JP2006163768A - User dictionary generating system, user dictionary generating device, and user dictionary generating method and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To allow a user to easily select question words effective for improving a translation result among the question words as a registration candidate to a user dictionary presented by a user dictionary generating device. <P>SOLUTION: A question portion analysis means 102 extracts an inconsistent part between a mechanical translation with respect to the original text and a reference translation, and extracts the words in the original text corresponding to the inconsistent part as the question words, and a question word appearance frequency analysis means 103 calculates the appearance frequency of the respective question words in the original text. A registration effect analysis means 104 calculates a degree of influence on the translation result based on the appearance frequency and a part of speech for each of the question words. A registration effect display specifying means 105 transmits a registration effect display screen for displaying the inconsistent part and the question words and the degree of influence by associating them to a dictionary user terminal. The user selects the question words as the registration candidate among the question words displayed on the registration effect display screen while considering the degree of influence. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a machine translation technique for translating a sentence in a first language into a sentence in a second language, and more particularly to a technique for supporting creation of a user dictionary.

  In recent years, machine translation systems for translating sentences in a first language into sentences in a second language using a computer have been actively used. In general, a machine translation system uses a machine translation dictionary built in the system, which is composed of a word dictionary, a compound word dictionary, a idiom dictionary, a syntax dictionary, a pattern dictionary that describes vocabulary relationships, and the like. We are translating. However, there are cases in which texts in specialized fields cannot be correctly translated using only the machine translation dictionary originally provided in the system. For this reason, a user dictionary in which dictionary data relating to words / phrases (including words, phrases, etc.) peculiar to specialized fields handled by the user is registered. The user dictionary is preferentially used over the machine translation dictionary during translation.

  By the way, when registering dictionary data in a user dictionary, conventionally, dictionary data is generally created manually based on a bilingual list owned by the user, and the created dictionary data is used as a registration function of a machine translation system. It is used to register in the user dictionary. However, this work has a problem that it is very time-consuming.

Therefore, in order to solve such problems, conventionally, the original text and the correct translation of the original text are input to extract the corresponding parts of the two, and dictionary data is created based on the extracted corresponding parts. A dictionary creation device for registering the dictionary data in the translation dictionary has been proposed (see, for example, Patent Documents 1 and 2).
JP 2002-366546 A JP-A-7-85056

  According to the conventional dictionary creation apparatus described above, it is possible to register dictionary data in the translation dictionary simply by inputting the original sentence and the correct translation sentence for the original sentence. However, when a user dictionary is created using a conventional dictionary creation device, the following problems arise. That is, since the conventional dictionary creation device automatically creates dictionary data based on the input original sentence and translated sentence, it is the same as the dictionary data registered in the machine translation dictionary originally provided in the system. There is a problem that the dictionary data may be registered in the user dictionary. When such dictionary data is registered in the user dictionary, storage resources are wasted. Furthermore, if the user has requested the user dictionary creation company to create a user dictionary and the dictionary creation fee is in accordance with the number of registered dictionary data, the registration cost of unnecessary dictionary data is also included. You have to pay for the included dictionary.

  In order to solve such problems, for example, there is a method in which dictionary data created based on the original sentence and the translated sentence is presented to the user, and the user selects dictionary data to be registered in the user dictionary. Conceivable. By doing so, the problem that dictionary data unnecessary for the user is registered in the user dictionary can be solved. However, when selecting dictionary data to be registered in the user dictionary from the presented dictionary data, it is checked whether or not the dictionary data is registered in the machine translation dictionary for each presented dictionary data. When the dictionary data is registered in the user dictionary, it is necessary to consider how much the translation result is improved, and this work is a heavy burden on the user.

(Object of invention)
Therefore, an object of the present invention is to register a problem word / phrase that becomes a problem when translation is performed using a machine translation dictionary and dictionary data corresponding to the problem word / phrase in a user dictionary, thereby obtaining a translation result. By presenting to the user information indicating how much improvement is to be made, the burden on the user when selecting dictionary data to be registered in the user dictionary is reduced.

The first user dictionary creation system according to the present invention is:
A user dictionary creation system including a user dictionary creation device and a user terminal,
The user dictionary creation device is
Machine translation means for translating the original text using a machine translation dictionary;
Extracting a mismatched portion between a machine translation sentence when the original sentence is translated by the machine translation means and a reference translation sentence for the original sentence, and extracting a word / phrase in the original sentence corresponding to the extracted mismatched part as a question word / phrase Problem location analysis means,
A problem word appearance frequency analyzing means for obtaining an appearance frequency in the original sentence of the problem word extracted by the problem part analyzing means;
A registration effect analyzing means for calculating an influence degree of the problem word / phrase on the translation result, which is a value related to the appearance frequency of the problem word / phrase obtained by the problem word / phrase appearance frequency analyzing means;
A registration effect display for displaying a mismatch effect part extracted by the problem location analysis means and a question word in association with each other and transmitting a registration effect display screen displaying the degree of influence calculated by the registration effect analysis means to the user terminal A designation means,
The user terminal is
Registration candidate selection means for transmitting registration candidate selection information indicating a problem word / phrase selected by a user among the question words / phrases displayed on the registration effect display screen to the user dictionary creating device is provided. And

A second user dictionary creation system according to the present invention is the first user dictionary creation system,
The registration effect analysis means
The degree of influence is calculated based on the appearance frequency of the problem word / phrase obtained by the problem word / phrase appearance frequency analyzing means and the part of speech of the problem word / phrase.

A third user dictionary creation system according to the present invention is the first user dictionary creation system,
The registration effect analysis means
The degree of influence is calculated on the basis of the appearance frequency of the problem word / phrase obtained by the problem word / phrase appearance frequency analysis means, the part of speech of the problem word / phrase, and the factor that caused the mismatched part corresponding to the problem word / phrase. It is characterized by.

The 4th user dictionary creation system concerning the present invention is the 1st-3rd user dictionary creation system,
The user dictionary creation device is
For each problem word specified by the registration candidate selection information, a registration candidate cost display screen for calculating a registration cost according to the part of speech of the problem word and displaying the calculated registration cost and the corresponding problem word Registration candidate / cost confirmation means to be transmitted to the user terminal,
A bilingual dictionary list creating means for creating a bilingual dictionary list including a problem word specified by registration target selection information and a word and phrase corresponding to the problem word in the reference translation sentence;
Dictionary registration means for registering the bilingual dictionary list created by the bilingual dictionary list creating means in a user dictionary;
The user terminal is
Registration object selection means for transmitting registration object selection information for specifying a problem word / phrase selected by the user among the problem words / phrases displayed on the registration candidate cost display screen to the user dictionary creation device is provided. Features.

A fifth user dictionary creation system according to the present invention is the fourth user dictionary creation system,
Instead of the registration candidate / cost determination means,
For each problem word specified by the registration candidate selection information, calculate the registration cost according to the part of speech of the problem word and the factor that caused the mismatched part corresponding to the problem word, and the calculated registration cost A registration candidate / cost determination means for transmitting a registration candidate cost display screen for displaying a corresponding problem word / phrase to the user terminal is provided.

A sixth user dictionary creation system according to the present invention is the fourth and fifth user dictionary creation system,
The apparatus further comprises cost billing means for creating a bill including the problem word / phrase registered in the user dictionary and the registration cost for the problem word / phrase calculated by the registration candidate / cost determination means.

The seventh user dictionary creation system according to the present invention is the first to sixth user dictionary creation systems,
The reference translated sentence is a translated sentence instructed by the user with respect to the original sentence.

The 8th user dictionary creation system concerning the present invention is the 1st-6th user dictionary creation system,
The reference translation sentence is a translation sentence created using a translation memory.

The first user dictionary creation device according to the present invention is:
Machine translation means for translating the original text using a machine translation dictionary;
Extracting a mismatched portion between a machine translation sentence when the original sentence is translated by the machine translation means and a reference translation sentence for the original sentence, and extracting a word / phrase in the original sentence corresponding to the extracted mismatched part as a question word / phrase Problem location analysis means,
A problem word appearance frequency analyzing means for obtaining an appearance frequency in the original sentence of the problem word extracted by the problem part analyzing means;
A registration effect analyzing means for calculating an influence degree of the problem word / phrase on the translation result, which is a value related to the appearance frequency of the problem word / phrase obtained by the problem word / phrase appearance frequency analyzing means;
A registration effect display designation for displaying a mismatch effect part extracted by the problem location analysis means and a question word in association with each other and transmitting a registration effect display screen displaying the degree of influence calculated by the registration effect analysis means to the user terminal Means.

A second user dictionary creation device according to the present invention is the first user dictionary creation device,
The registration effect analysis means
The degree of influence is calculated based on the appearance frequency of the problem word / phrase obtained by the problem word / phrase appearance frequency analyzing means and the part of speech of the problem word / phrase.

A third user dictionary creation device according to the present invention is the first user dictionary creation device,
The registration effect analysis means
The degree of influence is calculated on the basis of the appearance frequency of the problem word / phrase obtained by the problem word / phrase appearance frequency analysis means, the part of speech of the problem word / phrase, and the factor that caused the mismatched part corresponding to the problem word / phrase. It is characterized by.

The 4th user dictionary creation device concerning the present invention is the 1st-3rd user dictionary creation device,
When the registration candidate selection information indicating the problem word / phrase selected by the user among the question words / phrases displayed on the registration effect display screen is sent from the user terminal, the registration candidate selection information For each problem word specified, a registration cost corresponding to the part of speech of the problem word is calculated, and a registration candidate cost display screen for displaying the calculated registration cost and the corresponding problem word is transmitted to the user terminal. Registration candidate / cost determination means,
When registration object selection information for specifying a problem word / phrase selected by the user among the problem words / phrases displayed on the registration candidate cost display screen is sent from the user terminal, the registration object selection information A bilingual dictionary list creating means for creating a bilingual dictionary list including a specified problem word and the word corresponding to the problem word in the reference translation sentence;
And a dictionary registration unit for registering the bilingual dictionary list created by the bilingual dictionary list creating unit in a user dictionary.

A fifth user dictionary creation device according to the present invention is the fourth user dictionary creation device,
Instead of the registration candidate / cost determination means,
For each problem word specified by the registration candidate selection information, calculate the registration cost according to the part of speech of the problem word and the factor that caused the mismatched part corresponding to the problem word, and the calculated registration cost A registration candidate / cost determination means for transmitting a registration candidate cost display screen for displaying a corresponding problem word / phrase to the user terminal is provided.

