JP2004185306A - Dictionary construction supporting device and method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a supporting environment for allowing a user to prepare/verify a dictionary for information extraction where the user can set whether morphemic analysis is to be performed in collating expressions. <P>SOLUTION: A dictionary construction supporting program 100 stored in a storage device 50 is read and stored into a memory 40, and executed by a CPU 10 so that a dictionary construction supporting device can be realized. When a dictionary verifying part 104 extracts the expressions of each concept stored in a dictionary 109 for information extraction from given document data, a condition setting part 106 sets whether morphemic analysis is to be performed to the document data for each concept, and stores the setting in the dictionary 109 for information extraction as a collation condition. <P>COPYRIGHT: (C)2004,JPO&NCIPI

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a dictionary construction support apparatus and method for enabling a user to construct an electronic dictionary used in a document classification system for automatically classifying document data.
[0002]
[Prior art]
A variety of electronic data is used every day in various places, and it is said that more than 80% of the data is document data. Various electronic dictionaries are used for the document data depending on the application. For example, a kana-kanji conversion dictionary is used when creating Japanese document data, a morphological analysis dictionary is used when performing morphological analysis on Japanese document data, and a morphological analysis dictionary is used when performing syntax analysis on document data. A parsing dictionary is used. In such electronic dictionaries, there are basically invariable rules that are clearly defined, and are created and supplied in advance by the product maker. Is used without any particular awareness.
[0003]
In recent years, a text mining technique has attracted attention as a technique for extracting information necessary for a user from document data described in a free format, among various kinds of document data.
[0004]
The applicant of the present application has proposed a method of text mining technology in Patent Document 1. This involves extracting important information contained in the target document data using an information extraction dictionary that holds the concept to be extracted and one or more various expressions representing the concept in association with each other. A method of analyzing document data has been proposed. The expression referred to here means, for example, a word, a phrase, a set of words having a co-occurrence relation, and the like, and the concept referred to herein is a higher order common to each of such expressions. Corresponds to the meaning of For example, “owner”, “trainer”, “buyer”, etc. are individual “expressions”, and “person” is a “concept” having a higher meaning such as “owner”, “trainer”, “buyer”, etc. It is.
[0005]
The correspondence between concepts and expressions in such an information extraction dictionary does not have a clearly defined invariant rule, but depends on the field of document data to be analyzed and the viewpoint of analysis. Things.
[0006]
In these information extraction dictionaries, similarly to the electronic dictionary described above, the product manufacturer investigates the usage form of the user who introduces the text mining technology, and creates and provides the information extraction dictionary in accordance with the survey. At present, the results after use are verified, and maintenance such as registration, deletion, and correction in the information extraction dictionary is performed.
[0007]
As described above, the conventional information extraction dictionary used for the text mining technique is prepared in advance by a manufacturer other than the user, despite the fact that the quality of the dictionary greatly affects the extraction accuracy. No means has been provided for the user to create the dictionary or to edit the dictionary while verifying the created dictionary. Although it was possible to edit the dictionary for information extraction using an editor or the like, it was necessary to fully understand the structure of the dictionary and to edit at the program level, and it could be easily edited by ordinary users. It was not something.
[0008]
In the maintenance of these information extraction dictionaries, the product manufacturer investigates the user's problems after use, judges from the experience, etc., the parts that are considered to be resolved, and adds them to the information extraction dictionaries. There was no choice but to improve the accuracy of text mining by deleting or modifying.
[0009]
In view of such points, the applicant of the present application has proposed a method and an apparatus in Patent Literature 2 that facilitate creation and maintenance of an information extraction dictionary used in text mining.
[0010]
In addition, in view of the fact that a concept to be extracted may not require a morphological analysis such as a proper noun, the applicant of the present application disclosed in Patent Literature 3 when collating an expression in an electronic dictionary with document data is set in advance. We have proposed an information extraction system characterized by comprising an expression extracting means for extracting expressions by switching between collation after performing morphological analysis and collation without performing morphological analysis according to conditions. According to this, matching conditions can be set for each expression or concept to be extracted. For example, if you want to extract an expression that becomes a part of another word only by string matching, you can perform morphological analysis to correctly recognize word boundaries, so the extraction accuracy is higher than without morphological analysis . As a specific example, if you want to extract "Kyoto", it will match "Kyoto" in the character string "Tokyo" unless you perform morphological analysis. Can be prevented.
[0011]
Conversely, for example, when the morphological analysis fails and becomes an unknown word, there is a high possibility that the word boundary is erroneous. As a specific example, if a word is registered in the dictionary for extraction as one word in an attempt to extract “Kochijan”, if the word is divided into two words such as “Kochi / Jan” by morphological analysis, morphological analysis is performed. Will not match as it is.
