JP3838904B2 - Dictionary device and natural language processing system - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a dictionary device and a natural language processing system, and can be applied to, for example, a machine translation system used by a plurality of users who translate a source language into a target language and a dictionary device that is an element of the machine translation system. .
[0002]
[Prior art]
In general, a machine translation system has a basic system dictionary and a user dictionary registered by a user. When a plurality of users use the same machine translation system, if a user dictionary is registered for each user, the same term must be individually registered by the plurality of users. For example, almost all users register for newly appearing new words that are used by almost all users.
[0003]
As a conventional technique for avoiding such inconvenience, there is one disclosed in Japanese Patent Application Laid-Open No. 11-134339. In this gazette, when a machine translation system is shared by a plurality of users, a plurality of users can search by searching from a user dictionary having a higher priority or extracting a parallel translation with the latest registration update time from the user dictionary. A system that can efficiently share and use a created user dictionary is disclosed.
[0004]
[Problems to be solved by the invention]
In the prior art described above, it is necessary to specify a priority relationship in order to determine the priority order of the user dictionary. The priority relationship is a priority between users who own user dictionaries, and indicates that the user dictionaries owned by users with higher priorities have higher reliability.
[0005]
However, comparing the contents of the user dictionaries owned by each user and specifying the priority relationship between the users requires labor and judgment ability to confirm the translation in each user dictionary and determine superiority or inferiority. Also, if the number of registrations is large, there are superiority and inferiority depending on words, so ranking among users is difficult. Furthermore, there is no guarantee that the priority relationship will remain the same in the future, and the priority relationship must be reviewed frequently.
[0006]
Further, in the above-described conventional technique, when there are words in the source language in a plurality of user dictionaries, the latest translation with the latest registration update time is extracted. Bilingual translation is not always reliable. Therefore, in order to obtain a high-quality translation result, it is necessary to always check the translation result and a plurality of user dictionaries.
[0007]
The present invention has been made in view of the above problems, and a plurality of users can obtain high-quality natural language processing results by sharing dictionary data related to user registration that maintains high-quality contents. It is an object of the present invention to provide a dictionary device and a natural language processing system that can be used.
[0008]
[Means for Solving the Problems]
In order to solve such a problem, the first aspect of the present invention provides a dictionary device provided in a natural language processing system used by a plurality of users, a plurality of technical term dictionaries for each category, and a tree having a general term dictionary as a root node. A dictionary main body hierarchically structured, a user dictionary registration means for setting a user dictionary in association with any of the technical term dictionary or the general term dictionary, and the dictionary data common to a plurality of user dictionaries Based on dictionary data common to a plurality of associated user dictionaries, set to a specialized term dictionary and / or the above general term dictionary, set a dictionary data common to a plurality of subordinate dictionaries in a tree structure A dictionary changing means for changing the dictionary and / or the dictionary data, such as setting a subordinate technical term dictionary in the technical term dictionary It characterized by having a dictionary history recording means for recording the history of the dictionary and or the dictionary change by the dictionary changing means.
[0009]
The natural language processing system according to the second aspect of the present invention includes the dictionary apparatus according to the first aspect of the present invention and a natural language processing main body that performs natural language processing using the stored contents of the dictionary apparatus.
[0010]
DETAILED DESCRIPTION OF THE INVENTION
(A) First embodiment
Hereinafter, a first embodiment in which a dictionary apparatus and a natural language processing system according to the present invention are applied to a machine translation system will be described in detail with reference to the drawings.
[0011]
(A-1) Configuration of the first embodiment
FIG. 1 is a block diagram illustrating a functional configuration of the machine translation system according to the first embodiment.
[0012]
1, the machine translation system 100 according to the first embodiment includes a browsing input unit 101, a control unit 102, a machine translation unit 103, a dictionary management unit 104, a dictionary database 105, and a dictionary history recording unit 106.
[0013]
The machine translation system 100 according to the first embodiment may be configured by a single information processing apparatus (for example, a personal computer) having a predetermined processing program and data, or is configured by a plurality of information processing apparatuses. May be. As the latter, the browsing input means 101 is installed in an information processing apparatus (client terminal), and other control means 102, machine translation means 103, dictionary management means 104, dictionary database 105, and dictionary history recording means 106 are replaced with other information. An example of mounting on a processing device (server) and an example of mounting the machine translation means 103 on another information processing device can also be mentioned. Further, there may be a plurality of browsing input means 101 and a plurality of machine translation means 103.
[0014]
The browsing input unit 101 allows the user to input a translation request document (source language document), to browse the translation result, or to input a word (a translation of the source language and the target language) for user registration, The user registration contents are browsed and the operation guidance message is browsed. The input unit in the browsing input unit 101 corresponds not only to a general input configuration such as a keyboard and a mouse, but also to a drive configuration in the case where a translation request document is read and input from a flexible disk or the like. The browsing unit in the browsing input unit 101 mainly corresponds to a display, but a printer corresponds to a printer that outputs a translation result or the like instead of a display output.
[0015]
The control unit 102 is connected to the browsing input unit 101, the machine translation unit 103, and the dictionary management unit 104, and controls which document is translated and which dictionary is used for translation.
[0016]
The machine translation unit 103 translates a source language document given via the control unit 102 into a target language document and sends it back to the control unit 102. In the case of the first embodiment, the machine translation unit 103 Any machine translation method may be used. The machine translation means 103 is also provided with dictionary data necessary for machine translation via the control means 102.
[0017]
The dictionary management means 104 contains a dictionary database 105 and is a means for creating a user dictionary, editing dictionary data, creating a technical term dictionary, integrating, dividing, generating, and the like under the control of the control means 102. .