A sixth user dictionary creation device according to the present invention is the fourth and fifth user dictionary creation device,
The apparatus further comprises cost billing means for creating a bill including the problem word / phrase registered in the user dictionary and the registration cost for the problem word / phrase calculated by the registration candidate / cost determination means.

The seventh user dictionary creation device according to the present invention is the first to sixth user dictionary creation devices,
The reference translated sentence is a translated sentence instructed by the user with respect to the original sentence.

An eighth user dictionary creation device according to the present invention is the first to sixth user dictionary creation devices,
The reference translation sentence is a translation sentence created using a translation memory.

A first user dictionary creation method according to the present invention includes:
User dictionary creation device
A machine translation step for translating the original text using a machine translation dictionary;
Extracting a mismatched portion between the machine translation sentence when the original sentence was translated in the machine translation step and a reference translation sentence for the original sentence, and extracting a phrase in the original sentence corresponding to the extracted mismatched part as a problem phrase Problem location analysis step,
A problem word appearance frequency analysis step for obtaining an appearance frequency in the original sentence of the problem word extracted in the problem part analysis step;
A registration effect analysis step for calculating an influence degree of the problem word / phrase on the translation result, which is a value related to the appearance frequency of the problem word / phrase obtained in the problem word / phrase appearance frequency analysis step;
A registration effect display for displaying a mismatch effect part extracted in the problem location analysis step and a question word in association with each other and transmitting a registration effect display screen displaying the degree of influence calculated in the registration effect analysis step to the user terminal Execute the specified step,
The user terminal is
A registration candidate selection step of transmitting registration candidate selection information indicating a problem word / phrase selected by a user among the question words / phrases displayed on the registration effect display screen to the user dictionary creation device is executed. And

A second user dictionary creation method according to the present invention is the first user dictionary creation method,
The registration effect analysis step comprises:
The degree of influence is calculated based on the appearance frequency of the question word and the part of speech of the question word obtained in the question word appearance frequency analyzing step.

A third user dictionary creation method according to the present invention is the first user dictionary creation method,
The registration effect analysis step comprises:
The degree of influence is calculated based on the appearance frequency of the question word / phrase obtained in the question word / phrase appearance frequency analysis step, the part of speech of the question word / phrase, and the factor that caused the mismatched part corresponding to the question word / phrase. To do.

A fourth user dictionary creation method according to the present invention is the first to third user dictionary creation methods,
The user dictionary creation device is
For each problem word specified by the registration candidate selection information, a registration candidate cost display screen for calculating a registration cost according to the part of speech of the problem word and displaying the calculated registration cost and the corresponding problem word Registration candidate / cost confirmation step to be transmitted to the user terminal,
A bilingual dictionary list creating step of creating a bilingual dictionary list including a problem word specified by registration target selection information and a word and phrase corresponding to the problem word in the reference translation sentence;
A dictionary registration step of registering the bilingual dictionary list created in the bilingual dictionary list creating step in a user dictionary;
The user terminal is
Executing a registration object selection step of transmitting registration object selection information for specifying a problem word selected by the user among the problem words displayed on the registration candidate cost display screen to the user dictionary creation device; Features.

A fifth user dictionary creation method according to the present invention is the fourth user dictionary creation method,
Instead of the registration candidate / cost determination step,
For each problem word specified by the registration candidate selection information, calculate the registration cost according to the part of speech of the problem word and the factor that caused the mismatched part corresponding to the problem word, and the calculated registration cost A registration candidate / cost determination step for transmitting a registration candidate cost display screen for displaying a corresponding problem word to the user terminal is executed.

A sixth user dictionary creation method according to the present invention is the fourth or fifth user dictionary creation method,
The user dictionary creation device is
A cost billing step is performed for creating a bill including a problem word registered in the user dictionary and a registration cost for the problem word calculated by the registration candidate / cost determination unit.

The seventh user dictionary creation method according to the present invention is the first to sixth user dictionary creation methods,
The reference translated sentence is a translated sentence instructed by the user with respect to the original sentence.

An eighth user dictionary creation method according to the present invention is the first to sixth user dictionary creation methods,
The reference translation sentence is a translation sentence created using a translation memory.

The first program according to the present invention is:
Computer
Machine translation means for translating the original text using a machine translation dictionary,
Extracting a mismatched portion between a machine translation sentence when the original sentence is translated by the machine translation means and a reference translation sentence for the original sentence, and extracting a word / phrase in the original sentence corresponding to the extracted mismatched part as a question word / phrase Problem location analysis means,
A problem word appearance frequency analysis means for obtaining an appearance frequency in the original sentence of the problem word extracted by the problem part analysis means;
Registration effect analysis means for calculating the degree of influence of the problem word / phrase on the translation result, which is a value related to the appearance frequency of the problem word / phrase determined by the problem word / phrase appearance frequency analysis means;
A registration effect display designation for displaying a mismatch effect part extracted by the problem location analysis means and a question word in association with each other and transmitting a registration effect display screen displaying the degree of influence calculated by the registration effect analysis means to the user terminal It functions as a means.

The second program according to the present invention is the first program,
The registration effect analysis means
The degree of influence is calculated based on the appearance frequency of the problem word and the part of speech of the question word obtained by the problem word appearance frequency analysis means.

A third program according to the present invention is the first program,
The registration effect analysis means
The degree of influence is calculated based on the appearance frequency of the problem word / phrase obtained by the problem word / phrase appearance frequency analysis means, the part of speech of the problem word / phrase, and the factor that caused the mismatched part corresponding to the problem word / phrase. To do.

The fourth program according to the present invention is the first to third programs,
The computer,
When the registration candidate selection information indicating the problem word / phrase selected by the user among the question words / phrases displayed on the registration effect display screen is sent from the user terminal, the registration candidate selection information For each problem word specified, a registration cost corresponding to the part of speech of the problem word is calculated, and a registration candidate cost display screen for displaying the calculated registration cost and the corresponding problem word is transmitted to the user terminal. Registration candidate / cost determination means,
When registration object selection information for specifying a problem word / phrase selected by the user among the problem words / phrases displayed on the registration candidate cost display screen is sent from the user terminal, the registration object selection information A bilingual dictionary list creating means for creating a bilingual dictionary list including the specified problem word and the word corresponding to the problem word in the reference translation sentence;
The bilingual dictionary list created by the bilingual dictionary list creating means functions as a dictionary registering means for registering in the user dictionary.

A fifth program according to the present invention is the fourth program,
The registration candidate / cost determination means is
For each problem word specified by the registration candidate selection information, calculate the registration cost according to the part of speech of the problem word and the factor that caused the mismatched part corresponding to the problem word, and the calculated registration cost It has the structure which transmits the registration candidate expense display screen which displays a corresponding problem word to the said user terminal.

The sixth program according to the present invention is the fourth and fifth programs,
The computer,
It is made to function as a cost billing means for creating a bill including the problem word / phrase registered in the user dictionary and the registration cost for the problem word / phrase calculated by the registration candidate / cost determination means.

[Action]
When the original text to be translated is specified by the user, the machine translation means in the user dictionary creation device generates a machine translation sentence for the original text using the machine translation dictionary, and the problem location analysis means Extracting inconsistencies between the machine translation and the reference translation for the original (for example, an accurate translation specified by the user or a nearly accurate translation generated using a translation memory, etc.) Then, the phrase in the original text corresponding to the extracted inconsistent part is extracted as the question phrase. Thereafter, the problem word appearance frequency analysis means obtains the appearance frequency of each problem word in the original text, and the registration effect analysis means has a value related to the appearance frequency for each problem word phrase, and the problem word phrase is a translation result. The influence degree which shows the degree which influences is calculated. Here, since a phrase having a high appearance frequency is a phrase that is a keyword, the influence level, which is a value related to the appearance frequency, is a guideline for the user to select a phrase to be registered in the user dictionary. The registered effect display designation means uses a registration effect display screen that displays the inconsistent portion extracted by the problem location analysis means and the problem word in association with each other and displays the degree of influence of each problem word calculated by the registration effect analysis means. To the user terminal.

  When the registration effect display screen is sent from the user dictionary creation device, the user of the user terminal refers to the degree of influence displayed on the screen and refers to the problem words that are candidates for registration in the user dictionary. Select. Thereby, the registration candidate selection means in the user terminal sends the registration candidate selection information indicating the problem word / phrase selected by the user among the problem words / phrases displayed on the registration effect display screen to the user dictionary creation device. Send.

  The registration candidate / cost determination means in the user dictionary creation device calculates a registration cost corresponding to the part of speech of the question word for each question word specified by the registration candidate selection information from the dictionary user terminal. A registration candidate cost display screen for displaying the registered cost and the corresponding problem word is transmitted to the user terminal.

  When the user terminal user receives the registration candidate cost display screen from the user dictionary creation device, the registration cost displayed on the screen is taken into consideration and the problem of making it a target for registration in the user dictionary. Select a phrase. Thereby, the registration object selection means in the user dictionary transmits registration object selection information indicating the problem word / phrase selected by the user to the user dictionary creating apparatus.

  The bilingual dictionary list creating means in the user dictionary creating device creates a bilingual dictionary list including the problem word / phrase indicated by the registration target selection information and the word / phrase corresponding to the problem word / phrase in the reference translation sentence. The contents of the bilingual dictionary list are registered in the user dictionary.

  According to the present invention, the user can easily select a problem word / phrase effective for improving the translation result from the problem words / phrases to be registered in the user dictionary presented by the user dictionary creation device. It becomes possible. The reason for this is the mismatch between the machine translation and the reference translation for the original text, the problem word, and the degree of influence (value related to the frequency of occurrence of the problem word in the original text, indicating the degree of influence on the translation result). This is because a registration effect display screen for displaying the information in association with each other is transmitted to the user terminal. That is, the user recognizes that correct translation results cannot be obtained with dictionary data registered in the machine translation dictionary based on the mismatched portion between the machine translation sentence and the reference translation sentence, and translates based on the degree of influence. Since the influence on the result can be recognized, it becomes possible to easily select a problem word / phrase to be registered in the user dictionary.

  Next, embodiments of the present invention will be described in detail with reference to the drawings.