[0012]
In general, if you want to extract product names or symbol strings that are not registered in the morphological analysis dictionary, it is more appropriate not to perform morphological analysis, and extract general words that are registered in the morphological analysis dictionary. It can be said that it is more appropriate to perform morphological analysis.
[0013]
[Patent Document 1]
JP 2001-147937 A
[0014]
[Patent Document 2]
JP-A-2002-140338
[0015]
[Patent Document 3]
Japanese Patent Application No. 2002-278421
[0016]
[Problems to be solved by the invention]
However, in order to set the above conditions, it was necessary to edit the information extraction dictionary using a text editor or the like. The construction of an electronic dictionary without morphological analysis does not require knowledge of computational linguistics, so it is logically possible for ordinary users, but editing work at the program level is required. Therefore, editing work at the program level is actually required, and the editing work is extremely difficult for ordinary users.
[0017]
The present invention has been made in view of such circumstances, and provides a support environment for creating and verifying an information extraction dictionary that allows a user to set whether or not to perform morphological analysis in expression matching. The purpose is to:
[0018]
[Means for Solving the Problems]
Therefore, the dictionary construction support apparatus of the present invention provides a dictionary storage unit that stores an electronic dictionary that stores a plurality of expressions and a concept that is a higher-level expression common to those expressions, and that is extracted from the document data. Expression storage means for storing expressions, at least some concepts from the electronic dictionary stored in the dictionary storage means, and at least some of the extracted expressions stored in the expression storage means simultaneously Display means for displaying, designation of one or more expressions from the expressions displayed by the display means, and designation of one concept from the concepts displayed by the display means, the designated expression is designated. In addition to a registration unit for additionally registering the concept in the electronic dictionary in association with the concept, a unit for designating whether or not to perform morphological analysis at the time of matching for each concept and storing it in the electronic dictionary is provided.
[0019]
Further, when displaying the extraction result, a means is provided for distinguishing and displaying those that have been collated using morphological analysis and those that have not.
[0020]
This makes it possible to register expressions extracted from document data to desired concepts in the electronic dictionary with a simple operation, and to easily verify the quality of the electronic dictionary, including whether or not morphological analysis should be performed. Is now available.
[0021]
That is, while referring to the important expressions extracted from the document, necessary expressions can be selected and registered in the dictionary for information processing, and at the same time, whether or not to perform morphological analysis can be specified. The dictionary can be edited while verifying the performance of the dictionary by viewing the information processing results.
[0022]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to the drawings.
[0023]
FIG. 1 is a diagram showing a configuration of a dictionary construction support device according to an embodiment of the present invention.
[0024]
The dictionary construction support apparatus according to the present embodiment is configured such that, for example, in a computer such as a personal computer, a dictionary construction support program 100 stored in a storage device 50 is read into a memory 40 and executed by a CPU 10 that controls the entire system. Is achieved.
[0025]
The input unit 20 receives, from a mouse, a keyboard, a voice input device, or the like, an editing instruction such as insertion or deletion of a character string, an operation instruction for selecting a function, and a command input such as designation of a document or a dictionary to be processed. Is what you get. The output unit 30 is for supplying display information to a display device such as a display.
[0026]
The memory 40 is composed of a high-speed and volatile DRAM, for example, and stores the program 100 and the like executed by the CPU 10 as described above, and temporarily holds the internal memory when the program 100 is executed. It is used as the data holding unit 110.
[0027]
The internal data holding unit 110 includes a text buffer 110a, an expression list buffer 110b, a processing result buffer 110c, a dictionary buffer 110d, and a difference buffer 110e.
[0028]
The storage device 50 is a nonvolatile large-capacity storage device, and can be realized by, for example, an HDD, a CD-ROM, a DVD-ROM, or the like. The storage device 50 includes a dictionary construction support program 100 including a control unit 101, an expression registration unit 102 including an expression extraction unit 102a, a dictionary editing unit 103, a dictionary verification unit 104, a difference detection unit 105, and a condition setting unit 106; An information extraction dictionary 109 constructed by this program is stored.
[0029]
The control unit 101 is a main program that is executed first when the program 100 is started.
[0030]
The expression registration unit 102 is a program for registering expressions in the information extraction dictionary 109. The expression registration unit 102 is activated by a predetermined instruction from the control unit 102, and performs processing for registering “expressions” described later in the information extraction dictionary 109.