[0018]
The dictionary history recording means 106 is provided in association with the dictionary management means 104 and is a means for recording a history such as a dictionary and dictionary data registration update time for each dictionary data.
[0019]
The dictionary database 105 is a technical term dictionary database that stores user dictionaries classified by field. The dictionary database 105 may also be constructed by being distributed among a plurality of information processing apparatuses.
[0020]
The dictionary database 105 has a hierarchical structure as shown in FIG. In other words, the plurality of dictionaries stored in the dictionary database 105 have a tree structure, the general term dictionary is a root node, and the next term includes specialized term dictionaries when specialized fields and genres are roughly classified. In the lower hierarchy, there is a technical term dictionary when the specialized fields and genres of the upper hierarchy are further classified, and the number of branches (depth of the hierarchy) leading to the terminal node is not uniform, As shown in FIG. 2, user dictionaries Dm1 to DmN are provided in association with each technical term dictionary Dm.
[0021]
Each user may create a user dictionary in any technical term dictionary field. The contents of this user dictionary are used not only when the created user requests translations, but also when other users request translations when registered and transferred to the technical term dictionary associated with the contents. Is done. Due to the functions of the dictionary management means 104 described above, a dictionary in the field of the document and a technical term dictionary belonging to a higher hierarchy in the field are used for translation. In the example of the hierarchical structure of FIG. 1, when translating documents in the computer field, a computer dictionary, an engineering dictionary, and a general term dictionary are used.
[0022]
(A-2) Operation of the first embodiment
(A-2-1) User dictionary registration
Next, the operation when the user registers a word in the user dictionary will be described with reference to the flowchart of FIG.
[0023]
For example, the user uses the browsing input unit 101 to instruct a word user registration mode. At this time, the control means 102 and the dictionary management means 103 start the user dictionary registration process shown in FIG. 3, and first determine whether or not the user can perform word user registration (step S1). ). Note that user information that can be registered as a user is also stored in the dictionary database 105, although not shown.
[0024]
If the user is not a person who can execute word user registration, the control means 102 returns that fact and ends the user dictionary registration process (step S2).
[0025]
On the other hand, if the user can perform word user registration, the control unit 102 retrieves the tree structure of the general term dictionary and the technical term dictionary from the dictionary management unit 104, and registers the user dictionary including information on the tree structure. The input screen is transmitted to the browsing input means 101 (step S3), and the registered dictionary data from the browsing input means 101 is awaited (step S4).
[0026]
When the user dictionary registration input screen is displayed, the user inputs dictionary data to be registered, and the user dictionary is associated with the registered dictionary data based on the tree structure of the displayed general term dictionary and technical term dictionary. The information of the desired dictionary is added and transmitted to the control means 102. Note that a plurality of registered dictionary data may be transmitted together. In the following description, it is assumed that one registered dictionary data is transmitted for the sake of simplicity.
[0027]
When the registered dictionary data is given, the control means 102 gives it to the dictionary management means 104. The dictionary management means 104 determines the user dictionary (the user dictionary related to the current registered dictionary data) for the user based on the stored contents of the dictionary database 105. It is determined whether or not a user dictionary associated with the technical term dictionary or the general term dictionary has already been set (step S5).
[0028]
If no user dictionary has been set, such a user dictionary is set (step S6), and if it has already been set, the registered dictionary data given from the browsing input means 101 is registered immediately. Then, the history of registering the dictionary data is recorded in the dictionary history recording means 106 (step S7), and the series of user dictionary registration processing is terminated.
[0029]
The dictionary database 105 functionally has a hierarchical structure as shown in FIG. 1, but for example, in practice, the registered dictionary data represents the type of dictionary regardless of the type of dictionary. The data is sequentially stored together with the data, and the history of dictionary data by the dictionary history recording means 106 is recorded in association with such a record.
[0030]
(A-2-2) Enhancement of technical term dictionary
Next, the operation of enriching the technical term dictionary will be described using the flowchart of FIG.
[0031]
Note that FIG. 4 corresponds to a case where it is periodically started or a case where it is started by a system operator. FIG. 4 shows processing when a certain technical term dictionary is targeted, and the processing shown in FIG. 4 is executed for all technical term dictionaries.
[0032]
When the processing shown in FIG. 4 is started, the dictionary management means 104 extracts and buffers the information of all user dictionaries from the dictionary database 105 while aggregating the same dictionary data for each related technical term dictionary. S11).
[0033]
FIG. 5 shows an example taken out while counting. From the left, information numbered for each different dictionary data (dictionary data ID), headwords, translated words, and registered numbers are shown. The dictionary data ID is numbered sequentially from 0.
[0034]
When the extraction while counting is completed, the dictionary management unit 104 sets the dictionary data ID to 0 (initial value) (step S12).
[0035]
Then, the processing loop in steps S13 to S17 is repeated until the enrichment process is completed for all dictionary data.
[0036]
The dictionary management unit 104 determines whether there is dictionary data whose registered number is more than a majority of the number of user dictionaries associated with the technical term dictionary (steps S13 and S14).
[0037]
If there is even one such dictionary data, all the dictionary data are additionally registered in the technical term dictionary (step S15). In this case, it is confirmed that the dictionary data does not exist in the technical term dictionary and registered.
[0038]
After such additional registration of dictionary data to the technical term dictionary is completed, or if there is no dictionary data whose registration number is greater than the majority of the number of user dictionaries, the dictionary data ID is incremented by 1 (step S16). ), It is determined whether or not unprocessed dictionary data remains (step S17).
[0039]
If unprocessed dictionary data remains, the process returns to step S13 described above. If the enrichment process is completed for all dictionary data, the series of processes shown in FIG. 4 is terminated.