  Referring to FIG. 1, a first embodiment of a user dictionary creation system according to the present invention creates a user dictionary for a user dictionary creation device 1 and an operator of the user dictionary creation device 1. It consists of the dictionary user terminal 2 of the requesting user (dictionary user) and the dictionary creator terminal 3 of the operator of the user dictionary creation device 1, which are connected via a network 4 such as the Internet. Are connected to each other.

  The user dictionary creation device 1 is realized by a computer or the like, and has, for example, a configuration as shown in the block diagram of FIG. Referring to FIG. 2, the user dictionary creation device 1 includes a machine translation unit 101, a problem location analysis unit 102, a problem word appearance frequency analysis unit 103, a registration effect analysis unit 104, and a registration effect display designation unit 105. Problem difficulty level analysis means 106, registration candidate / cost determination means 107, bilingual dictionary list creation means 108, dictionary registration means 109, cost billing means 110, transmission / reception means 111, machine translation dictionary 112, original text User document storage unit 113 in which database (DB) 114, translation DB 115 and parallel translation DB 116 are stored, problem DB 117, translation change degree DB 118, dictionary difficulty DB 119, unit price DB 120, billing information DB 121, and creation target User dictionary 122, dictionary user DB 123, input unit 124 such as a keyboard, LCD, etc. And a display unit 125.. In FIG. 2, only one user document storage unit 113 is shown, but when there are a plurality of users, a user document storage unit 113 is prepared for each user. Further, the user document storage unit 113 is not necessarily arranged on the user dictionary creating apparatus 1 but may be arranged on another computer accessible from the user dictionary creating apparatus 1 via a network.

  The machine translation dictionary 112 includes various dictionaries necessary for the machine translation unit 101 to perform machine translation, such as a word dictionary, a compound word dictionary, a idiom dictionary, a syntax dictionary, and a pattern dictionary and a rule dictionary that describe vocabulary relationships. include. Further, the machine translation dictionary 112 includes a translation memory.

  The original sentence DB 114 stores the original sentence to be translated designated by the user of the dictionary user terminal 2. The translated sentence DB 115 stores the translated sentence designated by the user. The parallel translation DB 116 stores an original sentence designated by the user and a correct translation sentence for the original sentence. In addition, the pair of the original sentence and the translation sentence stored in the parallel translation DB 116 is created based on the parallel translation example sentence of a commercially available dictionary, or based on the original sentence and the translation sentence collected from the Japanese site and the English translation site on the network. can do.

  The machine translation means 101 translates the original text designated by the dictionary user terminal 2 out of the original text registered in the original text DB 114 and the parallel translation DB 116 using the machine translation dictionary 112, and the translation result (machine translation). Sentence) to the problem part analysis means 102. If the machine translation means 101 has two types of translation functions such as English-Japanese machine translation and Japanese-English machine translation, the original sentence DB 114 may be a translated sentence DB and the translated sentence DB 115 may be an original sentence DB. is there.

  The problem location analysis means 102 extracts a mismatched portion between the machine translation sentence for the original sentence passed from the machine translation means 101 and the reference translation sentence for the original sentence, and also a phrase (word, word) in the original sentence corresponding to the extracted mismatched portion. Etc.) as a problem word. For example, if the original sentence is “The power turns off.”, The machine translation sentence is “Power is turned off.”, And the reference translation sentence is “Power is turned off.”, “Power” and “Power supply” are extracted as mismatched parts. Then, “power” is extracted as a problem word. When the original sentence designated from the dictionary user terminal 2 is registered in the parallel translation DB 116, the reference translation sentence can be a correct translation sentence registered in association with the original sentence. . Further, when the original text specified from the dictionary user terminal 2 is registered in the original text DB 114, it can be a translation text generated using a translation memory, a pattern dictionary, or the like. For example, when creating a reference translation using a translation memory, select the source text registered in the translation memory that has the highest degree of agreement with the source text specified by the user. A translated sentence registered in association with the original sentence is used as a reference translated sentence.

  Further, the problem location analysis means 102 issues, for each extracted problem word, a data number unique to the problem word, a sentence in the original sentence including the problem word, a machine translation sentence, a reference translation sentence, a problem part, and a part of speech. It has a function of registering in the DB 117. FIG. 3 shows an example of the contents of the problem DB 117. Note that the items “appearance frequency” and “influence” in FIG. 3 are set by the problem word appearance frequency analysis unit 103 and the registered effect analysis unit 104, respectively.

  The problem word appearance frequency analysis means 103 has a function for obtaining the appearance frequency of the question word phrase in the original sentence for each question word phrase extracted by the problem location analysis means 102 and a function for registering the obtained appearance frequency in the problem DB 117.

  In the translation change degree DB 118, translation change degrees are registered in association with parts of speech. The translation change degree is a weighting coefficient used when calculating the influence degree on the translation result described later, and an example thereof is shown in FIG. In the example of the figure, the translational change degrees of nouns, verbs, auxiliary verbs, verb phrases, verb idioms,... Are 1, 3, 5, 8, 10,. The translation change is determined empirically by the dictionary creator.

  The registered effect analysis unit 104, for each problem word extracted by the problem location analysis unit 102, the appearance frequency of the problem word obtained by the problem word appearance frequency analysis unit 103, the part of speech of the problem word, and the translation change degree DB 118 Based on the content (translation change degree), it has a function of calculating the degree of influence of the question word on the translation result and a function of registering the calculated degree of influence in the problem DB 117. In the present embodiment, the influence degree is calculated based on the following equation (1).

Influence = (Frequency of problem phrase) × (Translation change according to part of speech of problem phrase) (1)
Note that the translation change degree corresponding to the part of speech of the problem phrase is obtained by referring to the translation change degree DB 118. Now, for example, if the appearance frequency of a certain problem phrase “power” is “125” and the contents of the translation change degree DB 118 are as shown in FIG. = 125 × (noun translation change) = 125 × 1 = 125. Also, if the appearance frequency of the question word “could” is “22”, the degree of influence on the question word “could” is influence = 22 × (translation change degree of auxiliary verb) = 22 × 5 = 110.

  The registered effect display designating means 105 is a registered effect display screen in which the inconsistent part and the question word extracted by the problem location analyzing means 102 and the influence degree calculated by the registered effect analyzing means 104 are arranged in association with each other. It has a function of creating a registration effect display screen that allows the user of the dictionary user terminal 2 to select a problem word / phrase to be registered in the dictionary 122 and transmitting it to the dictionary user terminal 2. Furthermore, the registration effect display designating unit 105 has a function of updating the problem DB 117 according to the problem word / phrase selected by the user of the dictionary user terminal 2 using the registration effect display screen.

  FIG. 5 shows an example of the registration effect display screen 50. Referring to the figure, the registration effect display screen 50 includes a data number part 51, an original sentence part 52, a machine translation sentence part 53, a reference translation sentence part 54, a problem part 55, a problem word phrase part 56, and a part of speech part. 57, an influence degree unit 58, and a selection unit 59.

  In the data number portion 51, a data number unique to the question word is set. An original sentence is set in the original sentence part 52, and a machine translation sentence and a reference translation sentence for the original sentence are set in the machine translation sentence part 53 and the reference translation sentence part 54, respectively. In addition, problem words and phrases in the original text set in the original text part 52 are enclosed in parentheses, and unmatched parts in the machine translation text and reference translation text set in the machine translation text part 53 and the reference translation text part 54 are also enclosed in parentheses. , They are visually distinguishable from other parts. In the problem part 55, information indicating the word number of the original sentence is set. In the question word / phrase part 56, a question word / phrase is set. In the part of speech part 57, the part of speech (noun, verb, auxiliary verb, verb phrase, verb idiom, etc.) is set. The influence degree 58 is set with the influence degree calculated by the registered effect analysis means 104. The selection unit 59 is a selection column used when the user selects a question word / phrase to be a candidate for registration in the user dictionary 122. In the example of FIG. 5, the problem word and the inconsistent part are enclosed in parentheses so that they can be visually distinguished from other parts. Other methods such as making the part different from the part may be adopted.

  In the dictionary difficulty DB 119, the difficulty level when registering the problem word / phrase of the part of speech in the user dictionary 122 is registered in association with the part of speech. FIG. 6 shows an example of the dictionary difficulty DB 119. In the example of the figure, the difficulty levels of nouns, verbs, auxiliary verbs, verb phrases, verb idioms,... Are 1, 3, 5, 8, 10,. Note that the difficulty level of each part of speech can be, for example, a value empirically obtained by the dictionary creator.

  The problem difficulty level analysis unit 106 has a function of acquiring the difficulty level of each question word / phrase registered in the problem DB 117 from the dictionary difficulty level DB 119 and passing it to the registration candidate / cost determination unit 107.

  A unit price for one point of difficulty is registered in the unit price DB 120. FIG. 7 shows an example of the contents of the unit price DB 120. In the example of the figure, the unit price for one point of difficulty is “100 yen”.

  The registration candidate / cost determination unit 107 registers each problem word in the user dictionary 122 based on the difficulty level of each problem word passed from the problem difficulty analysis unit 106 and the unit price registered in the unit price DB 120. It has a function to calculate the registration fee required for Further, the registration candidate / cost determination unit 107 is a registration candidate cost display screen in which the problem word and the registration cost of the word are associated with each other, and the problem word / phrase to be registered in the user dictionary 122 is displayed as a user. A registration candidate cost display screen that can be selected and transmitted to the dictionary user terminal 2. Further, the registration candidate / cost determination unit 107 has a function for creating a list of problem words / phrases (registration target list) selected by the user of the dictionary user terminal 2 using the registration candidate cost display screen, or a user's selection. It has a function of recalculating the registration cost (total amount) when registering the question word / phrase in the user dictionary 122.

  FIG. 8 is a diagram showing an example of the registration candidate cost display screen 80. The data number part 81, the problem word / phrase part 82, the appearance frequency part 83, the part of speech part 84, the difficulty part 85, and the phrase registration cost are shown. A part 86 and a selection part 87 are included. In the data number portion 81, the data number of each question word (each question word / phrase on the registration candidate cost display screen 80, each question word / phrase on the registration effect display screen 50, and each question DB 117 registered by this data number). Is associated with the question word). In the question word / phrase part 82, a question word / phrase to be registered as a candidate for registration in the user dictionary 122 selected by the user using the registration effect display screen 50 is set. In the appearance frequency section 83, the appearance frequency obtained by the problem word / phrase appearance frequency analysis means 103 is set. In the part of speech part 84, the part of speech of the question phrase is set. The difficulty level portion 85 is set with a difficulty level when registering a question word / phrase in the user dictionary 122. In the word / phrase registration cost section 86, a registration fee and a total amount for each question word / phrase required when registering the question / phrase in the user dictionary 122 are set. The selection unit 87 is a selection field used when the user selects a problem word (registration target question word) to be registered in the user dictionary 122.