[0031]
The expression extraction unit 102a is a program that is started when the expression registration unit 102 needs to perform an extraction process of the expression of the document data. The expression extraction unit 102a performs syntax analysis and the like on the content stored in the text buffer 110a, and lists words and phrases used in the document or co-occurrence relationships (hereinafter, these are referred to as expressions). The data is stored in the expression list buffer 110b. The list is collectively called an expression list. The method of creating this expression list is disclosed in, for example, JP-A-62-99865 and JP-A-2-42572, and the description is omitted.
[0032]
The expression registration unit 102 registers the expression specified by the user from the created expression list in the dictionary buffer 110d.
[0033]
The dictionary editing unit 103 is a program for editing the information extraction dictionary 109. The dictionary editing unit 103 can also be activated by the expression registration unit 102 and the dictionary verification unit 104.
[0034]
The dictionary verification unit 104 performs an extraction process specified by the user using the dictionary buffer 110d read and stored from the information extraction dictionary 109, and stores the extraction processing result in the processing result buffer 110c. Further, based on a user's instruction, information stored in the processing result buffer 110c is displayed.
[0035]
The difference detection unit 105 detects a difference between the two information extraction dictionaries 109.
[0036]
The condition setting unit 106 sets a collation condition for extracting information. The set conditions are stored in the information extraction dictionary 109 through the control unit 101. The stored conditions are read out to the dictionary buffer 110d to control the movement of information extraction in the dictionary verification unit 104, and to determine the movement of information extraction in an external information extraction system using the present information extraction dictionary. Control is performed as in Reference 3.
[0037]
The information extraction dictionary 109 is an electronic dictionary constructed by this program. FIG. 2 shows an example of the configuration of the information extraction dictionary 109. The information extraction dictionary 109 is composed of three layers. The highest layer is called a class, the middle layer is called a concept, and the lower layers are called expressions. In the case where one of these is referred to without any particular awareness of the class, concept, and expression hierarchy, it is simply referred to as a node hereinafter.
[0038]
The class table 121 indicates a class, and stores “place name”, “person name”, and the like. The concept table 122 stores classes and concepts in association with each other. The expression table 123 stores classes, concepts, and expressions in association with each other.
[0039]
The above is the configuration of the dictionary construction support device of the present embodiment.
[0040]
Next, the operation of the dictionary construction support apparatus of the present embodiment will be described with reference to a flowchart, taking information extraction as an example.
[0041]
FIG. 3 shows a flowchart of the control unit 101 in the dictionary construction program 100.
[0042]
When the user requests activation of the dictionary construction program 100, the control unit 101 is activated.
[0043]
After starting the own program, the control unit 101 waits for a key input from the user (S201). When a key input is received, first, it is determined whether or not the user input is an end instruction (S202). If it is determined that the instruction is an end instruction, the own program is ended.
[0044]
On the other hand, if it is determined in step S202 that the input is not an end instruction, it is next determined whether or not the user input is a document designation (S204). If the document is designated, the designated document data is read from the designated destination, for example, a memory, a magnetic disk, an optical disk, or the like, and stored in the text buffer 110a in the internal data holding unit 110 of the memory 40 ( S204). When this process ends, the process returns to the step S201 and waits for the next input from the user.
[0045]
On the other hand, if it is determined in step S203 that the input is not a document specification, it is next determined whether or not the user input is a dictionary specification (S205). If the dictionary is specified, the specified dictionary is read from the specified destination, for example, a memory, a magnetic disk, an optical disk, or the like, and stored in the dictionary buffer 110d in the internal data holding unit 110 of the memory 40 (S206). ). When this process ends, the process returns to the step S201 and waits for the next input from the user.
[0046]
On the other hand, if it is determined in step S205 that the dictionary is not specified, it is determined whether the user input is an expression registration instruction (S207). If it is determined that the instruction is an expression registration instruction, the expression registration unit 102 is activated (S208). The operation of the expression registration unit 102 will be described later. When the operation of the expression registration unit 102 ends, the process returns to step S201, and waits for the next input from the user.
[0047]
On the other hand, if it is determined in step S207 that the input is not an expression registration instruction, it is next determined whether or not the user input is a dictionary verification instruction (S209). If it is determined that the instruction is a dictionary verification instruction, the dictionary verification unit 104 is activated (S210). The operation of the dictionary verification unit 104 will be described later. When the operation of the dictionary verification unit 104 ends, the process returns to step S201, and waits for the next input from the user.