[0040]
In the example of FIG. 5, if the number of user dictionaries associated with the technical term dictionary is five, the dictionary data “pencil; pencil” has three user dictionary registrations and is more than a majority. Through the process, it is registered in the target technical term dictionary.
[0041]
In the above description, the registration conditions for the technical term dictionary are set to be more than half of the number of user dictionaries, but registration requirements with other ratios as threshold values may be used. Further, the threshold value defining the registration requirement may be different depending on the number of user dictionaries.
[0042]
Further, in the above description, after the additional registration in the technical term dictionary, the dictionary data is left in the user dictionary. However, the dictionary data additionally registered in the technical term dictionary may be deleted from the user dictionary. . In the former case, in the additional registration process in step S15, the dictionary history recording means 106 records the history of additionally registering dictionary data in the corresponding technical term dictionary, and in the latter case, the dictionary data belongs. A history indicating that the dictionary type has been changed or deleted from the user dictionary is recorded.
[0043]
Further, in the above description, the registration to the technical term dictionary is determined from the registration status in all user dictionaries associated with the technical term dictionary, but from the registration status in some user dictionaries. Whether to register in the technical term dictionary may be determined. Furthermore, in the above, the registration to the terminology dictionary has been completely automated. However, after the server operator has displayed and confirmed the dictionary data to be added to the terminology dictionary, the terminology dictionary is displayed. You may make it register additionally.
[0044]
If there is a user dictionary associated with the general term dictionary, the general term dictionary may be enriched in the same manner as the above processing.
[0045]
(A-2-3) Dictionary information integration operation
Next, the operation | movement which integrates the information of several technical vocabulary dictionary is demonstrated using the flowchart of FIG. Information integration of a plurality of technical term dictionaries refers to checking information on data in technical term dictionaries and registering data in lower technical term dictionaries as data in higher technical term dictionaries.
[0046]
Note that FIG. 6 corresponds to a case where it is periodically started or a case where it is started by a system operator. Further, FIG. 6 shows processing when one certain technical term dictionary is targeted, and the processing shown in FIG. 6 is executed for all technical term dictionaries.
[0047]
When the processing shown in FIG. 6 is started, the dictionary management unit 104 first extracts dictionary data registered in all technical term dictionaries subordinate to the target technical term dictionary (step S21), and sets the dictionary data ID. It is set to 0 (initial value) (step S22).
[0048]
And it is discriminate | determined whether the dictionary data registered into more than the majority of the number of subordinate technical term dictionaries exist (step S23, S24).
[0049]
If there is even one such dictionary data, all the dictionary data are additionally registered in the higher-level technical term dictionary, and those dictionary data are deleted from the lower-level technical term dictionary (step S25). In this case, additional registration may be performed after confirming that the dictionary data does not exist in the higher-level technical term dictionary.
[0050]
Here, in the additional registration processing in step S25, the dictionary history recording means 106 has confirmed that the dictionary type to which the dictionary data belongs has been changed to a higher terminology dictionary or deleted from the lower terminology dictionary. A representing history is recorded.
[0051]
After the above processing is completed, the dictionary data ID is incremented by 1 (step S26), and it is determined whether or not unprocessed dictionary data remains (step S27). If unprocessed dictionary data remains, the process returns to step S22 described above. If the integration process has been completed for all dictionary data, the series of processes shown in FIG. 6 is terminated.
[0052]
In the example of the dictionary database 105 in FIG. 1, if the same dictionary data is registered in both the physical dictionary and the mathematical dictionary, the dictionary data is registered in the higher-level natural science dictionary, and the dictionary is added from the physical dictionary and the mathematical dictionary. Data is deleted.
[0053]
Here, the siphoning processing (integration processing) is performed in order from the technical term dictionary with the lower hierarchy to the technical term dictionary with the higher hierarchy, so that the dictionary data (word information) that has just been sifted is further higher. Sometimes it is taken up by a technical term dictionary.
[0054]
In the above description, the registration condition for the higher-level technical term dictionary is set to be registered with more than a majority of the lower-level technical term dictionaries. Moreover, the threshold value that defines the registration requirement for the upper level may be made different depending on the number of lower-level technical term dictionaries. In addition, each terminology dictionary manages the frequency and usage rate of dictionary data, and it is also possible to use dictionary data (word information) that is used to some degree in multiple subordinate terminology dictionaries. You may make it include in the registration requirement to a dictionary.
[0055]
In addition, in the above description, it has been shown that when additionally registering in the higher terminology dictionary, the dictionary data is deleted from the lower terminology dictionary. It may be left in the dictionary. In addition, in the above, the registration to the upper terminology dictionary is shown as fully automated, but after displaying the dictionary data to be additionally registered in the upper terminology dictionary to the system operator etc. It may be additionally registered in a higher-level technical term dictionary.
[0056]
Note that dictionary data registered in a plurality of major terminology dictionaries may be registered in the general term dictionary.
[0057]
(A-2-4) Dictionary division / generation operation
Next, the operation of dividing / generating the technical term dictionary will be described with reference to the flowchart of FIG. Note that the processing in FIG. 7 shows processing for a certain technical term dictionary, and the processing in FIG. 7 is executed for each technical term dictionary.
[0058]
When the processing shown in FIG. 7 is started, the dictionary management unit 104 extracts and buffers related user dictionary information from the dictionary database 105 while counting the same dictionary data (step S31).
[0059]
FIG. 8 shows an example extracted from the user dictionary related to the sports dictionary while counting. From left, dictionary data ID, headword (key), translation, and number of registered user dictionaries are shown.
[0060]
When the extraction while counting is completed, the dictionary management unit 104 sets the dictionary data ID to 0 (initial value) (step S32). Then, the processing loop in steps S33 to S38 is repeated for all dictionary data.