  The bilingual dictionary list creating means 108 uses the dictionary creator terminal 3 to create a bilingual dictionary list including the registration target question words selected by the user and the contents registered in the user dictionary 122. The bilingual dictionary list thus created is transmitted to the dictionary user terminal 2 so that the user can confirm the contents registered in the user dictionary 122 in advance.

  FIG. 9 is a diagram showing an example of the bilingual dictionary list 90 created by the bilingual dictionary list creating means 108. In the example in the figure, the phrase (translation word such as removal) existing in the mismatched part in the reference translation sentence and the possible syntax and pattern (SVO etc.) are set in association with the problem word (remove etc.). However, other information (for example, part of speech or grammatical information) may be set. Although not shown in FIG. 9, a data number is also set in association with each question word.

  The dictionary registration unit 109 has a function of registering the contents of the bilingual dictionary list 90 in the user dictionary 122.

  In the dictionary user DB 123, user information (for example, name, address, company name, mail address, account number, etc.) of the user of the dictionary user terminal 2 is registered.

  The cost billing unit 110 calculates the billing amount based on the contents of the billing information DB 120 and the contents registered in the user dictionary 122 and the billing amount to the mail address registered in the dictionary user DB 123. It has a function of transmitting a bill including it. FIG. 10 shows an example of an invoice sent to the user.

  The transmission / reception means 111 has a function of transmitting / receiving data via the network 4.

  When the user dictionary creation device 1 is realized by a computer, for example, a disk, a semiconductor memory, or other recording medium that records a program for causing the computer to function as the user dictionary creation device 1 is prepared. Have the computer read the program. The computer controls its own operation according to the read program, so that the machine translation means 101, the problem part analysis means 102, the problem word appearance frequency analysis means 103, the registration effect analysis means 104, and the registration effect display designation are performed on the computer. Means 105, problem difficulty level analysis means 106, registration candidate / cost determination means 107, bilingual dictionary list creation means 108, dictionary registration means 109, cost billing means 110, and transmission / reception means 111 are realized.

  FIG. 11 is a block diagram showing a configuration example of the dictionary user terminal 2. The transmission / reception unit 21, the registration candidate selection unit 22, the registration target selection unit 23, the registration content confirmation unit 24, and an input unit 25 such as a keyboard. And a display unit 26 such as an LCD.

  The transmission / reception means 21 has a function of transmitting / receiving data via the network 4.

  The registration candidate selection means 22 is a function for displaying the registration effect display screen 50 sent from the user dictionary creation device 1 on the display unit 26, or the use of the problem words displayed on the registration effect display screen 50. The registration candidate selection information for specifying the problem word / phrase selected by the user is transmitted to the user dictionary creating apparatus 1.

  The registration target selection unit 23 has a function of displaying the registration candidate cost display screen 80 sent from the user dictionary creation device 1 on the display unit 26, or a problem word / phrase displayed on the registration candidate cost display screen 80. The registration target selection information for specifying the problem word / phrase selected by the user is transmitted to the user dictionary creating apparatus 1.

  The registered content confirmation unit 24 has a function of displaying the bilingual dictionary list sent from the user dictionary creation device 1 on the display unit 26. Further, in accordance with the deletion instruction input by the user from the input unit 25, the information regarding the problem word / phrase instructed to be deleted by the user is deleted from the bilingual dictionary list, or edited according to the transmission instruction input by the user from the input unit 25. The bilingual dictionary list (if the user does not input a deletion instruction, the bilingual dictionary list sent from the bilingual dictionary list creating means 108) is transmitted to the user dictionary creating apparatus 1.

  The dictionary user terminal 2 can be realized by a computer. When the dictionary user terminal 2 is realized by a computer, for example, the following operation is performed. A disk, a semiconductor memory, and other recording media recording a program for causing the computer to function as the dictionary user terminal 2 are prepared, and the program is read by the computer. The computer controls the operation of the computer according to the read program, thereby realizing the transmission / reception unit 21, the registration candidate selection unit 22, the registration target selection unit 23, and the registered content confirmation unit 24 on the computer.

  FIG. 12 is a block diagram showing a configuration example of the dictionary creator terminal 3, which includes a transmission / reception unit 31, a bilingual dictionary list creation support unit 32, an input unit 33 such as a keyboard, and a display unit 34 such as an LCD. ing.

  The transmission / reception means 31 has a function of transmitting / receiving data via the network 4.

  The bilingual dictionary list creation support means 32 has a function to display information (such as information stored in the problem DB 117) sent from the bilingual dictionary list creation means 108 in the user dictionary creation device 1 on the display unit 34, The dictionary creator has a function of transmitting the bilingual dictionary list created by using the input unit 33 to the bilingual dictionary list creating means 108.

  The dictionary creator terminal 3 can be realized by a computer. When the dictionary creator terminal 3 is realized by a computer, for example, the following operation is performed. A disk, a semiconductor memory, and other recording media recording a program for causing the computer to function as the dictionary creator terminal 3 are prepared, and the computer reads the program. The computer controls the operation of the computer according to the read program, thereby realizing the transmission / reception unit 31 and the bilingual dictionary list creation support unit 32 on the computer.

[Operation of First Embodiment]
Next, the operation of the present embodiment will be described in detail.

  When the original text to be translated is designated by the user of the dictionary user terminal 2 or the creator of the dictionary creator terminal 3, the machine translation means 101 reads the original text (text document) designated from the original text DB 114 or the parallel translation DB 116. Acquired (step S1301 in FIG. 13) and generates a machine translation sentence with reference to the machine translation dictionary 112 (step S1302). In the present embodiment, the original text to be translated is selected from the original text stored in the user document storage unit 113. However, the translation target is selected from the dictionary user terminal 2 or the dictionary creator terminal 3. The original text to be transmitted may be transmitted to the machine translation means 101.

  When the machine translation means 101 generates a machine translation sentence for the original sentence, the problem part analysis means 102 acquires a reference translation sentence for the original sentence (step S1303). For example, when the original sentence is stored in the parallel translation DB 116, the translation sentence registered in association with the original sentence is set as the reference translation sentence. If the original sentence is registered in the original sentence DB 114, a translation sentence is created using a translation memory, a pattern dictionary, etc., and is used as a reference translation sentence.

  After that, the problem part analysis unit 102 compares the corresponding part (for example, the corresponding sentence) between the machine translation sentence generated by the machine translation part 101 and the reference translation sentence acquired in step S1303, thereby identifying the inconsistent part. Extract (step S1304). Next, the problem part analysis means 102 extracts a word / phrase in the original text corresponding to the mismatched part as a problem word / phrase, and obtains a question part / part of speech of the question word / phrase indicating the word in the original text. (Step S1305). Thereafter, the original text, machine translation text, reference translation text, problem location, problem word / phrase, and part of speech are associated and registered in the problem DB 117 (step S1306).

The process in steps S1304 to S1306 described above will be described in detail as follows. Now, for example, if the machine translation for a sentence “The power turns off.” In the original text is “Power is turned off” and the reference translation is “Power is turned off.” In step S1304, “Power” and “Power supply” are extracted as inconsistent portions.

  In the next step S1305, the question word “power”, the question part “second word”, and the part of speech “noun” are extracted. Here, as a method of extracting the problem word “power”, for example, a word / phrase corresponding to the mismatched part “power” in the machine translation sentence is extracted from the machine translation dictionary 112, and is further present in the original sentence from the extracted word / phrase. A method of selecting a phrase to be performed can be employed. In addition, when generating a machine translation sentence, the dictionary content used at the time of machine translation is stored as internal data in association with the machine translation sentence, and a phrase corresponding to the mismatched portion “power” is extracted from this internal data, Furthermore, a method of selecting a word / phrase existing in the original sentence from the extracted words / phrases can be adopted.

  In the next step S1306, the original sentence “The power turns off.”, The machine translation sentence “The power turns off.”, The reference translation sentence “The power turns off.”, The problem part “second word”, and the problem phrase “power” ”And the part of speech“ noun ”are registered in the problem DB 117 in association with each other (see FIG. 3). In order to clarify the problem word and the mismatched part, the question word “power”, the mismatched part “power”, and “power” are enclosed in parentheses. The above is the details of the processing performed in steps S1304 to S1306.

  When the problem location analysis unit 102 registers information on each problem word in the problem DB 117 as shown in FIG. 3, the problem word appearance frequency analysis unit 103 selects the translation target registered in the user document storage unit 113. By referring to the original text, the appearance frequency of each question word registered in the problem DB 117 in the original text is obtained, and the obtained appearance frequency is registered in the problem DB 117 (step S1307).

  An example of the process performed in step S1307 will be described in detail as follows. The problem word / phrase appearance frequency analyzing unit 103 pays attention to one entry in the problem DB 117 and checks whether or not the appearance frequency is registered in the entry. For example, assume that attention is paid to the first entry in the problem DB 117 shown in FIG. At present, since the appearance frequency is not registered in the first entry, the appearance frequency of the problem word “power” registered in the first entry is obtained by referring to the user document storage unit 113. For example, if the appearance frequency is “125”, the value is set in the first entry. Thereafter, it is checked whether or not there is an entry in which the same problem word “power” registered in the first entry is registered. If there is an entry, the appearance frequency “125” is also present in the entry. Register. Thereby, since the processing for the first entry is completed, the second entry is noticed and it is checked whether or not the appearance frequency is registered. Since the appearance frequency is not registered at the present time in the second entry in which the question word “remove” is registered, the same processing as in the first entry described above is performed. However, if the problem word “power” is registered in the second entry, the appearance frequency is already registered at the time of processing the first entry. Focus on the entry (third entry). The above processing is repeated until there are no entries whose appearance frequency is not registered. The above is an example of the process performed in step S1307. In this example, the appearance frequency is obtained based only on the character face of the question word, but the appearance frequency may be obtained for each combination of the character face and the part of speech in consideration of the part of speech of the question word phrase. .