[0048]
On the other hand, if it is determined in step S209 that the input is not a dictionary verification instruction, then it is determined whether the user input is a dictionary edit instruction (S211). If the input is a dictionary editing instruction, the dictionary editing unit 103 is activated (S212). The operation of the dictionary editing unit 103 will be described later. When the operation of the dictionary editing unit 103 ends, the process returns to step S201, and waits for the next input from the user.
[0049]
On the other hand, if it is determined in step S211 that the input is not a dictionary editing instruction, then it is determined whether the user input is a difference detection instruction (S213). If the instruction is a difference detection instruction, the difference detection unit 105 is activated (S214). The operation of the difference detection unit 105 will be described later. When the operation of the difference detection unit 105 ends, the process returns to step S201, and waits for the next input from the user.
[0050]
On the other hand, if it is determined in step S213 that the received instruction is not a difference detection instruction, another designated process, for example, a process corresponding to a user input instruction such as an environment setting process is performed (S215), and the process returns to step S201. Wait for the next input from the user.
[0051]
As described above, the control unit 101 analyzes the input of the user, activates various processing units, and stores the processing target document and the dictionary in the internal data holding unit 110 based on the user's instruction.
[0052]
FIG. 4 shows a screen display example immediately after the control unit 101 is started. The upper part has buttons for registration, verification, editing, difference display, environment setting, and termination. The user selects these processings by pressing these buttons. For example, when the end button is clicked, it is determined in step S202 that the user's input is an end instruction, and end processing is performed. Further, for example, when the registration button is clicked, it is determined in step S207 that the user input is an expression registration instruction, and a process of activating the expression registration unit 102 is performed.
[0053]
Below these buttons, an area for inputting a dictionary name is provided on the left, and an area for inputting a document name is provided on the right. The user inputs a dictionary name and a document name in these areas and executes the The corresponding dictionary or document is read from the storage device 50. For example, when the user inputs a document name in an area for inputting a document name and executes the process, in step S203, it is determined that the user input is a document designation, and in step S204, based on the input document name, The relevant document is read from the storage device 50 or the like and stored in the text buffer 110a.
[0054]
Next, the processing operation of the expression registration unit 102 will be described. FIG. 5 is a flowchart showing the processing operation of the expression registration unit 102.
[0055]
If it is determined in step 207 that the instruction is an expression registration instruction during the operation of the control unit 101, the expression registration unit 102 is activated. The activated expression registration unit 102 first activates the expression extraction unit 102a. The activated expression extracting unit 102a analyzes the document data stored in the text buffer 110a by a predetermined analysis method, and creates an expression list (S301). The created expression list is stored in the expression list buffer 110b (S302). The predetermined analysis method may be, for example, when creating an expression list of words, morphological analysis or syntax analysis of document data, extract words, and list the extracted words. What is necessary is just to create by various well-known methods, without being limited to. After the extraction, the expression extraction unit 102a ends.
[0056]
Next, the dictionary data stored in the dictionary buffer 110d and the expression list stored in the expression list buffer 110b are read out and displayed simultaneously (S303).
[0057]
Next, it is determined whether or not the user input is a dictionary registration instruction (S304). If the instruction is a dictionary registration instruction, the user specifies an expression to be registered in the dictionary from the displayed expression list (S305). Thus, the expression registration unit 102 obtains the expression specified by the user. Further, the user designates a concept in which the obtained expression is to be registered from the displayed dictionary concepts (S306). Thereby, the expression registration unit 102 obtains the concept in which the expression is registered. Further, the user gives to the condition setting unit 106 whether or not the concept should be subjected to morphological analysis at the time of information extraction, using an input unit such as a check box (S307). Then, the obtained expressions, concepts, and conditions for information extraction are synthesized and registered as dictionary information (lower-level table) of the expressions (S308). This registration is added to the dictionary buffer 110d.
[0058]
On the other hand, if it is not a dictionary registration instruction in step S304, it is determined whether or not the user input is a dictionary editing instruction (S309). If the instruction is a dictionary editing instruction, the dictionary editing unit 103 described later is activated to edit the dictionary (S310). When the processing of the dictionary editing unit 103 ends, the process returns to step S304.
[0059]
On the other hand, if it is not a dictionary editing instruction in step S309, the information stored in the dictionary buffer 110d is stored on a magnetic disk or the like (S311), and the process ends.
[0060]
In this way, the expression registration unit 102 activates the expression extraction unit 102a to create an expression list, and registers a dictionary in the dictionary 109 from the expression list.
