JPH0991295A - Translation processor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To realize learning which can efficiently obtain the translation result intended by a user. SOLUTION: A result table 44 where transformation information including translations as translation candidates having the candidate order of each word obtained by translation processing, words corresponding to these translations, and their parts of speech are stored, a registration information taking-out part 45 which takes out a translation of translation information as the learning object, words corresponding to this translation, and their parts of speech from the result table in accordance with the instruction of learning execution to stored translation information, a first learning dictionary 48 where learning information to be referred to for the purpose of determining the order of translation candidates is stored, a learning registration part 47 which registers taken-out translation information in the first learning dictionary 48 as learning information in accordance with the order of learning, a learning dictionary reference part 50 which takes out learning information stored in the first learning dictionary so that the learning order can be discriminated, and a first result change part 53 which changes the candidate order of translation information stored in the result table 44 in accordance with taken-out learning information are provided.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a translation processing device having a function of learning a result obtained by a translation process for translating a sentence in a first language into a sentence in a second language.

[0002]

2. Description of the Related Art Generally, a translation processing device for translating a sentence in a first language into a sentence in a second language is provided with a function (learning function) for learning the result obtained by the translation process. . The learning function, for example, when there are multiple translation candidates for a word, learns the translation finally selected as the translation result and reflects it in the subsequent translation processing so that the learned translation is the correct translation. The output is prioritized as a candidate.

[0003] Conventionally, when learning a translation result, an analysis is made of the context of words and the interrelationship between words,
Along with the information of the analysis result, the information of the translation to be preferentially output is stored in the learning dictionary. The translation processing device performs a translation process on a sentence in a first language to analyze a context of words and a catching relationship between the words when obtaining a translated word of a word in a second language, and based on the analysis result. The learning dictionary is searched to find the corresponding translated word, which is given priority and output as a correct translation candidate.

Further, the translation processing device automatically carries out learning without the user's awareness, and refers to the content of the information stored in the learning dictionary by this learning and reflects it in the translation result. At that time, the translation processing apparatus considers that the recently learned content is the translation result requested by the user, and reads the latest learning content from the learning dictionary and reflects it in the translation processing.

[0005]

As described above, in the learning function in the conventional translation processing apparatus, when storing the learning content in the learning dictionary or extracting the learning content from the learning dictionary, the context of words and I had to analyze the relationship. In addition, the learning dictionary had to store information on these analysis results.

That is, it takes a lot of processing time to reflect the learning contents in the translation processing, which causes a decrease in the processing speed of the translation processing. In addition, since a large amount of information is registered in the learning dictionary, a large size (capacity) of the learning dictionary is required, and a large capacity requires a long time for searching, resulting in a decrease in processing speed.

[0007] Further, the translation processing device allows learning to function without the user being aware of it, and cannot reflect the user's intention for the learning function. In other words, the level (condition, learning strength) that reflects the learning effect on the translation result cannot be selected only by performing the one-method learning, and the translation processing is performed based on the latest learning, so that the user does not always request. There are some aspects that are difficult for users to understand, such as being unable to obtain the translated results.

The present invention has been made in consideration of the above circumstances, and an object of the present invention is to provide a translation processing apparatus capable of learning by which a translation result intended by a user can be efficiently obtained. .

[0009]

The first invention is a translation candidate having a candidate rank for each word, which is obtained by a translation process for translating a sentence in a first language into a sentence in a second language. A result table for storing translation information including a word and a part of speech corresponding to a translation, and a translation information to be learned from the result table according to a learning execution instruction for the translation information stored in the result table. A translation, registration information extracting means for extracting a word and a part of speech corresponding to the translation, a first learning dictionary for storing learning information referred to for determining the rank of candidates obtained by the translation process, and the registration Using the translation information extracted by the information extracting means as learning information, learning registration means for registering the learning information in the first learning dictionary in accordance with the learning order, and translation stored in the first learning dictionary. The learning information corresponding to the word to be processed is extracted in the result table in accordance with the learning dictionary reference unit that extracts the learning order so that the learned order can be identified, and the learning information extracted by the learning dictionary reference unit. It is characterized by comprising a first result replacing means for replacing the candidate ranks of the stored translation information.

As a result, it is not necessary to analyze the context of words or the relationship between words, and it is not necessary to store the information of the analysis result in the learning dictionary. Therefore, learning or translation processing using the learning dictionary is performed. Can be executed efficiently.

A second invention is obtained by a translation process for translating a sentence in the first language into a sentence in the second language,
A translation candidate having a candidate rank for each word, a result table for storing translation information including words and parts of speech corresponding to the translation, and a learning execution instruction for the translation information stored in the result table. Accordingly, the translation of the translation information to be learned, the word and the part of speech corresponding to the translation, the translation of the first candidate having the first candidate rank and the part of speech corresponding to the translation are extracted from the result table. Registration information retrieval means,
A second learning dictionary for storing learning information referred to for determining the rank of candidates obtained by the translation process;
The translation information retrieved by the registration information retrieval means is used as learning information, and the learning registration means for registering in the second learning dictionary in accordance with the order of learning and the translation processing object stored in the second learning dictionary. The learning information corresponding to the word is stored in the result table according to the learning dictionary reference means for extracting the learning information so that the learned order can be identified, and the learning order of the learning information extracted by the learning dictionary reference means. First to replace the candidate rankings for translated information
It is characterized by including a result replacement means.

As a result, it is not necessary to analyze the context of words or the relationship between words, and it is not necessary to store the analysis result information in the learning dictionary. Therefore, learning or translation processing using the learning dictionary is performed. Translation can be executed efficiently and the translation result intended by the user (first candidate)
Can be more easily obtained.

A third invention is obtained by a translation process for translating a sentence in the first language into a sentence in the second language.
A translation candidate having a candidate rank for each word, a result table for storing translation information including words and parts of speech corresponding to the translation, and a learning execution instruction for the translation information stored in the result table. Accordingly, the translation of the translation information to be learned, the word and the part of speech corresponding to the translation, the translation of the first candidate having the first candidate rank and the part of speech corresponding to the translation are extracted from the result table. Registration information retrieval means,
A second learning dictionary for storing learning information referred to for determining the rank of candidates obtained by the translation process;
With the translation information extracted by the registration information extracting means as learning information, learning registration means for registering in the second learning dictionary according to the order of learning, and a translation as an appropriate translation candidate as a translation result in the result table are provided. Appropriate translation determining means for determining whether or not it is stored, and the learning information corresponding to the word to be translated which is stored in the second learning dictionary are extracted so that the learning order can be identified. According to the learning dictionary reference means for selectively extracting a part of the learning information according to the result of the determination by the appropriate translation determination means, and the learning order of the learning information extracted by the learning dictionary reference means. And a first result replacement unit that replaces the candidate ranks of the translation information stored in the result table.

As a result, the learning effect is reflected more strongly in the translation result, and when there is no suitable translation, the learning effect is loosened and the translated word is obtained, so that the translated word cannot be obtained.
Alternatively, it is possible to avoid a situation in which an inappropriate translation is obtained.

Further, a fourth invention is obtained by a translation process for translating a sentence in the first language into a sentence in the second language,
A translation candidate having a candidate rank for each word, a result table for storing translation information including words and parts of speech corresponding to the translation, and a learning execution instruction for the translation information stored in the result table. Accordingly, the translation of the translation information to be learned from the result table, the registration information extracting means for extracting the word and the part of speech corresponding to the translation, and the reference for determining the order of the candidates obtained by the translation processing are referred to. A first learning dictionary for storing learning information, learning registration means for registering the translation information extracted by the registration information extracting means as learning information in the first learning dictionary according to a learning order, and the result An appropriate translation determining means for determining whether or not a translation which is an appropriate translation candidate is stored in the table as a translation result, and a determination result by the appropriate translation determining means, The learning dictionary reference means for extracting the learning information corresponding to the word to be subjected to the translation processing stored in the first learning dictionary so that the learning order can be identified, and the learning dictionary reference means for extracting the learning information. First, storing learning information as translation information in the result table
It is characterized by including a result replacement means.

As a result, when an appropriate translation is not obtained even when the learning content is reflected in the translation processing, the translation of the learning information stored in the learning dictionary other than the translation candidates obtained by the translation processing is obtained. Since it can be a translation candidate, it is possible to avoid a situation in which a translated word cannot be obtained or an inappropriate translated word is obtained.