[0061]
The dictionary management unit 104 checks whether there is dictionary data having different translation (value) with the same headword (key) in the registered dictionary data of the user dictionary associated with the technical term dictionary. Determines whether or not the number of user dictionaries registering the respective dictionary data is 20% or more of the total number of user dictionaries associated with the technical term dictionary (the ratio is not limited to this). (Steps S33 and S34).
[0062]
If there is user-registered dictionary data with the same headword (key) and different translated words (values), and the registration ratio of each dictionary data is 20% or more, the dictionary management means 104 is subordinate to the technical term dictionary. It is determined whether or not there is a technical term dictionary (step S35).
[0063]
When there is no lower-level technical term dictionary, the dictionary management unit 104 uses the same headword (key) with different translations (values) and lower technical term dictionaries by the number of types of dictionary data whose registration ratio is 20% or more. Is created and each dictionary data is registered (step S36). Here, the user dictionary including such dictionary data may be directly associated with the technical term dictionary defined by the dictionary data ID, or may be associated with the newly created technical term dictionary in which the registered dictionary data is registered. You may do it. In addition, the creation of the technical term dictionary at a lower level here is not complete creation but may be provisional creation.
[0064]
In the process of step S36 described above, the history of the creation of a new technical term dictionary as described above and the history of registration of certain dictionary data in the dictionary are recorded by the dictionary history recording means 106. Things are also done.
[0065]
On the other hand, when there is a low-level technical term dictionary, the dictionary management unit 104 stores the dictionary data with the same headword (key) with different translations (values) and a registration ratio of 20% or more, Move to the technical term dictionary (step S37). Here, the user dictionary including such dictionary data may be associated with the original technical term dictionary as it is, or may be associated with the moved lower technical term dictionary.
[0066]
In the processing of step S37 described above, the history that the dictionary data has been moved to the lower-level technical term dictionary is also recorded by the dictionary history recording means 106.
[0067]
The type of the lower-level technical term dictionary to which the dictionary data is moved is determined, for example, by recognizing the field or genre with reference to the contents of the dictionary database for reverse direction using the translated word as the headword. Further, for example, a part or all of the character string in the translated word of the current dictionary data is moved to the technical term dictionary on the lower side where dictionary data including the translated character string exists. Here, the movement of the dictionary data to the lower-level technical term dictionary may be a temporary movement, not a complete movement.
[0068]
The dictionary management unit 104 finishes the new creation of the lower-level technical term dictionary and the movement of the predetermined dictionary data to the lower-level technical term dictionary, or the translation (value) differs for the same headword (key). If there is no user-registered dictionary data, or even if it exists, if the registration ratio of any dictionary data is less than 20%, the dictionary data ID is incremented by 1 (step S38), and the dictionary data remains. It is discriminate | determined whether it is (step S39).
[0069]
If dictionary data remains, the process returns to step S33 described above, and if the dictionary dividing / generating process is completed for all dictionary data, the series of processes shown in FIG. 7 is terminated.
[0070]
In the dictionary division / generation process, complete automation is difficult, and it is necessary to perform post-processing by an operator on the system operator side (hereinafter referred to as a system operator) at the end of the process of FIG. For example, it is possible to assign a field or genre naming to a newly created technical term dictionary, determine whether the creation is necessary if the creation in step S36 is provisional creation, or if the movement in step S37 is provisional movement. It is necessary to determine whether or not this movement is necessary, or to perform processing for moving other dictionary data from a higher-level terminology dictionary when a new lower-level terminology dictionary is created.
[0071]
For example, in the example of FIG. 8, if the number of user dictionaries associated with the sports dictionary is 10, the two types of dictionary data with the headword “Pitcher” and the translations “Pitcher” and “7th Iron” Since the number of user registrations is 3, it satisfies the requirements for newly creating a lower-level technical term dictionary or the dictionary data movement requirement of the lower-level technical term requirements.
[0072]
In this case, if there is no lower-level technical term dictionary in the sports dictionary, two lower-level technical term dictionaries are created, dictionary data “Pitcher; 7th Iron” is registered on one side, and dictionary on the other side Data “Pitcher” is registered. In the post-processing by the system operator, the former will be given a golf dictionary and the latter a baseball dictionary.
[0073]
On the other hand, when the golf dictionary and the baseball dictionary exist below the sports dictionary, the dictionary data “Pitcher; 7th Iron” is registered in the golf dictionary, and the dictionary data “Pitcher; pitcher” is registered in the baseball dictionary. Is done. This is because, for example, if another dictionary data whose translation is “iron” exists in the golf dictionary, a common character string with the current translation “7th iron” exists. Further, for example, if there is another dictionary data whose translation is “right wing” in the baseball dictionary, there is a common character string with the current translation “pitcher”.
[0074]
FIG. 9 shows a change state of the dictionary database 105 when a golf dictionary and a baseball dictionary are newly created below the sports dictionary from the state where only such a sports dictionary is prepared, through the processing of FIG. FIG. 10 shows the movement of the associated user dictionary when the technical term dictionary is divided and generated as shown in FIG.
[0075]
In the above description, only the registration information of the user dictionary is used to determine whether or not a lower-level technical term dictionary is created, but the lower-level specialized term dictionary is determined from the registration information of the user dictionary and the registration information of the technical term dictionary. Whether or not a term dictionary is created may be determined. For example, when dictionary data of user registration with the same headword as the registration information of the technical term dictionary and different translated words is registered in a predetermined number or more or a predetermined ratio of user dictionaries, creation of a subordinate technical term dictionary, etc. May be performed.