[0061]
FIG. 6 shows an example of a screen display at the time of processing of the expression registration unit 102. A registration button is provided below the area for inputting the dictionary name. By clicking this button, dictionary registration processing (S305 to S308 in FIG. 5) becomes possible. The contents of the dictionary are displayed below the registration button. In this example, a part of the contents of the dictionary with the dictionary name “XXXXXXXXX” is displayed. In this display, the information is displayed in a three-layered format of a class, a concept, and an expression. Also, a check box is provided at the front of each word, and the user can specify the check box by clicking the check box. The concept has a check box (a1, a2,...) For designating whether to use morphological analysis as a collation condition at the time of information extraction, and the condition can be input by the user. One of the features of this dictionary construction support device is that the check boxes (a1, a2,...) Are provided so that whether or not to perform morphological analysis can be designated. This eliminates the need for editing work, and allows ordinary users to easily perform editing work. In this example, the check box (a1) for the concept “work” is off and the checkpoint (a2) for the concept “person” is on. No analysis is used. On the other hand, morphological analysis is used to extract expressions included in the concept “person”.
[0062]
On the other hand, below the area for inputting the document name, there are buttons for word, co-occurrence, and phrase. By pressing these buttons, the display contents of the expression list can be selected. The lower part of these buttons is an area for displaying an expression list. In this example, an expression list of a word having a document name “++++++” is displayed. A check box is provided in front of each word (word in this case) in the expression list, and can be specified by the user by clicking the check box.
[0063]
As described above, the screen at the time of processing of the expression registration unit 102 can simultaneously display the dictionary and the expression list, and the dictionary class, concept, expression, and each word of the expression list can be easily specified. This makes it possible for the user to easily register in the dictionary.
[0064]
FIG. 7 is a diagram schematically illustrating expression registration processing. This example shows processing of internal data when "employee" is selected from the expression list and added as one expression of the concept of "person".
[0065]
When "employee" is selected from the displayed expression list, "employee" is obtained from the expression list buffer 110b, and when the concept "person" is selected from the displayed dictionary, the concept stored in the dictionary buffer 110d. “Industry / person” is acquired from the table 122. Then, the obtained "industry, person" and "employee" are combined to obtain "industry / person / employee", which is registered in the expression table 123 of the dictionary buffer 110d. As a result, the expression table becomes like 123 '.
[0066]
Thereby, the expression extracted from the document data can be registered to a desired concept in the electronic dictionary by a simple operation.
[0067]
Next, the processing operation of the dictionary verification unit 104 will be described. FIG. 8 is a flowchart showing the processing operation of the dictionary verification unit 104.
[0068]
If it is determined in step 209 that the instruction is a dictionary verification instruction during the operation of the control unit 101, the dictionary verification unit 104 is activated. The activated dictionary verification unit 104 displays the dictionary data stored in the dictionary buffer 110d (S401).
[0069]
Next, information is extracted using the dictionary 109, and the result is stored in the processing result buffer 110c (S402). Then, the information stored in the processing result buffer 110d is displayed (S403).
[0070]
Next, it is determined whether or not the user input is a verification instruction of the information extraction result (S404).
[0071]
If it is determined that the instruction is a verification instruction of the information extraction result, the node to be verified is specified in the dictionary (S405). Then, the information extraction result corresponding to the designated node is displayed (S406). After this processing, the process returns to step S404.
[0072]
On the other hand, if it is determined in step S404 that the instruction is not a verification instruction of the information extraction result, it is next determined whether or not the user input is a dictionary editing instruction (S407). If it is determined that the instruction is a dictionary editing instruction, the dictionary editing unit 103 is activated to perform dictionary editing. When the dictionary editing unit 103 ends, the process returns to step S404.
[0073]
On the other hand, if it is determined in step S407 that the instruction is not a dictionary editing instruction, the information stored in the dictionary buffer 110d is stored on a magnetic disk or the like (S409). In addition, the information stored in the processing result buffer 110c is stored on a magnetic disk or the like (S410), and the processing of the dictionary verification unit 104 ends.
[0074]
In this way, the dictionary verification unit 104 verifies the dictionary while displaying the information extraction result by the dictionary 109.
[0075]
FIG. 9 shows an example of a screen display at the time of processing by the dictionary verification unit 104.
[0076]
On the display, the display content of the left half is the same as the display content of the left half of the expression registration unit 102 in FIG. 6 except that there is only a verification button instead of the registration button.
[0077]
By pressing this verification button, the result verification processing (S405 to S406 in FIG. 8) becomes possible.
[0078]
On the other hand, in the display content in the right half of FIG. 9, the extraction result processed by the dictionary 109 is displayed below the document name. Then, as shown in the figure, after verification of the extraction result, underlines (b1, b2, b3, b4, b4 …). It should be noted that any method other than underlining, for example, may be used as long as it can sufficiently identify colors and the like.