A fifth invention is obtained by a translation process for translating a sentence in the first language into a sentence in the second language,
A translation candidate having a candidate rank for each word, a result table for storing translation information including words and parts of speech corresponding to the translation, and a learning execution instruction for the translation information stored in the result table. Accordingly, the translation of the translation information to be learned from the result table, the registration information extracting means for extracting the word and the part of speech corresponding to the translation, and the reference for determining the order of the candidates obtained by the translation processing are referred to. Learning information, a first learning dictionary for storing the learning frequency of each learning information, and learning registration means for registering the translation information retrieved by the registration information retrieval means as learning information in the first learning dictionary, The frequency of storage in the first learning dictionary is determined according to the learning information that is registered by the learning registration unit and that is extracted by the registration information extracting unit. Further frequency changing means, a learning dictionary reference means for extracting the learning information corresponding to the translation target word stored in the first learning dictionary together with the corresponding frequency, and the learning dictionary reference means for extracting the learning information. It is characterized by further comprising a second result replacement means for replacing the candidate ranks of the translation information stored in the result table according to the frequency of the learning information.

As a result, the learning information stored in the learning dictionary is added with information on the frequency of each learning target, and the order of a plurality of translation candidates obtained by the translation process is changed according to the frequency. Not only the latest learning result but also the past learning result can be reflected in the translation result.

A sixth invention is obtained by a translation process for translating a sentence in the first language into a sentence in the second language,
A translation candidate having a candidate rank for each word, a result table for storing translation information including words and parts of speech corresponding to the translation, and a learning execution instruction for the translation information stored in the result table. Accordingly, the translation of the translation information to be learned from the result table, the registration information extracting means for extracting the word and the part of speech corresponding to the translation, and the reference for determining the order of the candidates obtained by the translation processing are referred to. The learning information, the first learning dictionary for storing the learning frequency of each learning information, and the translation information extracted by the registered information extracting means are used as learning information in the first learning dictionary according to the learning order. Learning registration means to register,
A frequency changing unit for changing the frequency stored in the first learning dictionary according to the learning information to be registered by the learning registration unit, which is extracted by the registration information extracting unit; Learning dictionary reference means for extracting the stored learning information corresponding to the word to be translated, together with the corresponding frequency, so that the learned order can be identified, and the learning extracted by the learning dictionary reference means. A second result interchange unit for exchanging the candidate ranks of the translation information stored in the result table according to the frequency of information, and a first result interchange unit stored in the result table.
It is characterized by further comprising a third result replacing means for replacing the translation information of the rank according to the latest learning information retrieved by the learning dictionary referring means.

As a result, the learning information stored in the learning dictionary is added with the information of the frequency of the learning target, and the order of the plurality of translation candidates obtained by the translation process is changed according to the frequency, and further updated. By changing according to the learning result of, the past learning result and the latest learning result can be reflected in the translation result.

Further, the seventh invention comprises dictionary frequency changing means for decreasing the frequencies corresponding to all the learning information stored in the first dictionary according to a specific rule so as to make a difference in a low frequency. Characterize.

Thus, when the frequency of addition to the learning information and storage in the learning dictionary is reduced as a whole, the learning information having a low frequency is differentiated so that the learning information having a low frequency is also reflected in the translation result. Can be made.

An eighth aspect of the present invention comprises switching means for switching a mode indicating the strength of reflecting the learning effect on the translation result, and the learning dictionary reference means has learning information according to the mode switched by the switching means. Is taken out and reflected in the replacement of the candidate rank of the translation information.

Thus, by switching the strength of the condition for reflecting the learning effect, it is possible to arbitrarily limit or relax the output of the translation that is not included in the translation candidates by the normal translation processing.

[0025]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing a configuration example of a system that realizes the translation processing device according to the present embodiment. The first language and the second language which are the objects of the present invention
The language may be any language such as Japanese, English, German, French, etc., but in this embodiment,
An example will be described in which English is treated as the first language and Japanese is treated as the second language, and an English sentence is translated into a Japanese sentence.

In FIG. 1, 11 is a microprocessor (hereinafter referred to as CPU), and 12 and 13 are CPU 11.
Memory (ROM (Read Only Memo
ry) and RAM (Randam Access Memory).

The CPU 11 controls the entire apparatus, and controls the translation processing function by accessing the ROM 12 and the RAM 13 and starting a program according to an input instruction.
The translation processing function is provided with a function (learning function) for learning the result obtained by the translation processing.

The ROM 12 has a program area 1 in which a translation processing program for determining the operation of the CPU 11 is stored.
In addition to 2a, there is provided a translation dictionary area 12b in which a translation dictionary to be referred to when performing a translation process is stored.

The RAM 13 is for storing the original sentence in English in the first language, which is the translation target when the translation processing function (including the learning function) is executed, and the translated sentence in Japanese generated by the translation processing. A buffer and the like are provided. A result table 1 for storing translation information including translations of translation candidates obtained by translation processing in the RAM 13.
3a, a first learning dictionary 13b and a second learning dictionary 13c for storing learning information referred to for determining the rank of candidates obtained by the translation process, and a first learning for reflecting the learning effect on the translation candidates. A learning memory 13d for storing learning information read from the dictionary 13b or the second learning dictionary 13c, and a mode determination memory 13 for storing a mode indicating the strength of reflecting the learning effect on the translation result.
Each area for e is provided.

Reference numeral 14 is a keyboard (KB), 15 is a keyboard controller (KB-CNT), 16 is a floppy disk device (FDD), 17 is an FDD controller (FDD-CNT), and 18 is a printer device (PR).
T) and 19 are printer controllers (PRT-CN)
T) and 20 are print buffers.

The keyboard 14 is an input device for inputting data and giving a specific instruction. The keyboard 14 includes a “character” key, a “numerical value” key, a “cursor” key, a “select / execute” key, a “cancel” key, and various keys for inputting instructions for various functions, for example, a plurality of keys. "Translation word conversion" key 14a for sequentially switching and displaying translations of translation candidates, "translation word determination" key 14b for selecting a translation to be a final translation result from translations of a plurality of translation candidates, learning effect A "mode switching" key 14c for instructing the conversion of the condition (mode) to be reflected is set.

The floppy disk device 16 is, for example, for 3.5 inch type high density recording and stores / recalls a created document or the like. The FDD controller 17 is a CPU
Input / output control of data between 11 and the floppy disk device 16 is performed.

The printer device 18 comprises, for example, a thermal / thermal transfer type dot printer, and prints a created document or the like. The printer controller 19 expands print data in the print buffer 20 under the control of the CPU 11 and outputs this to the printer device 18.

Further, 21 is a display, 22 is a display controller (display-CNT), and 23 is a VRAM (Video RAM).
) And 24 are display signal synthesizers.

The display 21 is composed of, for example, a color liquid crystal display (color LCD) with a backlight, a CRT, or the like. The display controller 22 is a CPU
The display data is expanded in the VRAM 23 under the control of 11.
It is output to the display 18 via the display signal synthesizer 24.

Reference numeral 25 is a video signal converter, 26 is a video input interface (video input I / F), 27 is a TV tuner, and 28 is a frame memory.

The video signal converter 25 is for inputting a moving image (video), and inputs a video signal input via the video input I / F 26 or the TV tuner 27 to perform A / D conversion. , Into data of a predetermined format to be displayed on the display 21 via the display signal synthesizer 24. The video signal converter 25 expands the display data obtained by converting the video signal in the frame memory 28. The display data stored in the frame memory 28 is provided for display on the display 21 via the display signal synthesizer 24.

The video input I / F 26 inputs various kinds of video signals from the video equipment into the document creating apparatus and outputs them to the video signal converter 25 via the frame memory 28. TV
The tuner 27 receives radio waves for television broadcasting and outputs a video signal of a predetermined channel to the video signal converter 25.

The display signal synthesizer 24 is composed of the frame memory 2
VRAM2 and display data for video display expanded in 8
The display data for displaying the document No. 3 and the like are combined and output to the display 21. The display signal synthesizer 24 uses CP
It has a function of controlling the display screen of the display unit 21 under the control of the display-CNT 22 in response to the notification from the U11.

Further, 29 is a monitor device (monitor), 30
Is a video output interface (video I / F), and 31 is a color scanner device (scanner). The document creation apparatus of this embodiment can display various information such as document information on the monitor 29. The monitor 29 is connected to a dedicated connector provided on the apparatus body. The video output I / F 30 is
The output control of the image information to the monitor 29 is performed.

Further, in this apparatus, the color scanner 31
The image information read in can be incorporated into the document information. The color scanner 31 is connected to a dedicated connector provided on the apparatus body. Scanner I / F 32
Controls the input of image information by the color scanner 31.

Next, a first embodiment of the present invention will be described.

FIG. 2 is a block diagram showing the configuration of the translation processing function in the first embodiment. As shown in FIG. 2, the translation processing function includes a translation processing unit 41, a translation dictionary 42, a result table 44, a registration information extracting unit 45, and a learning registration unit 4.
7, a first learning dictionary 48, a learning dictionary reference unit 50, a learning memory 51, and a first result replacement unit 53. (The second learning dictionary 49 is used in the second embodiment described later).

