JP2004234582A - Dictionary construction method, system, and screen - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To construct a dictionary with further high utility value by easily constructing a field term dictionary which has required an enormous man-hour for work in the past by comparison of a term extracted from history data of a retrieval function with an original dictionary term, and enabling adaptation of a term actually used at present as a registration candidate of the dictionary by using retrieval history data. <P>SOLUTION: This system comprises a means for extracting and storing a retrieval keyword or other retrieval attribute information from the use history data of the retrieval function; a means for comparing an existing dictionary or term classification data of dictionary with the extracted retrieval keyword and extracting and storing only terms not overlapped; a means for displaying the existing dictionary or the term classification data; a means for narrowing and displaying a dictionary registration candidate from the term data from which duplication is removed; an editing means for associating the narrowed registration candidate term with a term in the existing dictionary or the term classification data; and a means for storing the editing results. <P>COPYRIGHT: (C)2004,JPO&NCIPI

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a dictionary construction method for constructing a dictionary, a dictionary construction system, and a screen device.
[0002]
[Prior art]
In the Internet and document management systems, when the amount of contents such as documents and data files serving as information sources becomes enormous, the labor required for users to obtain necessary information generally increases. Therefore, in order to obtain necessary information with a small amount of time, the computer can understand in advance the technical terms for each field of information that the user wants to use and the relationships between terms such as superordinate and subordinate concepts, aliases, and synonyms. A dictionary database is prepared in this way, and this dictionary database is used for searching, extracting, and classifying information.
[0003]
However, the task of constructing a dictionary database of technical terms as described above has conventionally been performed manually by experts in the field, and there has been a problem that the number of work steps is considerably large depending on the number of words. Specifically, as a general dictionary construction method, a term set obtained by inputting a document in a specialized field and automatically extracting the term into words (morphological analysis process) is manually input. Unnecessary terms were removed and classified.
[0004]
In order to reduce these manual operations, Japanese Patent Application Laid-Open No. H11-296549 describes a user interface for editing a dictionary of concept information, and in particular, a method of listing related concept information candidates using a degree of association. Is described.
[0005]
[Patent Document 1]
JP-A-11-296549 [0006]
[Problems to be solved by the invention]
In the related art, a morphological analysis is performed by using a document in a specific field or a document group as an input, and the obtained term set is classified using the degree of association between terms or the like, or a classification candidate is presented.
[0007]
However, when the input is a document in a specific field, there are the following problems.
[0008]
First, since a document is a set of terms, it is necessary to use morphological analysis as a process for extracting technical terms. However, this generally requires time and effort to remove noise (unnecessary terms), and depends on the size of the document. Tend to increase. A parameter such as a term appearance frequency in a document may be used for this noise removal. However, the relationship between an extremely high or low appearance frequency and the possibility of a technical term cannot be generally described.
[0009]
Further, even if the tendency can be uniformly grasped based on the frequency of appearance of the term, the old term cannot be distinguished from the latest term, and the conventional technology cannot be used from the viewpoint of maintaining the freshness of the term.
[0010]
Therefore, an object of the present invention is to provide a dictionary construction method, system, and screen for reducing the number of dictionary construction steps.
[0011]
[Means for Solving the Problems]
One feature of the present invention is that in a method for constructing a dictionary, the dictionary is constructed from search keywords or search attribute information extracted from search history information.
[0012]
The other features of the present invention are as described in the claims of the present application.
[0013]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to the drawings.
[0014]
The first embodiment is an example in which the searcher 10 (user) uses the search history data at the time of using the search function to support the construction of a term dictionary, and FIG. 1 shows an example thereof.
[0015]
The search function in the present embodiment means a function of searching for information in a file required by a user by inputting a keyword in a search engine such as a file system, a document management system, a mail system, and the Internet.
[0016]
Here, "file" refers to document data created by a word processor or an editor or the like, standard document data on the Internet such as HTML (Hyper Text Markup Language), XML (extensible Markup Language), program data describing software, and program data. One unit of data, such as data, analysis data, image data, and moving image data.
[0017]
The search keyword means a keyword that can be finally digitized as a word, such as a keyboard input or a voice input.
[0018]
First, the user searcher 10 freely searches using the search function 11. Specifically, the user uses an Internet search engine. Thereby, the search history data 12 is obtained. It should be noted that as the number of times the searcher 10 performs a search using the search function 11 increases, the amount of the search history data 12 increases, and a more complete dictionary can be constructed.