[0079]
Furthermore, at this time, depending on whether or not morphological analysis was used for information extraction, by changing the display method such as changing the shape of the underline or changing the color, it is possible to simultaneously display the conditions for information extraction. it can. One of the features of this dictionary construction support device is that the display method is changed depending on whether or not a morphological analysis is used for extracting information at a location where an expression of the designated concept exists.
[0080]
In this way, the concepts in the hierarchical electronic dictionary can be specified by a simple operation, and the expressions included in the specified concept can be easily extracted and specified from the extraction results extracted by the electronic dictionary. Since the matching condition is also displayed at the same time, the user can easily verify the quality of the electronic dictionary.
[0081]
In this example, the underline (b1, b2, b3) added to "service", "development", and "sales" is a single line, whereas the underline (b4) added to "buyer" is a double line. However, this indicates that “service”, “development”, and “sales” do not use morphological analysis in extracting information, while “buyers” indicate that morphological analysis is used. I have. In other words, the concept “work” including “service”, “development” and “sales” shown in FIG. 6 is not used for morphological analysis, and the concept “person” including “buyer” is not analyzed by morphological analysis. The matching condition at the time of information extraction set to use is easily recognized by the underline.
[0082]
Next, the processing operation of the dictionary editing unit 103 will be described. FIG. 10 is a flowchart illustrating the processing operation of the dictionary editing unit 103.
[0083]
During the operation of the control unit 101, when it is determined that the instruction is a dictionary editing instruction in step S 211, during the operation of the expression registration unit 102, when it is determined that the instruction is a dictionary editing instruction in step S 308, and during the operation of the dictionary verification unit 104, If it is determined in step S407 that the instruction is a dictionary editing instruction, the dictionary editing unit 103 is activated.
[0084]
The activated dictionary editing unit 103 displays the dictionary data stored in the dictionary buffer 110d (S501).
[0085]
Next, it is determined whether or not the user input is a node addition instruction (S502). If the instruction is an addition instruction, an additional node is specified (S503). Then, a child node is added to the designated node (S504). Additional content is entered directly by the user. Then, the process returns to step S502.
[0086]
On the other hand, if it is not an addition instruction in step S502, it is next determined whether or not the user input is an instruction to delete a node (S505). If the instruction is a deletion instruction, a deletion node is designated (S506). Then, the designated node and all child nodes of the node are deleted (S507). Then, the process returns to step S502.
[0087]
On the other hand, if it is not a deletion instruction in step S505, it is next determined whether or not the user input is a node change instruction (S508). Here, if it is a change instruction, a change node is designated (S509). Then, the character string and value of the designated node are changed (S510). Then, the process returns to step S502.
[0088]
On the other hand, if it is not a change instruction in step S508, then it is determined whether or not the user input is a node copy instruction (S511). If the instruction is a copy instruction, a copy source node is designated (S512). Then, a copy destination node is designated (S513). Then, the specified copy source node and all of its child nodes are added to the child nodes of the copy destination node (S514). Then, the process returns to step S502.
[0089]
On the other hand, if it is not a copy instruction in step S511, it is determined whether or not the user input is a node move instruction (S515). Here, in the case of a move instruction, a move source node is designated (S516). Then, the destination node is specified (S517). Then, the designated source node and all of its child nodes are moved to child nodes of the destination node (S518). Then, the process returns to step S502.
[0090]
On the other hand, if it is not a move instruction in step S515, the information stored in the dictionary buffer 110d is stored on a magnetic disk or the like (S519), and the process ends.
[0091]
In this way, the dictionary editing unit 108 performs addition, deletion, change, copying, movement, and the like on the nodes of the dictionary 109 to perform dictionary editing.
[0092]
FIG. 11 shows an example of a screen display at the time of processing by the dictionary editing unit 103.
[0093]
This example is different from the display content of the left half of the expression registration unit 102 in FIG. 6 in that add, delete, change, move, and copy buttons are arranged in place of the registration button. Is the same. By clicking each button, various editing functions of the clicked button are performed as described above.
[0094]
As described above, since the editing of the electronic dictionary is made to be a user interface which is easy to handle, the dictionary editing can be easily realized for the user.
[0095]
Next, the processing operation of the difference detection unit 105 will be described. FIG. 12 is a flowchart showing the processing operation of the difference verification unit 105.
[0096]
During the operation of the control unit 101, if it is determined in step S213 that the instruction is a difference detection instruction, the difference detection unit 105 is activated.