The translation processing unit 41 translates the original sentence in the first language (English) to be translated into the translated sentence in the second language (Japanese) based on the dictionary information stored in the translation dictionary 42. Execute the translation process. In the translation process, translation information that includes translations that have a candidate rank for each word and words and parts of speech corresponding to the translations is obtained. Translation processing unit 41
Causes the result table 44 to store a word (original sentence, here, English spelling) and a part of speech corresponding to each translation in order from the translation of the first candidate with the first candidate rank. In addition, the translation processing unit 41 requests the learning dictionary reference unit 50 to execute the reference process in order to reflect the learning on the translation candidates.

In the translation dictionary 42, dictionary information used for translation processing by the translation processing unit 41 is registered.

The result table 44 is for storing the translation information obtained by the translation processing by the translation processing unit 41 for each word (English spelling). The translation information includes English spelling, English part-of-speech, translation of translation candidates (first candidate-) and part-of-speech corresponding to the translation. In addition,
It is not always possible to obtain multiple translation candidates.
In some cases, only the translation of the term or a candidate translation cannot be obtained.

The registered information extracting section 45 responds to the instruction to execute the learning with respect to the translation information stored in the result table 44, from the result table 44, the translation of the translation information to be learned, the word corresponding to the translation, "Information to learn" consisting of parts of speech is retrieved. That is, information to be learned = {English spelling, English part of speech, learned translation}.

The learning registration section 47 includes a registration information extraction section 4
Translation information retrieved by 5 ("information to learn")
As learning information, the first learning dictionary 4 is used according to the order of learning.
Register to 8.

The first learning dictionary 48 stores learning information ({English spelling, English part of speech, learned translation}) that is referred to for determining the rank of candidates obtained by the translation process. The structure is such that the order in which they are performed can be understood. The learning information is generated by the “information to be learned” extracted by the registration information extracting unit 45 being stored in the first learning dictionary 48 by the learning registration unit 47.

The learning dictionary reference unit 50, in response to the execution request of the reference process from the translation processing unit 41, performs the first learning corresponding to the word to be translated, that is, the translation information stored in the result table 44. The learning information stored in the dictionary 48 is extracted so that the learning order can be identified,
It is stored in the learning memory 51.

The learning memory 51 stores the learning information extracted by the learning dictionary reference unit 50 in order (so that the learning order can be known).

The first result replacement unit 53 replaces the candidate rank of the translation information stored in the result table 44 according to the learning information retrieved by the learning dictionary reference unit 50, and finally outputs the translation result. (Translation of the first candidate to be output) is determined.

Next, the operation of the first embodiment will be described.

As an example, the learning information about the word (English spelling) "live" is obtained by learning the translation result of the English sentence (1) "We cannot live without air." Shown in FIG. Register and write in English (2)
The registration process for registering the learning information and the reference process for reflecting the learning information in the learning result will be described by taking as an example the case where the learning is reflected when the translation of "We cannot live without water."

(Registration processing) As a result of the translation processing of the English sentence (1) by the translation processing unit 41, the translation of the English spell "live" becomes "live" according to the output order when there is no learning. At this time, in the result table 44, as shown in FIG. 4, for the English spell “live”, the English part-of-speech “verb”, the first candidate translation “live”, and the first candidate translation part-of-speech “verb”. Is stored. In addition, the translation of the second candidate “live” and the part of speech “verb” of the translation of the second candidate are stored (not shown).

Therefore, the translation result of the English sentence (1) when there is no learning is "we cannot survive without air".

At this point, the user operates the "translation decision" key 14a of the keyboard 14 to "live (liv)".
The translation of "e)" is changed to the translation of the second candidate, "living." As a result, the translation information as shown in FIG. 5 is stored for the English sentence (1) in the result table 44 (only the translation of the first candidate is shown).

(1) When the learning execution is instructed by the operation of the "translation word decision" key 14a, the registration information extracting section 45
Starts the learning registration process shown in the flowchart of FIG. That is, the registration information extracting unit 45 uses the “learning information” from the result table 44, that is, the English spelling,
English part-of-speech and learned translation (translation selected by the user) are retrieved. In this case, the information to be learned is {live, verb, live}.

(2) The learning registration unit 47 registers the "information to be learned" extracted by the registration information extraction unit 45 as learning information in the first learning dictionary 48 (step A).
1).

(Reference processing) The translation processing unit 41 translates the English sentence (2) "We cannot live without water." Shown in FIG.

(3) When the translation information obtained by the translation processing by the translation processing section 41 is stored in the result table 44, the learning dictionary reference section 50 starts the reference processing shown in the flowchart of FIG. The learning dictionary reference unit 50 extracts from the result table 44 the English spelling and English part-of-speech of the translation information targeted for learning reference (step B1).

The learning dictionary reference unit 50 refers to whether the learning information matching the extracted translation information {English spelling, English part of speech} is registered in the first learning dictionary 48 (step B2). Further, the learning dictionary reference unit 50 obtains the learning information in which the learned version (the translation selected by the user registered in the registration process) of the matching learning information matches the translation candidate translation stored in the result table 44. All are taken out and stored in the learning memory 51 (step B3).

Here, there are cases where the translation information for learning reference stored in the result table 44 has translations of a plurality of translation candidates. Just match. Further, there may be a plurality of pieces of learning information (learned translations) that match the translation information {English spelling, English part-of-speech}, so all the matching information is extracted.

For example, as a result of the translation processing of the English sentence (2) by the translation processing unit 41, if the learning is not reflected, the translation of the English spell "live" is shown in FIGS.
As shown in, it becomes "survived" according to the output order when there is no learning.

The learning dictionary reference unit 50 uses the translation information {live,
For the verb}, the first learning dictionary 48 is referenced to search for matching learning information. As a result, the corresponding learning information {li
There is one ve, verb, live. Further, the learning dictionary reference unit 50 compares the translation candidate “live, live” of the translation information with “live” which is the {learned translation> of the learning information, and there is a matching translation “live”. Therefore, the learning dictionary reference unit 50 stores this learning information in the learning memory 51. Here, since there is only one piece of learning information, other learning information is not extracted, but when there are a plurality of pieces of learning information, the learning memory 51 is provided so that the learning order can be known.
Learning information is stored in.

(4) When the learning information is stored in the learning memory 51, the first result replacement section 53 translates the corresponding translation candidates stored in the result table 44 according to the learning order of the learning information. Are replaced (step B
4). That is, the first result replacement unit 53 rearranges the order of the translation candidates from the translation corresponding to the new learning information to the translation corresponding to the old learning information. For translation candidates that have no corresponding learning information, the candidate rank is lowered.

Here, the first result replacement unit 53 sets the translation candidate translation “live” to the order of the translation candidate translation “live” stored for the word “live” stored in the result table 44. Change to become a translation of the first candidate and change to "live, live".

(5) As a result, the final translation result for the English sentence (2) becomes "we cannot live without water".

Next, the reference process will be described using another example. The translation processing unit 41 uses the English sentence (3) “Most p shown in FIG.
Translate "eople live by working".

As a result of the translation processing of the English sentence (3) by the translation processing unit 41, if learning is not reflected,
As shown in FIGS. 3 and 4, the English translation of “live” is “make a living” according to the output order when there is no learning.

(3) The learning dictionary reference section 50 refers to the first learning dictionary 48 for translation information {live, verb},
Search for matching learning information. As a result, there is one piece of relevant learning information {live, verb, live}. further,
The learning dictionary reference unit 50 translates the translation information “can make a living, live, live” and the learning information {learned translation}.
Comparing with "live", there is a matching translation "live". Therefore, the learning dictionary reference unit 50 stores this learning information in the learning memory 51.

(4) The first result replacement section 53 changes the translation of the translation candidates for the word "live" stored in the result table 44, "making a living, living, living", to the translation of the third candidate. Change to "live, earn a living, live" by replacing "living" so that it becomes the first translation.

(5) As a result, the final translation result for the English sentence (3) becomes "people live by working".

As described above, according to the translation processing function of the first embodiment, the learning information registration process and the reference process are performed without analyzing the front-rear relationship and the inter-relationship relationship between words, and the translation processing unit Learning can be reflected in the translation candidates obtained by 41. Therefore, the time required for the processing of the translation processing function including the learning function is shortened, and the translation processing is efficiently performed. Further, the learning dictionary stored in the first learning dictionary 48 includes only the English spelling, the English part of speech, and the learned translation (translation selected by the user), and the learning function is realized with a small amount of information. There is.

Next, a second embodiment of the present invention will be described.