  When the problem phrase appearance frequency analysis unit 103 extracts the appearance frequency of each problem word registered in the problem DB 117, the registration effect analysis unit 104 calculates the translation change degree and the problem phrase appearance frequency registered in the translation change degree DB 118. Based on the appearance frequency obtained by the analysis unit 103, the degree of influence on the translation result is calculated for each question word registered in the problem DB 117, and the calculated degree of influence is registered in the problem DB 117 (steps S1308 to S1310). ).

  An example of the processing performed in steps S1308 to S1310 will be described in detail as follows. The registered effect analysis unit 104 pays attention to one entry in the problem DB 117 and checks whether or not the influence degree is registered in the entry. If registered, pay attention to the next entry.

  On the other hand, when the influence degree is registered in the attention entry, first, the translation change degree corresponding to the part of speech registered in the attention entry is obtained from the translation change degree DB 118, and further, the obtained translation change degree is obtained. Is multiplied by the appearance frequency in the entry of interest to calculate the influence degree (step S1309). For example, if the part of speech registered in the entry of interest is a verb idiom and its appearance frequency is “25”, the influence = (translation change corresponding to the verb idiom) × (occurrence frequency) = 10 X25 = 250. When a plurality of parts of speech are stored in the target entry, the present embodiment uses the largest translation change degree corresponding to each part of speech in the target entry. A value obtained by adding the degree of change may be used. Thereafter, the calculated influence degree is registered in the attention entry. At this time, if there is another entry in which the same question word / part of speech as the target entry is registered, the calculated influence degree is also registered in the entry. The above processing is repeated until there are no more unregistered entries. The above is the details of the processing performed in steps S1308 to S1310.

  Thereafter, the registration effect display designating unit 105 creates a registration effect display screen 50 as shown in FIG. 5 based on the contents of the problem DB 117 and transmits it to the dictionary user terminal 2 (step S1311). The registration effect display screen 50 shown in FIG. 5 is merely an example, and the user can confirm the problem word / phrase, the original sentence, the reference translation sentence, the machine translation sentence, the influence degree, and the problem words / phrases to be registered as candidates. Anything can be selected. For example, in addition to the screen shown in FIG. 5, a bar graph as shown in FIG. 15 or a pie chart as shown in FIG. 16 showing the degree of influence of each question word may be displayed.

  The registration candidate selection means 22 in the dictionary user terminal 2 displays the registration effect display screen 50 sent from the user dictionary creation device 1 on the display unit 26. The user selects a problem word / phrase as a candidate for registration in the user dictionary 120 in consideration of the degree of influence of each problem word / phrase displayed on the registration effect display screen 50 (checks the selection unit 59). The registration candidate selection means 22 transmits registration candidate selection information (for example, including the data number of the selected problem word / phrase) indicating the selected problem word / phrase to the user dictionary creation device 1 (step S1312).

  When the registration candidate selection information is sent from the dictionary user terminal 2, the registration effect display designating means 105 in the user dictionary creation device 1 includes information related to each question word / phrase registered in the question DB 117. Information on the question phrase that was not selected by the user is deleted (step S1313). As a result of this processing, only the information related to the problem word / phrase selected by the user remains in the problem DB 117.

  Thereafter, the problem difficulty level analysis means 106 refers to the dictionary difficulty level DB 119 and acquires the difficulty level (difficulty level according to the part of speech) of each question word registered in the problem DB 117 (steps S1401 and S1402 in FIG. 14). ). Thereafter, the registration candidate / cost determination unit 107 uses the difficulty of each problem word acquired by the problem difficulty analysis unit 106 and the unit price registered in the unit price DB 120 for each problem word / phrase. The registration fee for the dictionary 122 is calculated (step S1403). For example, if the contents of the dictionary difficulty level DB 119 and the unit price DB 120 are as shown in FIGS. 6 and 7, respectively, the registration cost of the problem phrase “wear” is registration cost = difficulty × unit price = 3 × 100 = 300. Yen. Thereafter, the registration candidate / cost determination means 107 creates a registration candidate cost display screen 80 as shown in FIG. 8 and transmits it to the dictionary user terminal 2 (step S1404).

  The registration object selection means 23 in the dictionary user terminal 2 displays the registration candidate cost display screen 80 sent from the user dictionary creation device 1 on the display unit 26. The user selects a problem word to be registered in the user dictionary 122 from the problem words displayed on the registration candidate cost display screen 80 in consideration of the registration cost (checks the selection unit 87). ). As a result, the registration target selection unit 23 transmits registration target selection information (for example, including the data number of the selected problem word / phrase) indicating the selected problem word / phrase to the user dictionary creation device 1 (step S1405).

  When registration target selection information is sent from the dictionary user terminal 2, the registration candidate / cost determination unit 107 charges the data number of the problem word / phrase selected by the user in association with the registration cost of the word / phrase. The information is registered in the information DB 121 (step S1406). Thereafter, the registration candidate / cost determination unit 107 creates a list (data list) of the data numbers of the problem words / phrases that are registered by the user and passes them to the bilingual dictionary list creation unit 108 (step S1407).

  When the registration target list is delivered from the registration candidate / cost determination unit 107, the bilingual dictionary list creation unit 108 associates the data numbers included in the registration target list with each data number registered in the problem DB 117. The registered information is acquired, and the acquired information is transmitted to the dictionary creator terminal 3. The bilingual dictionary list creation support means 32 in the dictionary creator terminal 3 displays the information sent from the user dictionary creation device 1 on the display unit 34. The dictionary creator looks at this information and creates a bilingual dictionary list 90 as shown in FIG. 9 in which dictionary registration contents are set for each question word and returns it to the bilingual dictionary list creating means 108. The bilingual dictionary list creating means 108 sends the bilingual dictionary list sent from the dictionary creator terminal 3 to the dictionary user terminal 2 (step S1408). In the present embodiment, the bilingual dictionary list creation means 108 transmits only the information in the problem DB 117 specified by the data number in the registration target list to the dictionary creator terminal 3. Alternatively, information that helps to create a bilingual dictionary list may be extracted from the machine translation dictionary 112 or the bilingual DB 116, and the extracted information may also be transmitted to the dictionary creator terminal 3.

  The registered content confirmation unit 24 in the dictionary user terminal 2 displays the bilingual dictionary list sent from the user dictionary creation device 1 on the display unit 26. When the user looks at the displayed bilingual dictionary list and determines that it is necessary to register information about all problem words in the bilingual dictionary list in the user dictionary 122, the user transmits from the input unit 25. Enter instructions. On the other hand, if it is determined that the bilingual dictionary list includes information on a problem word / phrase that is not required to be registered in the user dictionary 122, a deletion instruction is input from the input unit 25 to obtain the corresponding problem word / phrase. After deleting the information on the bilingual dictionary list, a transmission instruction is input from the input unit 25. When the transmission instruction is input, the registered content confirmation unit 24 transmits the content of the currently displayed bilingual dictionary list to the bilingual dictionary list creating unit 108 (step S1409).

  When the bilingual dictionary list is sent from the dictionary user terminal 2, the bilingual dictionary list creating means 108 passes it to the dictionary registering means 109. Thereby, the dictionary registration means 109 registers the contents of the bilingual dictionary list in the user dictionary 122 (step S1410). After that, the cost billing means 110 is registered in association with the data number of the problem word registered in the user dictionary 122 in step S1410 and each problem word (data number registered in the charging information DB 121 in step S1406). ) And the contents of the dictionary user DB 123, a bill as shown in FIG. 10 is created and sent to the user by e-mail (steps S1411 to S1413). That is, based on the data number of each problem word registered in the user dictionary 122 and the registration cost of each problem word registered in the charging information DB 121 in association with the data number, the user dictionary 122 is actually registered. The registration fee and the total amount of each registered problem word are obtained, and an invoice as shown in FIG. 10 is created.

  In the above-described embodiment, the user dictionary 122 is arranged on the user dictionary creating apparatus 1 when creating a dictionary, but may be arranged on another computer on the network 4. Further, after the creation of the user dictionary 122 is completed, the user dictionary 122 may be used as it is on the user dictionary creation device 1 or on a computer in which machine translation means similar to the machine translation means 101 is introduced exclusively for the user. You may move to and use it. In the above-described embodiment, the bill is transmitted to the dictionary user. However, the registration fee may be deducted from a bank account designated by the dictionary user. Furthermore, in the above-described embodiment, the influence degree is calculated based on the appearance frequency and the part of speech of the problem word / phrase, but the appearance frequency itself may be used as the influence degree. Although not explained in the above-described embodiment, the mistranslation rate for each problem word / phrase on the registration effect display screen 50 (the ratio of mismatch between the machine translation sentence and the reference translation sentence. For example, the problem phrase “power ”Is 100, and the portion in which the machine translation and the reference translation do not match is 80, the mistranslation rate may be 80%).

[Effect of the first embodiment]
According to the present embodiment, it is possible for the user to easily select a problem word / phrase for which dictionary data should be registered in the user dictionary 122 from the problem words / phrases presented by the user dictionary creation device 1. Become. The reason for this is the mismatch between the machine translation and the reference translation for the original text, the problem word, and the influence (value related to the frequency of appearance and part of speech of the problem word in the original text, indicating the degree to which the translation results This is because the registration effect display screen 50 that displays in association with each other is transmitted to the dictionary user terminal 2.

  Further, according to the present embodiment, it is possible to present to the user the cost required for creating the user dictionary in a form that is clear and convinced by the user. The reason is that, for each problem word / phrase, a registration cost corresponding to the part of speech of the problem word / phrase is calculated, and a registration candidate cost display screen 80 for displaying the calculated registration cost and the problem word / phrase in association with each other is displayed on the dictionary user terminal. This is because a registration candidate / cost determination unit 107 to be transmitted to 2 is provided.

[Configuration of Second Embodiment]
Next, a second embodiment of the user dictionary creation system according to the present invention will be described. The present embodiment is characterized in that, in addition to the part of speech and the appearance frequency of the problem word, the influence on the translation result is calculated in consideration of the factor (problem type) that caused the mismatched part corresponding to the problem word. .

  The user dictionary creation system according to the present embodiment is realized by using the user dictionary creation device 1a shown in FIG. 17 instead of the user dictionary creation device 1 in FIG.