[0097]
The activated difference detection unit 105 specifies two dictionaries to be compared according to the user's input (S601). Then, a difference of the designated dictionary is created and stored in the difference buffer 110e (S602).
[0098]
Next, the difference stored in the difference buffer 110e is displayed (S603). The difference stored in the difference buffer 110e is stored on a magnetic disk or the like (S604), and the process ends.
[0099]
In this way, the difference detection unit 105 creates and displays a difference by comparing the dictionaries 109 with each other.
[0100]
FIG. 13 shows an example of a screen display at the time of processing of the difference detection unit 105.
[0101]
In this example, the upper part is provided with two areas for inputting a dictionary to be compared. The lower part of this area is an area for displaying the comparison result. Here, the comparison result for each concept is displayed.
[0102]
This makes it possible to easily designate two electronic dictionaries, detect a difference between the electronic dictionaries, and present the difference to the user in units of concepts.
[0103]
In the present embodiment described above, an expression extracted from document data can be registered in a desired concept in the electronic dictionary by a simple operation.
[0104]
In addition, the concept in the hierarchical electronic dictionary can be specified by a simple operation, and the expressions included in the specified concept can be easily extracted from the extraction results extracted by the electronic dictionary and specified. Therefore, the user can easily verify the quality of the electronic dictionary.
[0105]
Further, two electronic dictionaries can be easily specified, and a difference between the electronic dictionaries can be detected and presented to the user in units of concepts.
[0106]
In addition, it is possible to select necessary expressions and register them in the dictionary for information extraction while referring to important expressions extracted from the document. You can now edit it.
[0107]
It should be noted that the present invention is not limited to the above-described embodiment, and can be variously modified in an implementation stage without departing from the scope of the invention. Furthermore, the embodiments include inventions at various stages, and various inventions can be extracted by appropriately combining a plurality of disclosed constituent elements. For example, even if some components are deleted from all the components shown in the embodiment, the problem described in the column of the problem to be solved by the invention can be solved, and the effects described in the column of the effect of the invention can be solved. Is obtained, a configuration from which this configuration requirement is deleted can be extracted as an invention.
[0108]
【The invention's effect】
As described above, according to the present invention, it is possible to provide a support environment for creating and verifying an information extraction dictionary in which a user can set whether or not to perform morphological analysis in expression matching.
[Brief description of the drawings]
FIG. 1 is a diagram showing a configuration of a dictionary construction support device according to an embodiment of the present invention.
FIG. 2 is an example of a configuration of an information extraction dictionary 109 according to the embodiment.
FIG. 3 is a flowchart showing a processing operation of a control unit 101 in the embodiment.
FIG. 4 is a screen display example immediately after the control unit 101 according to the embodiment is started.
FIG. 5 is a flowchart showing a processing operation of the expression registration unit 102 in the embodiment.
FIG. 6 is a screen display example during processing of the expression registration unit 102 in the embodiment.
FIG. 7 is an exemplary view schematically showing registration processing of expressions in the embodiment.
FIG. 8 is a flowchart showing a processing operation of the dictionary verification unit 104 according to the embodiment.
FIG. 9 is an example of a screen display during processing of the dictionary verification unit 104 according to the embodiment.
FIG. 10 is an exemplary flowchart illustrating the processing operation of the dictionary editing unit 103 according to the embodiment.
FIG. 11 is a screen display example during processing of the dictionary editing unit 103 according to the embodiment.
FIG. 12 is a flowchart showing a processing operation of a difference detection unit 105 in the embodiment.
FIG. 13 is a screen display example during processing of the difference detection unit 105 in the embodiment.