The translation processing function of the second embodiment uses the second learning dictionary 49 in the configuration shown in FIG. 2 used in the description of the first embodiment (the first learning dictionary 48 is not used).

The second learning dictionary 49 stores learning information referred to for determining the rank of candidates obtained by the translation process, and has a structure in which the learning order is known. The learning information is generated by storing the “information to be learned” extracted by the registration information extracting unit 45 in the second learning dictionary 49 by the learning registration unit 47.

The "learning information" fetched from the result table 44 by the registration information fetching unit 45 in the second embodiment is: learning information = {English spelling, English part-of-speech, first candidate translation, first candidate The part of speech of the translation, the translation that was learned}. Therefore, the learning information consisting of the information to be learned ({English spelling, English part of speech, first candidate translation, first candidate translation part of speech, learned translation}) is registered in the second learning dictionary 49. It

The functions of the other parts are basically the same as those in the first embodiment except that the learning information to be handled is different because the second learning dictionary 49 is used, and therefore detailed description thereof will be omitted.

Next, the operation of the second embodiment will be described.

Here, in the same manner as the registration processing described in the first embodiment, the English sentence (1) "We cannot live w.
It is assumed that learning information about the word (English spelling) “live” is registered in the second learning dictionary 49 by learning the translation result for “ith out air.”.

(Registration processing) As a result of the translation processing of the English sentence (1) by the translation processing unit 41, the translation of the English spelling "live" becomes "live" according to the output order when there is no learning. At this time, in the result table 44, as shown in FIG. 4, for the English spell “live”, the English part-of-speech “verb”, the first candidate translation “live”, and the first candidate translation part-of-speech “verb”. Is stored. In addition, the translation of the second candidate “live” and the part of speech “verb” of the translation of the second candidate are stored (not shown).

At this point, the user operates the "translate word decision" key 14a of the keyboard 14 to "live (liv)".
The translation of "e)" is changed to the translation of the second candidate, "living." As a result, the translation information as shown in FIG. 5 is stored for the English sentence (1) in the result table 44 (only the translation of the first candidate is shown).

(1) When learning execution is instructed by operating the "translation word decision" key 14a, the registration information extracting section 45
Starts the learning registration process shown in the flowchart of FIG. That is, the registration information extracting unit 45 uses the “learning information” from the result table 44, that is, the English spelling,
The English part of speech, the translation of the first candidate, the part of speech of the translation of the first candidate, and the learned translation (translation selected by the user) are taken out. in this case,
The information to be learned is {live, verb, live, verb, live}.

(2) The learning registration unit 47 registers the “information to be learned” extracted by the registration information extraction unit 45 in the second learning dictionary 49 as learning information (step C).
1).

(Reference processing) The translation processing unit 41 translates the English sentence (2) "We cannot live without water." Shown in FIG.

(3) When the translation information obtained by the translation processing by the translation processing unit 41 is stored in the result table 44, the learning dictionary reference unit 50 starts the reference processing shown in the flowchart of FIG. The learning dictionary reference unit 50 extracts from the result table 44 the English spelling, the English part of speech, the translation of the first candidate, and the part of speech of the translation of the first candidate of the translation information targeted for learning reference (step D1).

The learning dictionary reference section 50 determines that the learning information that matches the extracted translation information {English spelling, English part of speech, first candidate translation, first candidate translation part of speech}
Whether it is registered in No. 9 is referred to (step D2). Further, the learning dictionary reference unit 50 obtains the learning information in which the learned version (the translation selected by the user registered in the registration process) of the matching learning information matches the translation candidate translation stored in the result table 44. All are taken out and stored in the learning memory 51 (step D3).

Here, since there are cases where a plurality of translation candidate translations exist in the translation information to be the learning reference target stored in the result table 44, in that case, any one of these translations is translated. Just match. Also, translation information {English spelling, English part-of-speech, translation of first candidate, part-of-speech of first candidate translation}
There may be a plurality of pieces of learning information (learned translations) that match with, so that all the pieces of matching information are extracted.

For example, as a result of the translation processing of the English sentence (2) by the translation processing unit 41, if the learning is not reflected, the translation of the English spelling "live" is as shown in FIGS.
As shown in, it becomes "survived" according to the output order when there is no learning.

The learning dictionary reference unit 50 uses the translation information {live,
For the verb, live, verb, live}, the second learning dictionary 49 is referred to and matching learning information is searched. As a result, the corresponding learning information {live, verb, survive, verb,
There is one). Furthermore, the learning dictionary reference unit 50
Comparing the translation candidate “live, live” of the translation information with the “live”, which is the {learned translation> of the learning information, there is a matching translation “live”. Therefore, the learning dictionary reference unit 50 stores this learning information in the learning memory 51. Here, since there is only one piece of relevant learning information, other learning information is not extracted, but when there are a plurality of learning information, the learning information is stored in the learning memory 51 so that the learning order can be known.

(4) When the learning information is stored in the learning memory 51, the first result replacement section 53 translates the corresponding translation candidates stored in the result table 44 according to the learning order of the learning information. Are replaced (step D
4). That is, the first result replacement unit 53 rearranges the order of the translation candidates from the translation corresponding to the new learning information to the translation corresponding to the old learning information. For translation candidate translations for which corresponding learning information does not exist, the candidate ranks are lowered in order after the translation of the translation candidate that has been learned.

Here, the first result replacement section 53 determines the order of the translation candidate translation “live” to be stored in the result table 44 as the translation candidate translation “live” to be the order of the second candidate translation “live”. Change to become a translation of the first candidate and change to "live, live".

(5) As a result, the final translation result for the English sentence (2) becomes "we cannot live without water".

Next, the reference process will be described using another example. The translation processing unit 41 uses the English sentence (3) “Most p shown in FIG.
Translate "eople live by working".

As a result of the translation processing of the English sentence (3) by the translation processing unit 41, if learning is not reflected,
The translation of the English spell "live" is "make a living" according to the output order when there is no learning, as shown in FIGS.

(3) The learning dictionary reference unit 50 refers to the second learning dictionary 49 for the translation information {live, verb, make a living, verb} and searches for matching learning information. As a result, there is no corresponding learning information.

(4) As a result, the first result replacement unit 53
Because the translation order of translation candidates cannot be changed by
The final translation result for English (3) is "People earn a living by working."

In this way, according to the translation processing function of the second embodiment, the learning information is added to the learning information of the first embodiment {translation of the first candidate, part of speech of the translation of the first candidate}. Therefore, the same effect as that of the first embodiment can be obtained. In addition, learning information {English spelling, English part of speech, first candidate translation, first candidate translation part of speech, learned translation}
With this, the condition for reflecting the learning effect can be made tighter (stronger) than in the case of the first embodiment. That is, when the learning is effective as in the English sentence (2) described above, the translation result may be improved, but when the learning is effective as in the English sentence (3), the translation result may be deteriorated. According to the second embodiment, it is possible to prevent learning from being applied to the translation process of the English sentence (3).

Next, a third embodiment of the present invention will be described.

FIG. 10 is a block diagram showing the configuration of the translation processing function in the third embodiment. As shown in FIG. 10, the translation processing function includes a translation processing unit 41, a translation dictionary 42,
The result table 44, the registration information extraction unit 45, the learning registration unit 47, the second learning dictionary 49, the learning dictionary reference unit 50a, the learning memory 51, the first result replacement unit 53, and the appropriate translation determination unit 54 are included.

The translation processing function of the third embodiment uses the second learning dictionary 49 in the same manner as in the case of the second embodiment described above, and uses all the information included in the learning information and uses some of the information. Case (equivalent to executing the first embodiment)
Is selected according to the translation candidate, the condition for reflecting the learning effect is switched.

In FIG. 10, the same parts as those in the first or second embodiment are designated by the same reference numerals and the description thereof will be omitted.

The learning dictionary reference unit 50a includes the translation processing unit 41.
In response to the execution request of the reference process from the learning information stored in the second learning dictionary 48 corresponding to the word to be translated, that is, the translation information stored in the result table 44 ({English spelling, English part of speech, first candidate translation, first
The part of speech of the candidate translation, the learned translation}) is extracted so that the learned order can be identified, and stored in the learning memory 51. In addition, the learning dictionary reference unit 50a includes an appropriate translation determination unit 54
In response to the notification from, part of the learning information ({English spelling, English part of speech, learned translation}) stored in the second learning dictionary 49 is extracted and stored in the learning memory 51.

The appropriate translation determination unit 54 determines whether or not a translation that is an appropriate translation candidate is stored as a translation result in the result table 44, and notifies the learning dictionary reference unit 50a.

Next, the operation of the third embodiment will be described.