  The difference between the user dictionary creation device 1a used in the present embodiment shown in FIG. 17 and the user dictionary creation device 1 used in the first embodiment shown in FIG. A problem location analysis means 102a instead of the registration effect analysis means 104, a registration effect analysis means 104a instead of the registration effect analysis means 104, and a problem difficulty analysis means 106a instead of the problem difficulty analysis means 106. In addition to the registration candidate / cost determination means 107, the registration candidate / cost determination means 107a is provided, the machine translation dictionary 112a is provided instead of the machine translation dictionary 112, and the problem DB 117 is replaced with a problem. The point that the type DB 117a is provided, the point that the translation change degree DB 118a is provided instead of the translation change degree DB 118, and the dictionary difficulty that is provided instead of the dictionary difficulty degree DB 119 That it includes a DB119a, and translation knowledge DB126 is a point that has been added.

  The machine translation dictionary 112a includes a compound dictionary, a syntax dictionary, a pattern dictionary describing a vocabulary relationship, and a rule dictionary in addition to a word dictionary. The translation knowledge DB 126 includes a translation memory and the like.

  The problem location analysis unit 102a has a function of extracting a factor (problem type) that causes a mismatch between the machine translation text and the reference translation text, in addition to the function of the problem location analysis means 102.

  In the translation change degree DB 118a, a translation change degree corresponding to the part of speech of the problem phrase and a translation change degree corresponding to the problem type are registered. FIG. 18 is a diagram illustrating an example of the translation change DB 118a. Referring to the figure, the translation change degree DB 118a includes a part-of-speech part 1181 and a problem type part 1182. The part of speech part 1181 registers a translation change degree (part of speech translation change degree) in association with the part of speech, and the problem type part 1182 registers a translation change degree (problem type translation change degree) in association with the problem type. ing.

  The registration effect analysis unit 104a uses each of the problem phrases extracted by the problem phrase appearance frequency extraction unit 103, the part-of-speech translation change degree, and the problem type translation change degree registered in the translation change degree DB 118a. It has a function of calculating the degree of influence on the translation result of the question word.

  In the dictionary difficulty level DB 119a, a difficulty level corresponding to the part of speech of the problem phrase and a difficulty level corresponding to the problem type are registered. FIG. 19 shows an example of the dictionary difficulty DB 119a. As shown in the figure, the dictionary difficulty level DB 119a is composed of a part-of-speech part 1191 and a problem type part 1192. The part-of-speech part 1191 registers the difficulty level (part-of-speech difficulty) in association with the part-of-speech, and the problem type part 1192 registers the difficulty (class difficulty) in association with the problem type.

  The problem difficulty level analysis unit 106a has a function of acquiring the part-of-speech difficulty level and the type difficulty level for each question word phrase from the dictionary difficulty level DB 119a and passing them to the registration candidate / cost determination unit 107a.

  The registration candidate / cost determination means 107a has a function of calculating a registration cost for each question word based on the part-of-speech difficulty level and the type difficulty level.

  Note that the user dictionary creation device 1a can be realized by a computer, and when realized by a computer, for example, is as follows. A disk, a semiconductor memory, and other recording media on which a program for causing the computer to function as the user dictionary creation device 1a is recorded are prepared, and the program is read by the computer. The computer controls its own operation according to the read program, so that the machine translation unit 101, the problem location analysis unit 102a, the problem word appearance frequency analysis unit 103, the registration effect analysis unit 104a, the registration effect display designation on the computer. Means 105, problem difficulty level analysis means 106a, registration candidate / cost determination means 107a, bilingual dictionary list creation means 108, dictionary registration means 109, cost billing means 110, and transmission / reception means 111 are realized.

[Operation of Second Embodiment]
Next, the operation of the present embodiment will be described. When the original text to be translated is designated by the user of the dictionary user terminal 2 or the creator of the dictionary creator terminal 3, the machine translation means 101 reads the original text (text document) designated from the original text DB 114 or the parallel translation DB 116. Acquired (step S1301 in FIG. 20) and generates a machine translation sentence with reference to the machine translation dictionary 112a (step S1302).

  When the machine translation means 101 generates a machine translation sentence for the original sentence, the problem part analysis means 102a acquires a reference translation sentence for the original sentence (step S1303). Thereafter, the problem part analysis unit 102a extracts a mismatched portion by comparing the machine translation sentence generated by the machine translation unit 101 with the corresponding part of the reference translation sentence acquired in step S1303 (step S1304). Further, a problem word / phrase, a problem location, a part of speech and a problem type corresponding to the inconsistent portion are obtained (step S1305a). The difference between step S1305a and step S1305 in FIG. 13 is whether or not a question type is obtained. Thereafter, the problem part analysis means 102a associates and registers the original sentence, machine translation sentence, reference translation sentence, problem part, problem word / phrase, part of speech, and problem type in the problem type DB 117a (step S1306a). FIG. 22 shows an example of the contents of the problem type DB 117a. The items “appearance frequency” and “influence” in the problem type DB 117a are set by the problem word / phrase appearance frequency analysis unit 103 and the registered effect analysis unit 104a, respectively.

  Here, an example of how the problem type is obtained in step S1305a will be described. If the question word is a noun, the question type is “translation error”. If the question word is a verb, if the corresponding translation is different, it will be judged as “translation error”, and grammatical information such as SVO obtained from the syntax dictionary, pattern dictionary, rule dictionary, translation memory, and translated particle “ If the correspondence of “ha” is wrong, it is “syntax error”, and if the noun that each particle can take has a condition, it is “verb pattern recognition error”. If the question word is a verb idiom, it is assumed that the idiom is recognized incorrectly. The above is merely an example, and the problem type may be determined by other methods.

  When the problem location analysis unit 102a registers information on each question word / phrase (items “appearance frequency” and “influence” are not registered at this time) as shown in FIG. 22 in the problem type DB 117a, the problem word / phrase appearance frequency The analysis unit 103 obtains the appearance frequency of each question word in the original text, and registers the obtained appearance frequency in the problem type DB 117a (step S1307).

  When the problem phrase appearance frequency analysis means 103 obtains the appearance frequency of each problem word phrase registered in the problem type DB 117a, the registration effect analysis means 104a displays the part-of-speech translation change degree and problem type registered in the translation change degree DB 118a. Based on the translation change degree and the appearance frequency obtained by the problem word appearance frequency analysis unit 103, the degree of influence on the translation result is calculated for each question word registered in the problem type DB 117a. Registration in the problem type DB 117a (steps S1308a to S1310a).

  An example of the processing performed in steps S1308a to S1310a will be described in detail as follows. The registered effect analysis unit 104a pays attention to one entry in the problem type DB 117a, and checks whether or not the influence degree is registered in the entry. If registered, pay attention to the next entry.

  On the other hand, if the degree of influence is registered in the entry of interest, first, the part of speech registered in the entry of interest, the part of speech translation change corresponding to the problem type, and the problem type translation change are obtained from the translation change degree DB 118a. Acquire and add both. Thereafter, the influence level is calculated by multiplying the addition result by the appearance frequency registered in the entry of interest. That is, the degree of influence is calculated by performing the following calculation: influence degree = (part of speech translation change degree + question type translation change degree) × appearance frequency. For example, if the part of speech, question type, and appearance frequency registered in the entry of interest are “noun”, “translation error”, and “125”, the influence degree = (1 + 1) × 125 = 250. When a plurality of parts of speech are stored in the attention entry, in this embodiment, the largest translation change degree corresponding to each part of speech in the attention entry is used. You may make it use the value which added the translation change corresponding to a part of speech. Thereafter, the calculated influence degree is registered in the attention entry. At this time, if there is another entry in which the same problem word or noun as the target entry is registered, the calculated influence degree is also registered in that entry. The above processing is repeated until there are no more unregistered entries. The above is the details of the processing performed in steps S1308a to S1310a.

  Thereafter, the registration effect display designating unit 105 creates a registration effect display screen 50 as shown in FIG. 5 based on the contents of the problem type DB 117a (see FIG. 22) and transmits it to the dictionary user terminal 2 (step S1311). ).

  The registration candidate selection means 22 in the dictionary user terminal 2 displays the registration effect display screen 50 sent from the user dictionary creation device 1 on the display unit 26. The user selects a question word / phrase to be a candidate for registration in the user dictionary 122 in consideration of the degree of influence of each question word / phrase displayed on the registration effect display screen 50 (checks the selection unit 59). The registration candidate selection means 22 transmits registration candidate selection information (for example, including the data number of the selected problem word / phrase) indicating the selected problem word / phrase to the user dictionary creation device 1 (step S1312).

  When registration candidate selection information is sent from the dictionary user terminal 2, the registration effect display designating means 105 in the user dictionary creation device 1a includes information related to each question word / phrase registered in the question type DB 117a. Then, the information on the problem word / phrase that has not been selected by the user is deleted (step S1313a). As a result of this processing, only information relating to the question word / phrase selected by the user remains in the question type DB 117a.

  Thereafter, the problem difficulty level analysis means 106a refers to the dictionary difficulty level DB 119a, and acquires the part of speech difficulty level and type difficulty level of each question word registered in the problem type DB 117a (steps S1401 and S1402a in FIG. 21). After that, the registration candidate / cost determination means 107a determines each problem word / phrase based on the part-of-speech difficulty and type difficulty of each problem word acquired by the problem difficulty analysis means 106a and the unit price registered in the unit price DB 120. Every time, the registration cost to the user dictionary 122 is calculated (step S1403a). For example, if the contents of the dictionary difficulty level DB 119a and the unit price DB 120 are as shown in FIGS. 19 and 7, respectively, and the contents of the problem type DB 117a are as shown in FIG. 22, the registration cost of the question phrase “wear” is , Registration cost = (part of speech difficulty + class difficulty) × unit price = (3 + 1) × 100 = 400 yen. Note that the registration cost calculation method is not limited to this, and the frequency of occurrence of the problem word may also be considered. That is, registration cost = (part of speech difficulty + class difficulty) × appearance frequency × unit price.

  Thereafter, the registration candidate / cost determination unit 107a creates a registration candidate cost display screen 2300 as shown in FIG. 23 and transmits it to the dictionary user terminal 2 (step S1404). As shown in FIG. 23, the registration candidate cost display screen 2300 includes a data number part 2301, a problem word / phrase part 2302, an appearance frequency part 2303, a part of speech part 2304, a part of speech difficulty part 2305, and a question type part 2306. , A type difficulty level part 2307, a total difficulty level part 2308, a registration cost part 2309, and a selection part 2310. In the data number part 2301, a data number unique to the question word is set. In the question word / phrase part 2302, a question word / phrase is set. In the appearance frequency section 2302, the appearance frequency of the problem word / phrase is set. In the part of speech part 2304, the part of speech of the question word is set. In the part of speech difficulty part 2305, the part of speech difficulty of the question phrase is set. A problem type is set in the problem type section 2306. In the type difficulty level section 2307, the type difficulty level of the question word / phrase is set. In the total difficulty 2308, a total value of the part of speech difficulty and the type difficulty is set. In the registration cost unit 2309, a registration cost is set. The selection unit 2310 is a selection field used when the user selects a problem word / phrase to be registered in the user dictionary 122.