[0076]
Moreover, in the above description, the case where the lower-level technical term dictionary is created only when the lower-level technical term dictionary does not exist is shown. It is also possible to create a technical term dictionary below. That is, regardless of the presence or absence of a lower-level technical term dictionary, another lower-level technical term dictionary may be created if the condition is satisfied (in other words, the lower-level migration process may be eliminated). For example, even if there is a judo dictionary or a land dictionary under the sports dictionary, “Pitcher; 7th Iron” or “Pitcher” may be registered as a user in connection with the sports dictionary as described above. In this case, a golf dictionary or a baseball dictionary (named later) is created as a subordinate technical term dictionary.
[0077]
(A-2-5) Dictionary browsing operation
Next, the operation when browsing the stored contents of the dictionary will be described with reference to the flowchart of FIG.
[0078]
The user may want to browse the contents of the dictionary in advance, for example, when registering dictionary data or when confirming what kind of translation the source language word is translated prior to machine translation processing.
[0079]
In such a case, the user inputs a browsing request through the browsing input unit 101 and transmits it to the control unit 102 (step S41).
[0080]
Here, the browsing request may be a request including only a headword (source language word), may be a request including only a translated word, and further includes a pair of a headword and a translated word. May be. Further, it may be a browsing request that limits the type of dictionary to be searched. In the case where a certain terminology dictionary is limited as a dictionary to be searched, only the terminology dictionary may be set as a search target, and the terminology dictionary and all user dictionaries associated with the terminology dictionary are set as search targets. You may make it do.
[0081]
The control unit 102 gives the received browsing request to the dictionary management unit 104. The dictionary management unit 104 searches the dictionary data in response to the browsing request, and the history information regarding the dictionary data obtained by the search is also stored in the dictionary history recording unit. It takes out from 106 (step S42).
[0082]
Thereafter, the dictionary management means 104 returns the dictionary data (including history information) obtained by the search to the browsing input means 101 via the control means 102, and the browsing input means 101 displays or prints out (step S43). ).
[0083]
(A-2-6) Machine translation processing
Next, a machine translation operation for an input sentence or an input document will be described with reference to the flowchart of FIG.
[0084]
If there is a sentence or document desired to be machine-translated, the user inputs it into the browsing input means 101 and transmits it to the control means 102, and the control means 102 gives the received input sentence or input document to the machine translation means 103 (step S51). When the input document is an HTML document, the control unit 102 cuts out only the sentence portion and gives it to the machine translation unit 103.
[0085]
In addition, when the user wants to limit the dictionary used in machine translation, the designated dictionary information is input to the browsing input means 101 and transmitted to the control means 102, and the dictionary management to which the designated dictionary information is given from the control means 102 The means 104 validates only the corresponding dictionary in the dictionary database 105 (step S52).
[0086]
As described above, the specified dictionary and the technical term dictionary or general term dictionary that belong to the higher level of the field are valid, and the user's associated with the valid technical term dictionary or general term dictionary is valid. The user dictionary is valid. In the example of the machine translation processing shown in FIG. 12, the user specifies a dictionary to be used. However, the dictionary management unit 104 applies an existing specialized field recognition method or the like from an input sentence or an input document. You may make it recognize automatically the dictionary to perform.
[0087]
Thereafter, the machine translation means 103 performs machine translation while referring to the stored contents of the activated dictionary (step S53). Here, all the stored contents of the validated dictionary may be given to the machine translation means 103 to execute machine translation, and the machine translation means 103 is connected via the control means 102 and the dictionary management means 104. Then, the machine translation may be executed by referring to the stored contents of the valid dictionary each time.
[0088]
If the dictionary history recording means 106 or the dictionary database 105 manages the usage frequency and usage rate of dictionary data, the usage frequency and usage rate of dictionary data used in machine translation are updated. The usage frequency and usage rate of dictionary data may also be browsed when a dictionary data browsing request is made.
[0089]
The obtained translation result is given to the control means 102 by the machine translation means 103, and the control means 102 returns it to the browsing input means 101 (step S54). When the input document (translation request document) is a document such as an HTML document and the control unit 102 cuts out only the sentence portion and gives it to the machine translation unit 103, the control unit 102 displays the translation result. From this, the HTML document is reconstructed and returned to the browsing input means 101.
[0090]
(A-3) Effects of the first embodiment
As described above, since a dictionary database that maintains high-quality contents is shared by a plurality of users, high-quality translation results can be obtained, and the dictionary database is frequently changed because it performs changes for quality improvement. Therefore, it can always be used in the latest state, and the maintenance man-hours can be reduced.
[0091]
In addition, the user can not only browse the dictionary data of such a dictionary database, but also can view the update history, so not only can grasp the necessity of user registration, but also when other users can You can also know when it has been updated and when it has been downloaded to the technical term dictionary, and can recognize the current state of the world for each term.
[0092]
(B) Second embodiment
Next, a second embodiment in which the dictionary device and the natural language processing system according to the present invention are applied to a machine translation system will be described in detail with reference to the drawings.
[0093]
(B-1) Configuration of the second embodiment
FIG. 13 is a block diagram illustrating a functional configuration of the machine translation system according to the second embodiment. The same reference numerals are given to the same or corresponding parts as those in FIG. 1 according to the first embodiment. .
[0094]
In FIG. 13, the machine translation system 100A of the second embodiment is similar to the first embodiment in the browsing input means 101, the control means 102, the machine translation means 103, the dictionary management means 104, the dictionary database 105, and the dictionary history record. In addition to the means 106, it further includes a network 201, a communication means 202, and an update information request database (update information request DB) 203.
[0095]
That is, in the machine translation system 100A of the second embodiment, the browsing input unit 101 is connected to the other system components 102 to 106, 202, and 203 via the network 201.
[0096]
For example, the network 201 is the Internet, and the system components 102 to 106, 202, and 203 other than the browsing input unit 101 are mounted on a server (for example, a provider server), and the browsing input unit 101 has a browser function. Corresponds to an information processing terminal having
[0097]
Note that the first embodiment described above may also include a network as in the second embodiment except for the update information request database 203.