Here, as an example, the learning information about the word (English spelling) “live” is obtained by learning the translation result of the English sentence (1) “We cannot live without air.” Shown in FIG. Register and write in English (3)
The registration process for registering the learning information and the reference process for reflecting the learning information in the learning result will be described by taking as an example the case where the learning is reflected when translating “Most people live by working”.

(Registration Processing) (1) The registration processing in the third embodiment is performed in the same manner as in the second embodiment, so description thereof will be omitted. The second learning dictionary 49 stores learning information having information of {live, verb, live, verb, live}.

(Reference Processing) (2) When the translation information obtained by the translation processing by the translation processing section 41 is stored in the result table 44, the learning dictionary reference section 50 performs the learning reference processing shown in the flowchart of FIG. Start. First, a translation process by learning reference using the second learning dictionary 49 is performed (step E1). The process in step E1 is the learning reference process shown in the flowchart of FIG. 9 used in the description of the second embodiment.

In other words, the learning dictionary reference unit 50 extracts from the result table 44 the English spelling, the English part of speech, the translation of the first candidate, and the translation of the first candidate of the translation information targeted for learning reference, Second learning information that matches this translation information
Obtained by referring to the learning dictionary 49 (steps D1 and D
2).

If there is the corresponding learning information, the learning dictionary reference unit 50 extracts all learning information in which the learned translation of the learning information matches the translation of the translation candidate stored in the result table 44 and retrieves the learning memory 51. Stored in (step D
3). The first result replacement unit 53 replaces the translation of the corresponding translation candidate stored in the result table 44 according to the learned order of the learning information stored in the learning memory 51 (step D4).

(3) Here, the appropriate translation determination unit 54 refers to the result table 44 and determines whether or not a translation that is an appropriate translation candidate is stored as the translation result (step E).
2). That is, in the case where no translation candidates are stored in the result table 44 or only candidates with low priority (should not be output as translation results),
The appropriate translation determination unit 54 notifies the learning dictionary reference unit 50a.

(4) Upon receiving the notification from the proper translation determining unit 54, the learning dictionary reference unit 50a performs a translation process by a learning process using a part of the learning information stored in the second learning dictionary 49 (step). E3). The process in step E3 is the learning reference process shown in the flowchart of FIG. 7 used in the description of the first embodiment. Although there is a difference that the first embodiment targets the first learning dictionary 48,
Part of the learning information stored in the second learning dictionary 49 ({English spelling, English part of speech, learned translation}), that is, the first
By using the same information as the learning information stored in the learning dictionary 48, substantially the same processing is performed.

The learning dictionary reference section 50 uses the result table 44.
From the English spelling of translation information, which is the target of learning reference,
The English part-of-speech is extracted, and learning information that matches this translation information is obtained by referring to a part of the learning information ({English spelling, English part-of-speech, learned translation}) stored in the second learning dictionary 49 ( Steps B1, B2).

If there is the corresponding learning information, the learning dictionary reference unit 50 extracts all the learning information in which the learned translation of the learning information matches the translation candidate translation stored in the result table 44 and retrieves the learning memory 51. Stored in (step B
3). The first result replacement unit 53 replaces the translation of the corresponding translation candidate stored in the result table 44 according to the learned order of the learning information stored in the learning memory 51 (step B4).

For example, in English (3) "Most people live b
When translating for "y working", it becomes as follows.

First, in step E1, with respect to the translation information {live, verb, make a living, verb} obtained by the translation process, the second learning dictionary 49 is referred to, and when matching learning information is searched, the corresponding learning information is obtained. There is no information.
Therefore, the learning effect is not reflected in the translation candidates. As a result, in step E2, it is assumed that the translation candidate stored in the result table 44 is determined by the translation translation determination unit 54 as a translation result that is not a translation.

In this case, in step E3, with respect to a part {live, verb} of the translation information obtained by the translation processing, a part of the learning information stored in the second learning dictionary 49 is referred to, and matching learning information is obtained. To search. Second learning dictionary 49
Stores learning information {live, verb, live, verb, live} in the registration processing, a part of {live, verb, live} is extracted by the learning dictionary reference unit 50a, and the learning memory 51 is stored. Stored in.

Therefore, the first result replacing section 53 changes the translation of the translation candidates for the word "live" stored in the result table 44, "making a living, living, living".
The third candidate's translation "live" is replaced with the first candidate's translation and changed to "live, earn a living, live".

(5) As a result, the final translation result for the English sentence (3) is "people live by working" (step E4).

As described above, according to the translation processing function of the third embodiment of the present invention, the learning information stored in the second learning dictionary 49 is used to more strongly reflect the learning effect on the translation result. If there is no suitable translation, a part of the learning information is used to slow down the learning effect and a translated word is required, so avoid a situation where a translated word cannot be obtained or an inappropriate translated word is obtained. can do.

Next, a fourth embodiment of the present invention will be described.

The translation processing function of the fourth embodiment can be realized by the same configuration as that shown in FIG. 10 used in the description of the third embodiment. However, since the functions of the learning dictionary reference unit 50a, the first result replacement unit 53, and the appropriate translation determination unit 54 are slightly different from those of the third embodiment, only the different units will be described and detailed description of other units will be given. The description is omitted.

The learning dictionary reference unit 50a includes the translation processing unit 41.
The learning information stored in the second learning dictionary 49 corresponding to the translation target word, that is, the translation information stored in the result table 44, is learned in the learning order in response to the execution request of the reference processing from Are stored so that they can be identified and stored in the learning memory 51 (the same as the learning dictionary reference unit 50 of the second embodiment).

The first result replacement unit 53 replaces the candidate ranks of the translation information stored in the result table 44 according to the learning information retrieved by the learning dictionary reference unit 50, and finally outputs the translation result. (Translation of the first candidate to be output) is determined. In addition, the first result replacement unit 53 replaces the translation candidates stored in the result table 44 and adds the learning information stored in the learning memory 51, when notified by the proper translation determination unit 54.

The appropriate translation determining unit 54 determines whether or not a translation that is an appropriate translation candidate is stored as the translation result in the result table 44, and notifies the learning dictionary reference unit 50a and the first result replacement unit 53 of the result. .

Next, the operation of the fourth embodiment will be described.

(Registration Processing) (1) The registration processing in the fourth example is performed in the same manner as in the second embodiment, so description thereof will be omitted. In the second learning dictionary 49, learning information about the word (English spelling) “live” {English spelling, English part of speech, first candidate translation,
As the part of speech of the translation of the first candidate, the learned translation}, learning information having information of {live, verb, live, verb, live) is stored.

(Reference Processing) (2) When the translation information obtained by the translation processing by the translation processing section 41 is stored in the result table 44, the learning dictionary reference section 50 performs the learning reference processing shown in the flowchart of FIG. Start. First, a translation process by learning reference using the second learning dictionary 49 is performed (step F1). The process of step F1 is the same as the process of step E1 of the flowchart of FIG. 11 used in the description of the third embodiment, and therefore the description is omitted.

(3) Here, the appropriate translation determining unit 54 refers to the result table 44 and determines whether or not a translation that is an appropriate translation candidate is stored as the translation result (step F).
2). When a translation that is an appropriate translation candidate is not stored as the translation result, the appropriate translation determination unit 54 notifies the learning dictionary reference unit 50a and the first result replacement unit 53.

(4) Upon receipt of the notification from the proper translation determining unit 54, the learning dictionary reference unit 50a learns the learning information stored in the second learning dictionary 49 corresponding to the word (English spelling) to be translated. All the translated texts are extracted and stored in the learning memory 51 (step F3). That is, the learning dictionary reference unit 50a executes the processes of steps D1 to D3 of the flowchart shown in FIG. 9 described in the second embodiment.

(5) The first result replacement unit 53 adds the learned translation of the learning information stored in the learning memory 51 to the translation of the corresponding translation candidate in the result table 44, and replaces the translation candidate order. Do. Here, the learned translation included in the learning information is added as the translation of the first candidate. When a plurality of pieces of learning information are present, the first candidate is added to the learning order (from the latest one).

For example, in English (1) "We cannot live wit
When you translate "hout air."

First, in step F1, with respect to the translation information {live, verb, live, verb} obtained by the translation process, the second learning dictionary 49 is referred to, and when matching learning information is searched, the corresponding learning information is found. There are {live, verb, live, verb, live}. However, the learned version of this learning information (which can be alive) does not exist in the translation of the translation candidate in the result table 44, and therefore is not reflected in the translation result.

As a result, in step F2, it is assumed that the translation candidate stored in the result table 44 is determined by the translation translation determination unit 54 as a translation result that is not a translation.

In this case, in step F3, the learned version of the learning information {live, verb, live, verb, alive} corresponding to the translation information {live, verb, live, verb} Is taken out, and the learning memory 51
Stored in.