  The registration target selection means 23 in the dictionary user terminal 2 displays the registration candidate cost display screen 2300 sent from the user dictionary creation device 1 a on the display unit 26. The user selects a problem word / phrase to be registered in the user dictionary 122 from the problem words / phrases displayed on the registration candidate cost display screen 2300 in consideration of the registration cost (checks the selection unit 2310). ). As a result, the registration target selection unit 23 transmits registration target selection information indicating the selected problem phrase to the user dictionary creation device 1a (step S1405).

  When the registration candidate selection information is sent from the dictionary user terminal 2, the registration candidate / cost determination unit 107a charges the data number of the problem word / phrase selected by the user in association with the registration cost of the word / phrase. The information is registered in the information DB 121 (step S1406). After that, the registration candidate / cost determination unit 107a creates a list of data numbers (registration target list) of problem words / phrases that are registered by the user and passes them to the bilingual dictionary list creation unit 108 (step S1407).

  When the registration target list is passed from the registration candidate / cost determination unit 107a, the bilingual dictionary list creation unit 108 corresponds to the data number included in the registration target list from among the information registered in the problem type DB 117a. The registered information is acquired, and the acquired information is transmitted to the dictionary creator terminal 3. The bilingual dictionary list creation support means 32 in the dictionary creator terminal 3 displays the information sent from the user dictionary creation device 1 a on the display unit 34. The dictionary creator looks at this information and creates a bilingual dictionary list 90 as shown in FIG. 9 in which dictionary registration contents are set for each question word and returns it to the bilingual dictionary list creating means 108. The bilingual dictionary list creating means 108 sends the bilingual dictionary list sent from the dictionary creator terminal 3 to the dictionary user terminal 2 (step S1408).

  The registered content confirmation means 24 in the dictionary user terminal 2 displays the bilingual dictionary list sent from the user dictionary creation device 1 a on the display unit 26. If the contents of the bilingual dictionary list are correct, the user inputs confirmation OK from the input unit 25. If the contents are not correct, the user corrects the bilingual dictionary list. When the confirmation OK is input, the registered content confirmation unit 24 notifies the bilingual dictionary list creating unit 108 to that effect. When the bilingual dictionary list is modified, the registered bilingual dictionary list is displayed as the bilingual dictionary list. The message is transmitted to the creating means 108 (step S1409).

  When the confirmation OK is sent from the dictionary user terminal 2, the bilingual dictionary list creation means 108 passes the bilingual dictionary list created in step S1408 to the dictionary registration means 109, and the revised bilingual dictionary list is sent. If it does, it is passed to the dictionary registration means 109. Thereby, the dictionary registration means 109 registers the contents of the bilingual dictionary list in the user dictionary 122 (step S1410). Thereafter, the cost billing means 110 creates a bill as shown in FIG. 24 based on the data number of the problem word / phrase registered in the user dictionary 122, the contents of the charging information DB 121, and the contents of the dictionary user DB 123. Is transmitted to the user by electronic mail (steps S1411 to S1413).

[Effects of Second Embodiment]
In addition to the effect of the first embodiment, the present embodiment has an effect that the influence can be set finely. The reason is that the registration effect analysis means 104a for calculating the degree of influence in consideration of not only the appearance frequency and part of speech of the problem phrase but also the factor (problem type) that caused the mismatch between the machine translation sentence and the reference translation sentence is provided. Because.

  Furthermore, this embodiment has an effect that the registration fee can be set finely. The reason is that it includes a registration candidate / cost determination means 107a for calculating a registration cost in consideration of not only the part of speech of the problem phrase but also the factor (problem type) that caused the mismatch between the machine translation and the reference translation. Because.

  According to the present invention, it can be applied to a user dictionary development business of a machine translation system. Moreover, according to this invention, it can apply to uses, such as a user knowledge DB development business of a machine translation system.

It is a block diagram which shows the structural example of 1st Embodiment of the user dictionary creation system concerning this invention. It is a block diagram which shows the structural example of the user dictionary creation apparatus. It is a figure which shows an example of problem DB117. It is a figure which shows an example of translation variation DB118. 5 is a diagram illustrating an example of a registration effect display screen 50. FIG. It is a figure which shows an example of dictionary difficulty DB119. It is a figure which shows an example of unit DB120. 5 is a diagram illustrating an example of a registration candidate cost display screen 80. FIG. It is a figure which shows an example of the bilingual dictionary list. It is a figure which shows an example of an invoice. It is a block diagram which shows the structural example of the dictionary user terminal. 3 is a block diagram illustrating a configuration example of a dictionary creator terminal 3. FIG. 10 is a flowchart showing a processing example of machine translation means 101, problem location analysis means 102, problem word appearance frequency analysis means 103, registration effect analysis means 104, and registration effect display designation means 105. 7 is a flowchart showing a processing example of a problem difficulty level analysis unit 106, a registration candidate / cost determination unit 107, a bilingual dictionary list creation unit 108, a dictionary registration unit 109, and a cost claim unit 110. It is a figure which shows the example of a display (bar graph) of the influence degree in a registration effect display screen. It is a figure which shows the example of a display (pie graph) of the influence degree in a registration effect display screen. It is a block diagram which shows the structural example of the user dictionary creation apparatus 1a used by the 2nd Embodiment of this invention. It is a figure which shows an example of translation variation DB118a. It is a figure which shows an example of dictionary difficulty DB119a. 10 is a flowchart showing a processing example of a machine translation unit 101, a problem part analysis unit 102a, a problem word / phrase appearance frequency analysis unit 103, a registration effect analysis unit 104a, and a registration effect display designation unit 105. 7 is a flowchart showing a processing example of a problem difficulty level analysis unit 106a, a registration candidate / cost determination unit 107a, a bilingual dictionary list creation unit 108, a dictionary registration unit 109, and a cost billing unit 110. It is a figure which shows an example of problem classification DB117a. It is a figure which shows an example of the registration candidate expense display screen 2300. FIG. It is a figure which shows an example of an invoice.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1,1a ... User dictionary creation apparatus 101 ... Machine translation means 102, 102a ... Problem location analysis means 103 ... Problem word appearance frequency analysis means 104, 104a ... Registration effect analysis means 105 ... Registration effect display designation means 106, 106a ... Problem Difficulty level analysis means 107, 107a ... registration candidate / cost determination means 108 ... bilingual dictionary list creation means 109 ... dictionary registration means 110 ... cost billing means 111 ... transmission / reception means 112, 112a ... machine translation dictionary 113 ... user document storage section 114 ... Original DB
115 ... Translation DB
116 ... Bilingual DB
117 ... Problem DB
117a ... Problem type DB
118, 118a ... Translation change DB
119, 119a ... Dictionary difficulty DB
120 ... Unit price DB
121 ... Billing information DB
122 ... User dictionary 123 ... Dictionary user DB
124 ... Input unit 125 ... Display unit 126 ... Translation knowledge DB
2 ... Dictionary user terminal 21 ... Transmission / reception means 22 ... Registration candidate selection means 23 ... Registration object selection means 24 ... Registration content confirmation means 25 ... Input section 26 ... Display section 3 ... Dictionary creator terminal 31 ... Transmission / reception means 32 ... Bilingual dictionary List creation support means 33 ... input unit 34 ... display unit 4 ... network

Claims (30)