[0098]
The communication unit 201 performs communication between the browsing input unit 101 and the control unit 102 via the network 201.
[0099]
An update information request database 202 is connected to the communication unit 201, and the communication unit 201 updates the update information when accessed from any of the browsing input units 101 (in other words, from any of the users). It has a function of confirming the contents of the request database 202. Access that triggers confirmation of the contents of the update information request database 202 may be limited to access of processing related to machine translation (including not only machine translation processing but also dictionary browsing request processing). It may be an access of a process not related to (for example, a simple document search process).
[0100]
The update information request database 202 stores update information type data related to the dictionary database 105 for which each user requests automatic transmission (automatic notification).
[0101]
FIG. 14 shows an example of the contents stored in the update information request database 202. In the case of the example in FIG. 14, the user 1 has updated the computer dictionary and the mathematical dictionary (which may or may not include the user dictionary associated therewith) in the dictionary database 105. The user 2 requests automatic transmission of the dictionary data related to the update, and the user 2 requests automatic transmission of the dictionary data related to the update when the natural science dictionary in the dictionary database 105 is updated. When the dictionary 3 whose entry word is “pencil” in the dictionary database 105 is updated, the user 3 requests automatic transmission of dictionary data or the like related to the update. In the update information request database 202, the previous automatic transmission time may be recorded in order to clarify to what time the dictionary data related to the update is already transmitted.
[0102]
The functions of the browsing input unit 101 and the dictionary management unit 104 are slightly different from those of the first embodiment in relation to the components newly added in the second embodiment as described above. This point will be clarified in the explanation of the operation described later.
[0103]
(B-2) Operation of the second embodiment
Also in the machine translation system 100A of the second embodiment, various processes described in the first embodiment are executed.
[0104]
Furthermore, the machine translation system 100A according to the second embodiment also performs a process of automatically notifying the user of update information such as dictionary data updated in the dictionary database 105.
[0105]
The update information notification process will be described in detail below with reference to the flowchart of FIG. FIG. 15 shows processing for one user.
[0106]
Although illustration of the flowchart is omitted, the user inputs update information type data for requesting automatic transmission of update information from the browsing input unit 101, and an update provided in association with the communication unit 202 in advance. It is registered in the information request database 203. Registration in the update information request database 203 may be performed when the system operator enters into a user contract or the like instead of the method based on the input operation by the user as described above.
[0107]
When a certain user accesses the communication unit 202 from the browsing input unit 101 for some processing (for example, machine translation processing), for example, before starting the processing or after finishing the processing, The processing in FIG. 15 is started. First, the communication unit 202 determines whether or not there is update information type data (update information request) in the update information request database 203 for the user (step S61).
[0108]
The process shown in FIG. 15 may be applied by a method in which the communication unit 202, the control unit 102, or the dictionary management unit 104 is periodically started instead of the method that is started by access by the user as described above. Further, a method in which the dictionary management unit 104 recognizes the update of the dictionary database 105 and starts automatically may be applied. In the case of the two start-up methods mentioned as modifications, there is a possibility that the power of the browsing input means 101 is off, so the notification operation in step S65 described later is e-mail sending, for example.
[0109]
If the update information type data (update information request) is not stored in the update information request database 203 for the accessing user, the processing shown in FIG. 15 is immediately terminated.
[0110]
On the other hand, if update information type data (update information request) is stored in the update information request database 203 for the user who has accessed, the communication means 202 cooperates with the dictionary management means 104 to create a dictionary. After confirming that the database 105 has been updated, the dictionary history recording means 106 is searched to determine whether there is update information for which the user is requesting automatic transmission of update information (steps S62 to S64).
[0111]
More specifically, it is confirmed that the dictionary database 105 has been updated since the time of the last automatic transmission to the user, and dictionary data whose update time is the time after the last automatic transmission for the user, etc. In this case, the dictionary data specified by the update information type data stored in the update information request database 203 is searched for from the dictionary history recording means 106.
[0112]
For example, in the case where the accessing user is the user 2 shown in FIG. 14, if it is confirmed that the dictionary database 105 has been updated since the time of the last automatic transmission to the user 2, Whether the dictionary data contains dictionary data that has been updated since the time of the last automatic transmission to the user 2 is determined by searching the dictionary history recording means 106.
[0113]
If the dictionary database 105 has not been updated (negative result in step S63), or if there is no dictionary data to be notified (negative result in step S65), the processing shown in FIG. 15 ends.
[0114]
If there is corresponding dictionary data to be notified to the user, the communication unit 202 notifies the browsing input unit 101 of the dictionary data and / or update history, and ends the processing shown in FIG.
[0115]
Note that FIG. 15 shows a case where update information is extracted at a timing at which automatic transmission should be performed. However, when the dictionary database 105 is updated, automatic transmission is performed to each user based on the stored contents of the update information request database 203. Update information stored in the update information request database 203 without copying the dictionary database 105 and the dictionary history recording means 106 at the timing when the update information should be copied to the update information request database 203 and automatically transmitted. May be automatically transmitted.
[0116]
(B-3) Effects of the second embodiment
According to the second embodiment, the same effects as those of the first embodiment described above can be obtained.
[0117]
Furthermore, according to the second embodiment, the update information of the dictionary database can be automatically obtained according to the user's desire, so that even if the dictionary database is updated by user registration or the like by another user, It can be grasped early. As a result, the user can prevent unnecessary user registration and can accurately grasp the translation quality.
[0118]
(C) Third embodiment
Next, a third embodiment in which the dictionary device and the natural language processing system according to the present invention are applied to a machine translation system will be described in detail with reference to the drawings.