Therefore, the first result replacement unit 53 adds the learned translation {existence of existence} to the translation candidate translation "live, live" stored in the result table 44 for the word "live". , "Live" is replaced by the first candidate and changed to "live, live, live".

(5) As a result, the final translation result for the English sentence (1) is "we cannot live without air" (step F4).

In this way, even if the learning contents are reflected in the translation process, if the appropriate translation determination unit 54 determines that an appropriate translation cannot be obtained, the translation process stored in the result table 44 is performed. Since translations of the learning information stored in the second learning dictionary 49, other than the obtained translation candidates, can be added as translation candidates, a situation in which a translation cannot be obtained or an inappropriate translation is obtained. It can be avoided.

Next, a fifth embodiment of the present invention will be described.

FIG. 13 is a block diagram showing the structure of the translation processing function in the fifth embodiment. As shown in FIG. 13, the translation processing function includes a translation processing unit 41, a translation dictionary 42,
The result table 44, the registration information extraction unit 45a, the learning registration unit 47, the second learning dictionary 49a, the learning dictionary reference unit 50.
b, the learning memory 51, the second result replacing section 56, and the table frequency changing section 57.

The same parts as those in FIG. 2 used in the description of the second embodiment are designated by the same reference numerals and the description thereof will be omitted.

The registered information fetching unit 45a responds to the learning execution instruction for the translation information stored in the result table 44, from the result table 44, the translation of the translation information to be learned and the words and words corresponding to the translation. "Information to learn" consisting of parts of speech is retrieved. That is, learning information = {English spelling, English part-of-speech, translation of first candidate, part-of-speech of first candidate translation, learned translation}. In addition, when the registration information extracting unit 45a extracts the “information to be learned”, the learning information corresponding to the “information to be learned” is the second learning dictionary 4
The learning dictionary reference unit 50b is notified to determine the situation stored in 9a.

The second learning dictionary 49a stores learning information referred to for determining the rank of candidates obtained by the translation process, and has a structure in which the learning order is known. Further, the second learning dictionary 49a stores not only the learning information but also the frequency indicating the number of times of learning for each learning information.

The learning dictionary reference unit 50b is the translation processing unit 41.
In response to the execution request of the reference process from, the learning information stored in the second learning dictionary 49a corresponding to the word to be translated, that is, the translation information stored in the result table 44, together with the corresponding frequency. , So that the learning order can be identified and stored in the learning memory 51. Further, the learning dictionary reference unit 50b refers to whether the same learning information corresponding to the “information to be learned” to be learned exists according to the notification from the registration information extracting unit 45a,
The learning registration unit 47 and the table frequency changing unit 57 are notified according to the reference result.

The second result replacement unit 56 replaces the candidate ranks of the translation information stored in the result table 44 according to the learning information extracted by the learning dictionary reference unit 50 and the corresponding frequency, and finally outputs it. The translation result (translation of the first candidate to be output) to be determined is determined.

The table frequency changing unit 57, according to the learning information to be registered by the learning registration unit 47, which is extracted by the registration information extracting unit, the learning dictionary reference unit 5
According to the notification from 0b, the frequency stored in the second learning dictionary 49a is changed.

Next, the operation of the fifth embodiment will be described.

(Registration processing) Here, learning information about the word (English spelling) "live" {English spelling, English part of speech, first candidate translation, first candidate translation part of speech, learned translation} As {live, verb, live, verb, live}
An example is the case of registering learning information having the information of. Fifth
The registration process in the example is basically performed in the same manner as in the second embodiment, but the process of the flowchart shown in FIG. 14 is further added.

(New Registration Processing) (1) First, when learning execution is instructed, the registration information fetching unit 45 reads “information to be learned” from the result table 44.
Take out {live, verb, live, verb, live},
The learning dictionary reference unit 50b is notified.

(2) The learning dictionary reference unit 50b refers to whether the same learning information corresponding to the "information to be learned" notified from the registration information extracting unit 45a is already registered in the second learning dictionary 49a (step G1).

(3) Since the same learning information does not exist here, the learning dictionary reference unit 50b notifies the learning registration unit 47 and the table frequency changing unit 57 of this fact. The learning registration unit 47 stores the “information to be learned” extracted by the registration information extraction unit 45a in the second learning dictionary 49a as learning information. Further, the table frequency changing unit 57
The frequency corresponding to the learning information newly registered by the learning registration unit 47 is set to the initial value 1 (step G3).

(Second and subsequent registration processing) (1) First, when the learning execution for the same "information to be learned" as described above is instructed, the registration information extracting section 45 similarly performs "learning" from the result table 44. "Information to be processed" is extracted and notified to the learning dictionary reference unit 50b.

(2) The learning dictionary reference unit 50b refers to whether the same learning information corresponding to the "information to be learned" notified from the registration information extraction unit 45a is already registered in the second learning dictionary 49a (step G1).

(3) Since the same learning information exists here, the learning dictionary reference unit 50b notifies the learning registration unit 47 and the table frequency changing unit 57 of this. The learning registration unit 47 re-registers the learning information instructed to be registered in a new storage position in the second learning dictionary 49a, that is, a position according to the learning order (it can be determined that the learning is the latest learning). . Further, the table frequency changing unit 57 adds 1 to the frequency corresponding to the relevant learning information (step G2).

(Reference Processing) (4) The learning dictionary reference unit 50b stores the translation information obtained by the translation processing by the translation processing unit 41 in the result table 44.
When stored in, the reference processing shown in the flowchart of FIG. 15 is started. The processing of the flowchart shown in FIG. 15 is basically the same as the processing of the flowchart shown in FIG. 9 described in the second embodiment, and processing using the frequency of learning information is added. The process of steps H1 and H2 of FIG. 15 is performed by step D shown in FIG.
Since the processing is the same as that of 1 and D2, the description is omitted.

The learning dictionary reference unit 50b takes out all the learning information in which the learned translations of the matching learning informations match the translations of the translation candidates stored in the result table 44, including the corresponding frequencies, and stores them in the learning memory. The data is stored in 51 (step H3).

(5) When the learning information and the frequency are stored in the learning memory 51, the second result replacement section 56 translates the corresponding translation candidates stored in the result table 44 according to the frequency of the learning information. The replacement is performed (step H4).
That is, the second result replacement unit 56 rearranges the order of the translation candidates from the translation corresponding to the learning information with high frequency to the translation corresponding to the learning information with low frequency. For translation candidate translations for which corresponding learning information does not exist, the candidate ranks are lowered in order after the translation of the translation candidate that has been learned.

In this way, according to the translation processing function of the fifth embodiment, the learning information stored in the second learning dictionary 49a is added with the information of the frequency of each learning target, and is obtained by the translation processing. Since the order of the obtained plurality of translation candidates is changed according to the frequency, not only the latest learning result but also the past learning result can be reflected in the translation result.

The method of changing the order of translation of translation candidates stored in the result table 44 according to the frequency added to the learning information in the fifth example is applied to the first to fourth embodiments, The same effect can be obtained.

Next, a sixth embodiment of the present invention will be described.

FIG. 16 is a block diagram showing a part of the configuration of the translation processing function in the sixth embodiment. The translation processing function of the sixth embodiment is obtained by adding a third result replacement unit 59 to the configuration shown in FIG. 13 used in the fifth embodiment.
FIG. 16 shows only the part related to the third result replacement unit 59. Illustration and detailed description of other parts are omitted.

The third result replacement unit 59 uses the learning dictionary reference unit 50 to convert the translation of the first candidate stored in the result table 44.
It is replaced according to the latest learning information extracted by b and stored in the learning memory 51.

Next, the operation of the sixth embodiment will be described.

The registration process is performed in the same manner as in the case of the fifth embodiment, and therefore its explanation is omitted.

(Reference Process) (1) In the learning dictionary reference unit 50b, the translation information obtained by the translation process by the translation processing unit 41 is used as the result table 44.
When stored in, the reference process shown in the flowchart of FIG. 17 is started. The process of the flowchart shown in FIG. 17 is basically the same as the process of the flowchart shown in FIG. 15 described in the fifth embodiment, and the process of changing the priority order of translation of translation candidates (step I4) is different. It is a thing.

Since the processing of steps I1, I2 and I3 of FIG. 17 is the same as the processing of steps H1, H2 and H3 shown in FIG. 15, description thereof will be omitted.

(2) When the learning information and the frequency are stored in the learning memory 51, the third result replacement unit 59 stores the corresponding translation candidates stored in the result table 44 according to the learning information recently learned. Replace the translation of with the translation of the first candidate. In addition, the second result replacement unit 56 uses the learning memory 51.
According to the frequency of the learning information stored in the result table 4
The translations of the second and subsequent candidates of the translation candidates stored in 4 are replaced (step I4). That is, the second result replacement unit 56 rearranges the order of the translations of the second and subsequent translation candidates in the order of the translation corresponding to the learning information with high frequency to the translation corresponding to the learning information with low frequency. For translation candidate translations for which corresponding learning information does not exist, the candidate ranks are lowered in order after the translation of the translation candidate that has been learned.