[Explanation of symbols]
10 ... CPU
20 ... input section
30 ... Output unit
40 ... Memory
50 ... Storage device
100: Dictionary construction program
101 ... Control unit
102 ... Expression registration unit
102a ... Expression extraction unit
103 ... Dictionary editor
104: Dictionary verification unit
105: Difference detection unit
106: Condition setting section
109: Information extraction dictionary
110: Internal data holding unit
110a ... text buffer
110b ... Expression list buffer
110c: Processing result buffer
110d: Dictionary buffer
110e ... difference buffer
121 ... Class table
122 ... Concept table
123 ... Expression table

Claims (12)

Dictionary storage means for storing an electronic dictionary stored in association with a plurality of expressions and concepts which are higher-level expressions common to those expressions,
An expression storage means for storing an expression extracted from the document data;
Display means for simultaneously displaying at least some of the concepts from the electronic dictionary stored in the dictionary storage means and at least some of the extracted expressions stored in the expression storage means;
Upon receiving designation of one or more expressions from the expressions displayed by the display means and designation of one concept from the concepts displayed by the display means, the specified expression is associated with the specified concept. A dictionary construction support device comprising a registration unit for additionally registering in the electronic dictionary,
When matching the expression in the electronic dictionary stored in the dictionary storage unit with the document data, it is determined whether the document data is subjected to morphological analysis and then collated or collated without morphological analysis. A dictionary construction support device comprising a setting means for setting each concept. 2. The dictionary construction support apparatus according to claim 1, wherein said setting means includes means for storing the setting as a collation condition in the electronic dictionary. 3. The dictionary construction support apparatus according to claim 2, wherein said setting means includes means for displaying the setting stored in said electronic dictionary. Dictionary storage means for storing an electronic dictionary stored in association with a plurality of expressions and concepts which are higher-level expressions common to those expressions,
An expression storage means for storing an expression extracted from the document data;
A dictionary comprising display means for simultaneously displaying at least some concepts from the electronic dictionary stored in the dictionary storage means and at least some of the extracted expressions stored in the expression storage means A construction support device,
When displaying a part of the extracted expressions on the display unit, the display unit includes a display control unit that changes a display method according to a difference in a matching condition whether or not to use a morphological analysis used at the time of matching. A dictionary construction support device characterized by the following. Dictionary storage means for storing an electronic dictionary stored in association with a plurality of expressions and concepts which are higher-level expressions common to those expressions,
An expression storage means for storing an expression extracted from the document data;
Display means for simultaneously displaying at least some of the concepts from the electronic dictionary stored in the dictionary storage means and at least some of the extracted expressions stored in the expression storage means;
Upon receiving designation of one or more expressions from the expressions displayed by the display means and designation of one concept from the concepts displayed by the display means, the specified expression is associated with the specified concept. A dictionary construction support method of a dictionary construction support device comprising: a registration unit for additionally registering in the electronic dictionary;
When matching the expression in the electronic dictionary stored in the dictionary storage unit with the document data, it is determined whether the document data is subjected to morphological analysis and then collated or collated without morphological analysis. A dictionary construction supporting method, comprising a step of setting for each concept of: 6. The dictionary construction support method according to claim 5, wherein the setting step includes a step of storing the setting as a collation condition in the electronic dictionary. 3. The dictionary construction support method according to claim 2, wherein the setting step includes a step of displaying the setting stored in the electronic dictionary. Dictionary storage means for storing an electronic dictionary stored in association with a plurality of expressions and concepts which are higher-level expressions common to those expressions,
An expression storage means for storing an expression extracted from the document data;
A dictionary comprising display means for simultaneously displaying at least some concepts from the electronic dictionary stored in the dictionary storage means and at least some of the extracted expressions stored in the expression storage means A dictionary construction support method of a construction support device,
When displaying a part of the extracted expressions on the display unit, a display control step of changing a display method depending on a difference in a matching condition whether or not to use a morphological analysis used at the time of matching is provided. A dictionary construction support method characterized by the following. Dictionary storage means for storing an electronic dictionary stored in association with a plurality of expressions and concepts which are higher-level expressions common to those expressions,
An expression storage means for storing an expression extracted from the document data;
Display means for simultaneously displaying at least some of the concepts from the electronic dictionary stored in the dictionary storage means and at least some of the extracted expressions stored in the expression storage means;
Upon receiving designation of one or more expressions from the expressions displayed by the display means and designation of one concept from the concepts displayed by the display means, the specified expression is associated with the specified concept. A computer comprising a registration unit for additionally registering in the electronic dictionary,
When matching the expression in the electronic dictionary stored in the dictionary storage unit with the document data, it is determined whether the document data is subjected to morphological analysis and then collated or collated without morphological analysis. A program for functioning as setting means for setting each concept. The program according to claim 9, wherein the setting unit includes a unit that stores the setting as a collation condition in the electronic dictionary. The program according to claim 10, wherein the setting unit includes a unit configured to display the setting stored in the electronic dictionary. Dictionary storage means for storing an electronic dictionary stored in association with a plurality of expressions and concepts which are higher-level expressions common to those expressions,
An expression storage means for storing an expression extracted from the document data;
A computer including a display unit for simultaneously displaying at least some of the concepts from the electronic dictionary stored in the dictionary storage unit and at least some of the extracted expressions stored in the expression storage unit To
When displaying a part of the extracted expressions on the display unit, the display unit may function as a display control unit that changes a display method depending on a difference in a matching condition whether or not to use a morphological analysis used at the time of matching. Program.

Images