[0119]
(C-1) Configuration of the third embodiment
FIG. 16 is a block diagram illustrating a functional configuration of the machine translation system according to the third embodiment. The same reference numerals are given to the same or corresponding parts as those in FIG. 13 according to the second embodiment. .
[0120]
In FIG. 16, a machine translation system 100B according to the third embodiment includes a translation history recording unit 301 and a translation result database 302 in addition to the configuration of the second embodiment described above.
[0121]
The machine translation system 100B of the third embodiment is based on the premise that a request for translation of the same document is generated from a plurality of browsing input means 101. For example, when the network 201 is the Internet, a plurality of browsing input units 101 may retrieve the same document from a document search device (not shown), and a plurality of browsing inputs may be used for machine translation of the same searched document. Means 101 may find out. The machine translation system 100B of the third embodiment is based on such a network system.
[0122]
The translation result database 302 stores a translation request document (original text), the latest translation result corresponding to the translation request document, and a registration time (may be an update time) when the latest translation result is recorded in the translation result database 302. Is.
[0123]
The translation history recording unit 301 executes update processing of the translation result database 302 and the like under the control of the control unit 102.
[0124]
Note that the functions of the control means 102 and the like are slightly different from those of the second embodiment. The update function of the translation result will be clarified in the operation explanation described later.
[0125]
In addition, the control means 102 of the third embodiment also has the following functions. When there is a machine translation request for a document from the browsing input unit 101, the control unit 102 checks whether the translation result of the document is registered in the translation result database 302 via the translation history recording unit 301. If the translation result of the document is not registered in the translation result database 302, the machine translation unit 103 causes the document to be translated, and the translation result is returned to the browsing input unit 101. Thus, information such as a translation result is registered in the translation result database 302. On the other hand, if the translation result of the translation request document is registered in the translation result database 302, it is taken out and returned to the browsing input means 101.
[0126]
(C-2) Operation of the third embodiment
Also in the machine translation system 100B of the third embodiment, various processes described in the second embodiment are executed.
[0127]
Furthermore, the machine translation system 100B of the third embodiment also performs a process of updating the translation result that has already been machine translated.
[0128]
The translation result update process according to the third embodiment will be described in detail below with reference to the flowchart of FIG.
[0129]
For example, if the received translation result does not reflect dictionary data registered by the user, or if it is desired to improve the quality of the received translation result, the document (which may be a translation request document or a translation result document) or An instruction to update the translation result with the document specifying information (for example, URL) is given to the browsing input means 101 (step S71).
[0130]
Thereby, the concrete update process of a translation result like step S72 and after is started. It should be noted that the specific update processing of the translation result as in step S72 and after may be started by the system operator, or automatically when the dictionary management means 104 recognizes the update of the dictionary data in the dictionary database 105. It may be performed automatically or automatically at a predetermined cycle. Alternatively, it may be automatically performed when a new machine translation is requested for a document in the source language in which the translation result is registered.
[0131]
When an instruction to update the translation result in which the translation result is designated directly or indirectly is given from the browsing input means 101 side, the control means 102 sends the designated translation result to the translation result via the translation history recording means 301. It is confirmed whether it exists in the database 302 (steps S72 and S73). If not, the process shown in FIG. 17 ends (note that there is no reply to the browsing input means 101).
[0132]
If the designated translation result is in the translation result database 302, the control means 102 obtains the update time immediately before the translation result (which may be the initial registration time) via the translation history recording means 301 ( In step S74), the dictionary history recording unit 106 is accessed via the dictionary management unit 104, and dictionary data whose update time is later than the update time is searched (step S75), and the conditions for such update time are satisfied. It is determined whether or not dictionary data exists (step S76). If there is no such dictionary data, the processing shown in FIG. 17 is terminated (a reply that the updating to the browsing input means 101 cannot be performed is performed).
[0133]
If there is dictionary data that satisfies the update time condition, the control unit 102 accesses the translation result database 302 via the translation history recording unit 301 and searches for a sentence including such dictionary data in the source language document ( Step S77), it is determined whether or not such a sentence exists (S78). If there is no source language sentence including dictionary data that satisfies the update time condition, the processing shown in FIG. 17 is terminated (a reply that there is no source language sentence that is affected by the dictionary data update to the browsing input means 101) Do).
[0134]
If there is a source language sentence including dictionary data that satisfies the update time condition, the control means 102 fetches the source language sentence (if there are a plurality) from the translation result database 302 and gives it to the machine translation means 103 for machine translation. (Steps S79 and S80). Thereafter, the control unit 102 gives the translation result for the sentence cut out by the machine translation unit 103 and the update time to the translation history recording unit 301 to update the translation result database 302 (step S81), as shown in FIG. The process is terminated (replying the update to the browsing input means 101 is performed.
[0135]
As long as the machine translation means 103 translates in consideration of the context and the like, not only the sentence including the updated dictionary data but also the entire source language document may be given for machine translation. Further, the updated translation result may be returned to the browsing input means 101 instead of being updated.
[0136]
The update of the translation result is set in the update information request database 203 as an automatic transmission target, and the update of the translation result corresponding to the update information notification process described in the second embodiment (viewing input) Means 101) may be notified.
[0137]
(C-3) Effects of the third embodiment
According to the third embodiment, the same effect as that of the second embodiment described above can be obtained.
[0138]
Furthermore, according to the third embodiment, when the translation result is cached, only sentences including words updated after the cached time can be translated. It can be updated to the latest dictionary translation results.
[0139]
(D) Other embodiments
In the description of each of the above embodiments, various modified embodiments have been mentioned, but further modified embodiments as exemplified below can be cited.