As described above, according to the translation processing function of the sixth embodiment, the learning information stored in the second learning dictionary 49a is added with the information of the frequency of each learning target, and the second result is replaced. By transposing the order of translation of the plurality of translation candidates obtained by the translation processing by the unit 56 according to the frequency, and further determining the first candidate by the third result interchange unit 59 according to the latest learning result, The learning result and the latest learning result can be reflected in the translation result.

Next, the seventh embodiment will be described.

FIG. 18 is a block diagram showing a part of the configuration of the translation processing function in the seventh embodiment. The translation processing function of the seventh embodiment can be applied to the configuration in which the learning dictionary stores the frequency indicating the number of times of learning for learning information, that is, the translation processing function of the fifth and sixth embodiments. A section 50c is provided. FIG. 18 shows only the portion related to the learning dictionary reference unit 50c. Illustration and detailed description of other parts are omitted.

The learning dictionary reference section 50c has the same function as the learning dictionary reference section 50b in the fifth embodiment. Furthermore, the learning dictionary reference unit 50c displays the frequencies corresponding to all the learning information stored in the second learning dictionary 49a.
A dictionary frequency changing function for lowering a low frequency according to a specific rule, that is, a dictionary frequency changing unit 61 and a frequency calculating unit 62 are provided.

The dictionary frequency changing unit 61 uses the second learning dictionary 49.
The frequency added to each learning information stored in a is taken out, transferred to the frequency calculation unit 62, and the frequency stored in the second learning dictionary 49a is changed to the new returned frequency.

The frequency calculation section 62 lowers the frequency corresponding to each learning information transferred from the dictionary frequency changing section 61 according to a specific rule to obtain a new frequency.

Next, the operation of the seventh embodiment will be described.

When performing the registration process in the same manner as in the fifth embodiment, if the table frequency changing unit 57 adds 1 to the frequency for certain learning information, the preset allowable value is exceeded. A process for changing (lowering) the frequency shown in is started.

In this case, the learning dictionary reference unit 50c uses the dictionary frequency changing unit 61 and the frequency calculating unit 62 to repeat the following processing from the beginning for all the learning information stored in the second learning dictionary 49a. To do.

The dictionary frequency changing unit 61 uses the second learning dictionary 49.
The frequency added to the learning information is extracted from a and transferred to the frequency calculation unit 62 (step J1). Frequency calculator 62
Reduces the frequency according to a specific rule to obtain a new frequency, and returns it to the dictionary frequency changing unit 61 (step J2).

FIG. 20 shows the frequencies corresponding to the plurality of pieces of learning information stored in the second learning dictionary 49a and the new frequencies. As shown in FIG. 20, for low frequencies, for example, "0-10", the values are not changed so that a difference in frequency appears. For frequencies of 11 or more, new frequency = (frequency−10) ÷ 2 + 10 is calculated to obtain a new frequency with a reduced frequency value.

The dictionary frequency changing unit 61 writes the new frequency obtained by the frequency calculating unit 62 back into the second learning dictionary 49a.

Thus, when the frequency of addition to the learning information 49a and stored in the learning dictionary is reduced as a whole, the learning information having a low frequency is differentiated from each other so that the translation candidates corresponding to the learning information having a low frequency are obtained. As for the translation of, the priority can be changed according to the frequency of actual learning, so that all the learning information can be reflected in the translation result.

Next, an eighth embodiment of the present invention will be described.

FIG. 21 is a block diagram showing the configuration of the translation processing function in the eighth embodiment. As shown in FIG. 21, the translation processing function includes a translation processing unit 41, a translation dictionary 42,
The result table 44, the registration information extraction unit 45, the learning registration unit 47, the second learning dictionary 49, the learning dictionary reference unit 50a, the learning memory 51, the first result replacement unit 53, and the appropriate translation determination unit 5
4a, mode determination memory 64, and mode switching unit 6
It is composed of 5.

The translation processing function of the eighth embodiment uses the second learning dictionary 49 as in the case of the second embodiment, and the second learning dictionary 49 is used.
The conditions for reflecting the learning effect are switched by arbitrarily switching and executing the functions in the embodiments, the third embodiment, and the fourth embodiment. In FIG. 21, the same parts as those in the second, third, or fourth embodiment are designated by the same reference numerals, and the description thereof will be omitted.

The appropriate translation determining section 54a is provided with the mode switching section 6
Executes the operation according to the instruction to switch the mode according to 5.
By stopping, the learning function in the above-described second, third, or fourth embodiment is executed.

The mode judgment memory 64 stores information indicating the condition (mode) for reflecting the learning effect.

[0187] The mode switching unit 65 performs switching to execute the learning information reference process with the learning intensity according to the mode stored in the mode determination memory 64.

Next, the operation of the eighth embodiment will be described with reference to FIG.
This will be described with reference to the flowchart shown in FIG.

First, when the user operates the "mode switching" key 14c of the keyboard 14 to instruct the conversion of the learning strength, that is, the condition (mode) for reflecting the learning effect, the mode is determined in the mode judgment memory 64. The information shown is held (step K1).

The mode switching unit 65 changes the mode by executing / stopping the operation of the appropriate translation determining unit 54a according to the information indicating the mode held in the mode determining memory 64 (step K2).

In the eighth embodiment, for example, it is assumed that the user can select the learning intensity in any of four levels {1, 2, 3, 4} (note that {1} gives the strongest learning effect). Mode). Each mode causes the learning function (reference processing) of the following embodiments to be executed. That is, in the case of mode 1, the third embodiment and the fourth embodiment, in the case of mode 2, the second embodiment and the fourth embodiment, in the case of mode 3, only the third embodiment, and in the case of mode 4, the second embodiment. It is only in form.

In the case of mode 1, the suitable translation judging unit 54a refers to the result table 44 to judge whether or not a translation that is an appropriate translation candidate is stored as the translation result, and there is no suitable translation. Even if it does, the reference process using a part of the learning of the second learning dictionary 49 is executed (third embodiment), and even if there is no translation candidate translation target to be learned in the result table 44, The translation of the learning information stored in the 2 learning dictionary 49 is added as a translation candidate (4th Embodiment), and a final translation is calculated | required.

In the case of mode 2, the appropriate translation determination unit 54a executes the reference process using the second learning dictionary 49 (second embodiment), and further refers to the result table 44 to refer to the translation candidates and perform the appropriate translation. If there is no translation, the translation of the translation candidate is added by the function of the fourth embodiment to obtain the final translation.

In the case of mode 3, the appropriate translation determining unit 54a refers to the result table 44 to refer to the translation candidates, and if there is no appropriate translation, adds the translation of the translation candidate by the function of the fourth embodiment, and finally. Ask for a translation. In the case of mode 4, the proper translation determination unit 54a does not operate and causes the second learning dictionary 49 to perform learning.

In the description of the eighth embodiment, the learning strength is switched by combining the functions described in the second, third, or fourth embodiments, but the learning function by another method is used. The strength of the learning effect may be switched by changing the learning effect.

As described above, according to the eighth embodiment of the present invention, by switching the strength of the condition for reflecting the learning effect in accordance with the instruction from the user, the translation candidate that is not included in the normal translation processing can be translated. The output of can be arbitrarily limited or relaxed as the user intends.

In the fourth to seventh embodiments described above, the second learning dictionary 49 (or the second learning dictionary 49a) is used, but the first learning dictionary 48 can also be used. is there. Further, even if the second learning dictionary 49 is used, if a part of the information included in the learning information {English spelling, English part of speech, learned translation} is targeted, the first learning dictionary 48 is substantially used. Is the same as.

[0198]

As described above in detail, according to the present invention, it is possible to provide a translation processing device capable of learning, which can efficiently obtain a translation result intended by a user.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration example of a system that realizes a translation processing device according to an embodiment of the present invention.

FIG. 2 is a block diagram showing a configuration of a translation processing function in the first and second embodiments.

FIG. 3 is a diagram showing an example of an English sentence used in the description of the embodiment.

FIG. 4 is a diagram showing an example of translation information about the English spelling “live” in the English sentence shown in FIG. 3.

5 is a diagram showing an example of translation information of each word obtained for the English sentence (1) shown in FIG.

FIG. 6 is a flowchart for explaining learning registration processing according to the first embodiment.

FIG. 7 is a flowchart for explaining learning reference processing according to the first embodiment.

FIG. 8 is a flowchart for explaining learning registration processing according to the second embodiment.

FIG. 9 is a flowchart illustrating a learning reference process according to the second embodiment.

