means, and outputting data representing the translation sentence A program to make it work.

Images