[0140]
In the description of the first embodiment, it is described that the contents of the user dictionary of another user can be browsed, but browsing of the user dictionary of another user may not be permitted. However, when it is taken up in the technical term dictionary, it can be browsed naturally.
[0141]
The technical term dictionary includes not only those classified in so-called specialized fields, but also those classified in categories related to hobbies and the like.
[0142]
In the second embodiment, the update history is automatically transmitted. However, the update history of the type of dictionary data desired by the user is displayed in response to the update history transmission request (not including specification information such as the field). You may make it transmit.
[0143]
In each of the above-described embodiments, the technical terms dictionary is not directly registered by the user. However, the technical terms dictionary may be directly registered by the user. Even in this case, registration in some user dictionaries may be used as a condition.
[0144]
In the third embodiment, the update of the translation result record is performed after the translation result reflecting the new dictionary data and the translation result not reflected are presented to the user or the system operator and confirmed. Anyway.
[0145]
In each of the above embodiments, the case where the dictionary device and the natural language processing system of the present invention are used in a machine translation system has been shown, but the application target of the present invention is not limited to this. For example, the dictionary apparatus of the present invention can also be applied to a kana-kanji conversion dictionary used in speech recognition or the like.
[0146]
【The invention's effect】
As described above, according to the present invention, a dictionary device in which a plurality of users can obtain high-quality natural language processing results by sharing dictionary data related to user registration that maintains high-quality content. And a natural language processing system.
[Brief description of the drawings]
FIG. 1 is a block diagram illustrating a functional configuration of a machine translation system according to a first embodiment.
FIG. 2 is an explanatory diagram illustrating a relationship between a user dictionary and a technical term dictionary according to the first embodiment.
FIG. 3 is a flowchart illustrating user dictionary registration processing according to the first embodiment;
FIG. 4 is a flowchart illustrating a technical term dictionary enrichment process according to the first embodiment;
5 is an explanatory diagram of user-registered dictionary data registered in the technical term dictionary by the processing of FIG. 4;
FIG. 6 is a flowchart illustrating a process for integrating technical term dictionary information according to the first embodiment.
FIG. 7 is a flowchart showing a jargon dictionary dividing / generating process according to the first embodiment;
FIG. 8 is an explanatory diagram of user-registered dictionary data in which a lower-level technical term dictionary is generated by the process of FIG.
FIG. 9 is an explanatory diagram showing the structure of a dictionary database before and after a lower-level technical term dictionary is generated by the processing of FIG. 7;
FIG. 10 is an explanatory diagram showing movement of a user dictionary when a lower-level technical term dictionary is generated by the processing of FIG.
FIG. 11 is a flowchart showing dictionary browsing processing according to the first embodiment;
FIG. 12 is a flowchart showing machine translation processing according to the first embodiment;
FIG. 13 is a block diagram illustrating a functional configuration of a machine translation system according to a second embodiment.
FIG. 14 is an explanatory diagram illustrating an example of contents stored in an update information request database according to the second embodiment;
FIG. 15 is a flowchart showing a dictionary update information notification process according to the second embodiment;
FIG. 16 is a block diagram illustrating a functional configuration of a machine translation system according to a third embodiment.
FIG. 17 is a flowchart showing translation result update processing according to the third embodiment;
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 100, 100A, 100B ... Machine translation system, 101 ... Browsing input means, 102 ... Control means, 103 ... Machine translation means, 104 ... Dictionary management means, 105 ... Dictionary database, 106 ... Dictionary history recording means, 201 ... Network, 202 ... Communication means 203 ... Update information request database 301 ... Translation history recording means 302 ... Translation result database

Claims (6)

In a dictionary device provided in a natural language processing system used by a plurality of users,
A dictionary body in which multiple terminology dictionaries for each category are hierarchized in a tree structure with the general term dictionary as the root node,
User dictionary registration means for setting a user dictionary in association with any of the technical term dictionary and / or the general term dictionary,
Set the dictionary data common to multiple user dictionaries to the above-mentioned technical term dictionary or general term dictionary, or set the dictionary data common to multiple lower-level dictionaries in the tree structure to the higher-level dictionary, or associate A dictionary changing means for changing the dictionary and / or the dictionary data, such as setting a subordinate technical term dictionary in the technical term dictionary based on dictionary data common to a plurality of user dictionaries,
A dictionary apparatus comprising a dictionary history recording means for recording a dictionary and / or dictionary data change history by the dictionary changing means. A browsing request fetching unit for fetching a browsing request for dictionary data from a browsing input unit operated by a user;
In response to the browsing request, the dictionary data is retrieved and retrieved from the dictionary body, and a change history corresponding to the dictionary data is retrieved from the dictionary history recording unit, and a browsing request response unit for returning to the browsing input unit is provided. The dictionary device according to claim 1, further comprising: A notification type setting means for setting a type of change history to be notified for each user;
Based on the type information set in the notification type setting means, it further has a change history notifying means for taking out a change history to be notified to each user from the dictionary history recording means and notifying the browsing input means of each user. The dictionary device according to claim 1, wherein: The dictionary device according to any one of claims 1 to 4,
A natural language processing system comprising: a natural language processing main body that performs natural language processing using the stored contents of the dictionary device. Natural language processing result recording means for recording the natural language processing result executed by the natural language processing main body or the updated processing result together with the recording time information;
From the recording time of the natural language processing result recorded in the natural language processing result recording means, so that the update time recorded and managed by the dictionary history recording means reflects the dictionary data in the dictionary device after A natural language processing review unit for causing the natural language processing body to execute all or part of the natural language processing and updating the natural language processing result recorded in the natural language processing result recording unit according to the result. The natural language processing system according to claim 4, wherein 6. The natural language processing system according to claim 4, wherein the natural language processing executed by the natural language processing main body is machine translation processing.

Images