FIG. 10 is a block diagram showing the configuration of a translation processing function in the third and fourth embodiments.

FIG. 11 is a flowchart for explaining learning reference processing according to the third embodiment.

FIG. 12 is a flowchart for explaining learning reference processing according to the fourth embodiment.

FIG. 13 is a block diagram showing the configuration of a translation processing function in the fifth embodiment.

FIG. 14 is a flowchart illustrating a learning registration process according to the fifth embodiment.

FIG. 15 is a flowchart for explaining learning reference processing according to the fifth embodiment.

FIG. 16 is a block diagram showing a part of the configuration of a translation processing function in the sixth embodiment.

FIG. 17 is a flowchart for explaining learning reference processing according to the sixth embodiment.

FIG. 18 is a block diagram showing a part of the configuration of a translation processing function in the seventh embodiment.

FIG. 19 is a flowchart illustrating a process of reducing the frequency according to the seventh embodiment.

FIG. 20 is a diagram for explaining a process of reducing the frequency according to the seventh embodiment.

FIG. 21 is a block diagram showing the configuration of a translation processing function according to the eighth embodiment.

FIG. 22 is a flowchart illustrating a mode switching process according to the eighth embodiment.

[Explanation of symbols]

 41 ... Translation processing unit 42 ... Translation dictionary 44 ... Result table 45 ... Registration information retrieval unit 47 ... Learning registration unit 48 ... First learning dictionary 49 ... Second learning dictionary 50 ... Learning dictionary reference unit 51 ... Learning memory 53 ... First Result replacement unit 54 ... appropriate translation judgment unit 56 ... Second result replacement unit 57 ... Table frequency change unit 59 ... Third result replacement unit 61 ... Dictionary frequency change unit 62 ... Frequency calculation unit 64 ... Mode determination memory 65 ... Mode switching unit

Claims (8)

[Claims] 1. A translation that is a translation candidate having a candidate rank for each word obtained by a translation process that translates a sentence in a first language into a sentence in a second language, and a translation including a word and a part of speech corresponding to the translation. A result table for storing information, a translation of the translation information to be learned from the result table, and a word corresponding to the translation, in response to a learning execution instruction for the translation information stored in the result table. The registration information extracting means for extracting the part of speech, the first learning dictionary for storing the learning information referred to for determining the rank of the candidates obtained by the translation process, and the translation information extracted by the registration information extracting means As learning information, a learning registration unit that registers in the first learning dictionary according to a learning order, and a word that is a target of translation processing stored in the first learning dictionary Learning dictionary reference means for extracting the learning information so that the learning order can be identified, and a candidate rank of the translation information stored in the result table according to the learning information extracted by the learning dictionary reference means. A translation processing device, comprising: a first result replacing means for replacing 2. A translation that is a translation candidate having a candidate rank for each word, obtained by a translation process that translates a sentence in a first language into a sentence in a second language, and a translation including a word and a part of speech corresponding to the translation. A result table for storing information, a translation of the translation information to be learned from the result table, and a word corresponding to the translation, in response to a learning execution instruction for the translation information stored in the result table. Part of speech,
To store the translation of the first candidate having the first candidate rank, the registration information extracting means for extracting the part of speech corresponding to the translation, and the learning information referred to for determining the rank of the candidate obtained by the translation process. No. 2 learning dictionary, learning registration means for registering the translation information extracted by the registration information extracting means as learning information in the second learning dictionary according to the order of learning, and stored in the second learning dictionary. A learning dictionary reference unit that extracts learning information corresponding to a word that is a translation processing target so that the learning order can be identified, and a learning dictionary reference unit that extracts the learning information that is extracted by the learning dictionary reference unit. And a first result replacing means for replacing the candidate ranks of the translation information stored in the result table. 3. A translation that is a translation candidate having a candidate rank for each word obtained by a translation process that translates a sentence in a first language into a sentence in a second language, and a translation including a word and a part of speech corresponding to the translation. A result table for storing information, a translation of the translation information to be learned from the result table, and a word corresponding to the translation, in response to a learning execution instruction for the translation information stored in the result table. Part of speech,
To store the translation of the first candidate having the first candidate rank, the registration information extracting means for extracting the part of speech corresponding to the translation, and the learning information referred to for determining the rank of the candidate obtained by the translation process. Of the second learning dictionary, learning registration means for registering the translation information extracted by the registration information extracting means as learning information in the second learning dictionary according to the order of learning, and a suitable translation result in the result table. Appropriate translation determining means for determining whether or not a translation that is a candidate for translation is stored, and learning information corresponding to a word to be translated which is stored in the second learning dictionary, in the learning order. A learning dictionary reference unit for selectively extracting a part of the learning information according to a result of the determination made by the appropriate translation determination unit, and the learning dictionary reference unit. Depending on the order of learning of the learning information extracted I, the translation processing apparatus characterized by comprising a first result interchanging means for interchanging the candidate ranking translation information stored in the results table. 4. A translation which is a translation candidate having a candidate rank for each word and which is obtained by a translation process for translating a sentence in a first language into a sentence in a second language, and a translation including a word and a part of speech corresponding to the translation. A result table for storing information, a translation of the translation information to be learned from the result table, and a word corresponding to the translation, in response to a learning execution instruction for the translation information stored in the result table. The registration information extracting means for extracting the part of speech, the first learning dictionary for storing the learning information referred to for determining the rank of the candidates obtained by the translation process, and the translation information extracted by the registration information extracting means As learning information, learning registration means for registering in the first learning dictionary according to the order of learning, and a translation that becomes an appropriate translation candidate as a translation result in the result table An appropriate translation determining means for determining whether or not it has been paid, and learning information corresponding to a word to be subjected to translation processing stored in the first learning dictionary according to a determination result by the appropriate translation determining means, A learning dictionary reference means for extracting the learned order so that the learning order can be identified; and a first result replacement means for storing the learning information extracted by the learning dictionary reference means in the result table as translation information. A translation processing device. 5. A translation that is a translation candidate having a candidate rank for each word obtained by a translation process that translates a sentence in a first language into a sentence in a second language, and a translation including a word and a part of speech corresponding to the translation. A result table for storing information, a translation of the translation information to be learned from the result table, and a word corresponding to the translation, in response to a learning execution instruction for the translation information stored in the result table. Registration information extracting means for extracting a part of speech, learning information referred to for determining the order of candidates obtained by the translation process, a first learning dictionary for storing the learning frequency of each learning information, and the registration information Learning registration means for registering the translation information extracted by the extraction means as learning information in the first learning dictionary; and the learning information extracted by the registered information extraction means, A frequency changing unit that changes the frequency stored in the first learning dictionary according to the learning information to be registered by the learning registration unit, and a word that is a target of the translation process stored in the first learning dictionary. A learning dictionary reference means for extracting corresponding learning information together with the corresponding frequency, and a second order for exchanging candidate ranks of the translation information stored in the result table according to the frequency of the learning information extracted by the learning dictionary reference means. A translation processing device, comprising: result replacement means. 6. A translation that is a translation candidate having a candidate rank for each word, obtained by a translation process for translating a sentence in a first language into a sentence in a second language, and a translation including a word and a part of speech corresponding to the translation. A result table for storing information, a translation of the translation information to be learned from the result table, and a word corresponding to the translation, in response to a learning execution instruction for the translation information stored in the result table. Registration information extracting means for extracting a part of speech, learning information referred to for determining the order of candidates obtained by the translation process, a first learning dictionary for storing the learning frequency of each learning information, and the registration information By using the translation information extracted by the extraction means as learning information, the learning registration means for registering in the first learning dictionary according to the learning order, and the registered information extraction means. Frequency changing means for changing the frequency stored in the first learning dictionary according to the retrieved learning information to be registered by the learning registration means, and a translation process stored in the first learning dictionary. Depending on the frequency of the learning information corresponding to the target word, together with the corresponding frequency, the learning dictionary reference means for extracting the learned order so that the learned order can be identified, and the frequency of the learning information extracted by the learning dictionary reference means A second result replacement unit that replaces the candidate ranks of the translation information stored in the result table, and the first learning information stored in the result table, which is the latest learning information extracted by the learning dictionary reference unit. And a third result interchange means for exchanging according to the above. 7. The dictionary frequency changing means for lowering the frequencies corresponding to all the learning information stored in the first dictionary according to a specific rule so as to have a low frequency difference. The translation processing device according to claim 5 or claim 6. 8. A switching means for switching a mode indicating a strength of reflecting a learning effect on a translation result, wherein the learning dictionary reference means extracts learning information according to a mode switched by the switching means, and translates the learning information. 7. The translation processing apparatus according to claim 1, wherein the translation candidate order is reflected in the replacement.