A user dictionary creation system including a user dictionary creation device and a user terminal,
The user dictionary creation device is
Machine translation means for translating the original text using a machine translation dictionary;
Extracting a mismatched portion between a machine translation sentence when the original sentence is translated by the machine translation means and a reference translation sentence for the original sentence, and extracting a word / phrase in the original sentence corresponding to the extracted mismatched part as a question word / phrase Problem location analysis means,
A problem word appearance frequency analyzing means for obtaining an appearance frequency in the original sentence of the problem word extracted by the problem part analyzing means;
A registration effect analyzing means for calculating an influence degree of the problem word / phrase on the translation result, which is a value related to the appearance frequency of the problem word / phrase obtained by the problem word / phrase appearance frequency analyzing means;
A registration effect display for displaying a mismatch effect part extracted by the problem location analysis means and a question word in association with each other and transmitting a registration effect display screen displaying the degree of influence calculated by the registration effect analysis means to the user terminal A designation means,
The user terminal is
Registration candidate selection means for transmitting registration candidate selection information indicating a problem word / phrase selected by a user among the question words / phrases displayed on the registration effect display screen to the user dictionary creating device is provided. User dictionary creation system. In the user dictionary creation system according to claim 1,
The registration effect analysis means
A system for creating a user dictionary, characterized in that an influence degree is calculated based on an appearance frequency of the question word / phrase obtained by the question word / phrase appearance frequency analyzing means and a part of speech of the question word / phrase. The user dictionary creation system according to claim 1,
The registration effect analysis means
The degree of influence is calculated on the basis of the appearance frequency of the problem word / phrase obtained by the problem word / phrase appearance frequency analysis means, the part of speech of the problem word / phrase, and the factor that caused the mismatched part corresponding to the problem word / phrase. A user dictionary creation system characterized by In the user dictionary creation system according to claim 1, 2, or 3,
The user dictionary creation device is
For each problem word specified by the registration candidate selection information, a registration candidate cost display screen for calculating a registration cost according to the part of speech of the problem word and displaying the calculated registration cost and the corresponding problem word Registration candidate / cost confirmation means to be transmitted to the user terminal,
A bilingual dictionary list creating means for creating a bilingual dictionary list including a problem word specified by registration target selection information and a word and phrase corresponding to the problem word in the reference translation sentence;
Dictionary registration means for registering the bilingual dictionary list created by the bilingual dictionary list creating means in a user dictionary;
The user terminal is
Registration object selection means for transmitting registration object selection information for specifying a problem word / phrase selected by the user among the problem words / phrases displayed on the registration candidate cost display screen to the user dictionary creation device is provided. A featured user dictionary creation system. In the user dictionary creation system according to claim 4,
Instead of the registration candidate / cost determination means,
For each problem word specified by the registration candidate selection information, calculate the registration cost according to the part of speech of the problem word and the factor that caused the mismatched part corresponding to the problem word, and the calculated registration cost A user dictionary creation system comprising registration candidate / cost determination means for transmitting a registration candidate cost display screen for displaying a corresponding problem word to the user terminal. In the user dictionary creation system according to claim 4 or 5,
User dictionary creation comprising cost billing means for creating a bill including a problem word registered in the user dictionary and a registration cost for the problem word calculated by the registration candidate / cost determination means system. In the user dictionary creation system according to any one of claims 1 to 6,
The user dictionary creation system, wherein the reference translated sentence is a translated sentence instructed by the user with respect to the original sentence. In the user dictionary creation system according to any one of claims 1 to 6,
The user dictionary creation system, wherein the reference translation is a translation created using a translation memory. Machine translation means for translating the original text using a machine translation dictionary;
Extracting a mismatched portion between a machine translation sentence when the original sentence is translated by the machine translation means and a reference translation sentence for the original sentence, and extracting a word / phrase in the original sentence corresponding to the extracted mismatched part as a question word / phrase Problem location analysis means,
A problem word appearance frequency analyzing means for obtaining an appearance frequency in the original sentence of the problem word extracted by the problem part analyzing means;
A registration effect analyzing means for calculating an influence degree of the problem word / phrase on the translation result, which is a value related to the appearance frequency of the problem word / phrase obtained by the problem word / phrase appearance frequency analyzing means;
A registration effect display designation for displaying a mismatch effect part extracted by the problem location analysis means and a question word in association with each other and transmitting a registration effect display screen displaying the degree of influence calculated by the registration effect analysis means to the user terminal And a user dictionary creation device. In the user dictionary creation device according to claim 9,
The registration effect analysis means
An apparatus for creating a user dictionary, characterized in that an influence degree is calculated based on an appearance frequency of the question word / phrase obtained by the question word / phrase appearance frequency analyzing means and a part of speech of the question word / phrase. In the user dictionary creation device according to claim 9,
The registration effect analysis means
The degree of influence is calculated on the basis of the appearance frequency of the problem word / phrase obtained by the problem word / phrase appearance frequency analysis means, the part of speech of the problem word / phrase, and the factor that caused the mismatched part corresponding to the problem word / phrase. A user dictionary creation device characterized by The user dictionary creation device according to claim 9, 10 or 11,
When the registration candidate selection information indicating the problem word / phrase selected by the user among the question words / phrases displayed on the registration effect display screen is sent from the user terminal, the registration candidate selection information For each problem word specified, a registration cost corresponding to the part of speech of the problem word is calculated, and a registration candidate cost display screen for displaying the calculated registration cost and the corresponding problem word is transmitted to the user terminal. Registration candidate / cost determination means,
When registration object selection information for specifying a problem word / phrase selected by the user among the problem words / phrases displayed on the registration candidate cost display screen is sent from the user terminal, the registration object selection information A bilingual dictionary list creating means for creating a bilingual dictionary list including a specified problem word and the word corresponding to the problem word in the reference translation sentence;
A user dictionary creation apparatus comprising: a dictionary registration unit for registering a bilingual dictionary list created by the bilingual dictionary list creation unit in a user dictionary. The user dictionary creation device according to claim 12,
Instead of the registration candidate / cost determination means,
For each problem word specified by the registration candidate selection information, calculate the registration cost according to the part of speech of the problem word and the factor that caused the mismatched part corresponding to the problem word, and the calculated registration cost A user dictionary creation device comprising registration candidate / cost determination means for transmitting a registration candidate cost display screen for displaying a corresponding problem word to the user terminal. The user dictionary creation device according to claim 12 or 13,
User dictionary creation comprising cost billing means for creating a bill including a problem word registered in the user dictionary and a registration cost for the problem word calculated by the registration candidate / cost determination means apparatus. The user dictionary creation device according to any one of claims 9 to 14,
The user dictionary creation device, wherein the reference translated sentence is a translated sentence instructed by the user with respect to the original sentence. The user dictionary creation device according to any one of claims 9 to 14,
The user dictionary creation apparatus, wherein the reference translation is a translation created using a translation memory. User dictionary creation device
A machine translation step for translating the original text using a machine translation dictionary;
Extracting a mismatched portion between the machine translation sentence when the original sentence was translated in the machine translation step and a reference translation sentence for the original sentence, and extracting a phrase in the original sentence corresponding to the extracted mismatched part as a problem phrase Problem location analysis step,
A problem word appearance frequency analysis step for obtaining an appearance frequency in the original sentence of the problem word extracted in the problem part analysis step;
A registration effect analysis step for calculating an influence degree of the problem word / phrase on the translation result, which is a value related to the appearance frequency of the problem word / phrase obtained in the problem word / phrase appearance frequency analysis step;
A registration effect display for displaying a mismatch effect part extracted in the problem location analysis step and a question word in association with each other and transmitting a registration effect display screen displaying the degree of influence calculated in the registration effect analysis step to the user terminal Execute the specified step,
The user terminal is
A registration candidate selection step of transmitting registration candidate selection information indicating a problem word / phrase selected by a user among the question words / phrases displayed on the registration effect display screen to the user dictionary creation device is executed. User dictionary creation method. The user dictionary creation method according to claim 17,
The registration effect analysis step comprises:
A user dictionary creation method, wherein an influence degree is calculated based on an appearance frequency of the question word / phrase obtained in the question word / phrase appearance frequency analysis step and a part of speech of the question word / phrase. The user dictionary creation method according to claim 17,
The registration effect analysis step comprises:
The degree of influence is calculated based on the appearance frequency of the question word / phrase obtained in the question word / phrase appearance frequency analysis step, the part of speech of the question word / phrase, and the factor that caused the mismatched part corresponding to the question word / phrase. To create a user dictionary. The user dictionary creation method according to claim 17, 18 or 19,
The user dictionary creation device is
For each problem word specified by the registration candidate selection information, a registration candidate cost display screen for calculating a registration cost according to the part of speech of the problem word and displaying the calculated registration cost and the corresponding problem word Registration candidate / cost confirmation step to be transmitted to the user terminal,
A bilingual dictionary list creating step of creating a bilingual dictionary list including a problem word specified by registration target selection information and a word and phrase corresponding to the problem word in the reference translation sentence;
A dictionary registration step of registering the bilingual dictionary list created in the bilingual dictionary list creating step in a user dictionary;
The user terminal is
Executing a registration object selection step of transmitting registration object selection information for specifying a problem word selected by the user among the problem words displayed on the registration candidate cost display screen to the user dictionary creation device; Characteristic user dictionary creation method. The user dictionary creation method according to claim 20,
Instead of the registration candidate / cost determination step,
For each problem word specified by the registration candidate selection information, calculate the registration cost according to the part of speech of the problem word and the factor that caused the mismatched part corresponding to the problem word, and the calculated registration cost A method for creating a user dictionary, comprising: performing a registration candidate / cost determination step of transmitting a registration candidate cost display screen for displaying a corresponding problem word to the user terminal. The user dictionary creation method according to claim 20 or 21,
The user dictionary creation device is
A user dictionary creation characterized by executing a cost billing step of creating a bill including a problem word registered in the user dictionary and a registration cost for the problem word calculated by the registration candidate / cost determination unit Method. The user dictionary creation method according to any one of claims 17 to 22,
The user dictionary creation method, wherein the reference translated sentence is a translated sentence instructed by the user with respect to the original sentence. The user dictionary creation method according to any one of claims 17 to 22,
The user dictionary creation method, wherein the reference translation sentence is a translation sentence created using a translation memory. Computer
Machine translation means for translating the original text using a machine translation dictionary,
Extracting a mismatched portion between a machine translation sentence when the original sentence is translated by the machine translation means and a reference translation sentence for the original sentence, and extracting a word / phrase in the original sentence corresponding to the extracted mismatched part as a question word / phrase Problem location analysis means,
A problem word appearance frequency analysis means for obtaining an appearance frequency in the original sentence of the problem word extracted by the problem part analysis means;
Registration effect analysis means for calculating the degree of influence of the problem word / phrase on the translation result, which is a value related to the appearance frequency of the problem word / phrase determined by the problem word / phrase appearance frequency analysis means;
A registration effect display designation for displaying a mismatch effect part extracted by the problem location analysis means and a question word in association with each other and transmitting a registration effect display screen displaying the degree of influence calculated by the registration effect analysis means to the user terminal Program to function as a means. The program according to claim 25,
The registration effect analysis means
A program characterized in that the degree of influence is calculated based on the appearance frequency of the problem word / phrase obtained by the problem word / phrase appearance frequency analyzing means and the part of speech of the problem word / phrase. The program according to claim 25,
The registration effect analysis means
The degree of influence is calculated on the basis of the appearance frequency of the problem word / phrase obtained by the problem word / phrase appearance frequency analysis means, the part of speech of the problem word / phrase, and the factor that caused the mismatched part corresponding to the problem word / phrase. A program characterized by The program according to claim 25, 26 or 27,
The computer,
When the registration candidate selection information indicating the problem word / phrase selected by the user among the question words / phrases displayed on the registration effect display screen is sent from the user terminal, the registration candidate selection information For each problem word specified, a registration cost corresponding to the part of speech of the problem word is calculated, and a registration candidate cost display screen for displaying the calculated registration cost and the corresponding problem word is transmitted to the user terminal. Registration candidate / cost determination means,
When registration object selection information for specifying a problem word / phrase selected by the user among the problem words / phrases displayed on the registration candidate cost display screen is sent from the user terminal, the registration object selection information A bilingual dictionary list creating means for creating a bilingual dictionary list including the specified problem word and the word corresponding to the problem word in the reference translation sentence;
A program for causing a bilingual dictionary list created by the bilingual dictionary list creating means to function as a dictionary registering means for registering in a user dictionary. The program according to claim 28,
The registration candidate / cost determination means is
For each problem word specified by the registration candidate selection information, calculate the registration cost according to the part of speech of the problem word and the factor that caused the mismatched part corresponding to the problem word, and the calculated registration cost A program characterized in that a registration candidate cost display screen for displaying a corresponding problem word is transmitted to the user terminal. The program according to claim 28 or 29,
The computer,
A program for functioning as a cost billing means for creating a bill including a problem word registered in the user dictionary and a registration cost for the problem word calculated by the registration candidate / cost determination means.

Images