[0019]
Therefore, the search keyword extraction / save processing unit 13 extracts the search keyword from the search history data 12 and stores the result as the first registration candidate term data 14 in the database. Next, if a technical term dictionary already exists, the term comparison processing unit 16 performs character string pattern matching of the first registration candidate term data 14 with the term as the pre-edit term data 15, and outputs the term data before edit. The registration candidate terms that do not exist in 15 are extracted, and the result is stored in the database as second registration candidate term data 17.
[0020]
If the pre-edit term data 15 does not exist in the first use, basic term classification data may be created and used as the pre-edit term data 15.
[0021]
This makes it possible to distinguish between the old term and the latest term, and maintain the freshness of the term.
[0022]
Next, the term configuration display / edit processing unit 18 reads the pre-edit term data 15 first, and then reads the second registration candidate term data 17 according to the narrow-down condition specified by the dictionary editor 19.
[0023]
Here, as an example of a method for narrowing the registration candidate terms,
[Notation] [Meaning]
X *: All terms preceded by X are X? ? ? : There is a method of performing character string matching using a regular expression such as X followed by any three characters.
[0024]
The term composition display means that character string term data is generally displayed on a screen in a tree format because it has a classification hierarchy of upper words, lower words, and the like.
[0025]
Thereafter, the term selected from the second registration candidate term data is added below the optimal node (term) in the pre-edit term data 15 displayed in a tree format on the screen by the operation of the dictionary editor. To execute term editing.
[0026]
Finally, after the editing is completed, the edited result is stored in the database as the edited term data 100.
[0027]
In the above description, the pre-edit term data 15 and the post-edit term data 100 are distinguished, but they may be overwritten after the edit as one data.
[0028]
According to the above-described embodiment, it is possible to construct a dictionary with high term freshness while reducing the number of work steps from the search history information.
[0029]
The second embodiment is an example in which the search attribute data is further used to improve the dictionary editing efficiency in the first embodiment, and FIG. 2 shows an example thereof.
[0030]
From the search history data 12 output by the search function 11, the search keyword and search attribute data extraction / save processing unit 20 extracts the search keyword and the search attribute data, and the search keyword is used as the first registration candidate term. As data 21, search attribute data is stored in a database as 23. At this time, the data of 21 and 23 are associated with a term ID or the like.
[0031]
Here, the search attribute data is information on one search operation, such as the date and time and the number of hits when a searcher performs a search using the search function. The search attribute data also includes data that can identify the searcher, for example, the ID (IP address or the like) of the machine used and user information obtained from login information to the system.
[0032]
The term configuration display / edit processing unit 24 first reads the pre-edit term data 15, and then reads the second registration candidate term data 22 according to the narrowing conditions specified by the dictionary editor 19.
[0033]
Here, in order to narrow down the registration candidate terms, in addition to the method of using the regular expression of the first embodiment as a narrowing condition, the narrowing condition is set using search attribute data 23 such as a search date and time, a searcher, and a search hit rate. . Examples of search conditions are as follows.
Search date and time: January 1, 2000-December 31, 2001 Searcher: Taro Yamada Search hits: Less than 10 (or more)
When the organization information of the user can be obtained from the user management information of the system, organization information of a company, department, section, or the like may be specified in the searcher. This facilitates the construction of specialized dictionaries for, for example, “Company A” and “B Design Department”. In addition, by using the number of search hits as a narrowing condition, terms that are not commonly used yet, or conversely, terms that are already generally used can be narrowed down to some extent. When the number of search hits is 0, it may be determined that there is a high possibility that the search keyword is incorrect.
[0034]
The third embodiment is an example of the data structure of the first registration candidate term data 21, the second registration candidate term data 22, and the search attribute data 23 in the second embodiment, and FIG. 3 is an example of this.
[0035]
The table 30 is a data structure of the first registration candidate term data 21 and the second registration candidate term data 22 in the second embodiment, and associates the registration candidate term ID 31 and the registration candidate term 32 as one row.
[0036]
Further, the table 33 has the data structure of the search attribute data 23 in the second embodiment, and associates the registration candidate term ID 34 with the search attribute data such as the search date and time 35 and the searcher 36 as one line. Here, the IP address of the machine is used to identify the searcher 36.
[0037]
With such a data structure, it is easy to add a new search attribute item.
[0038]
The fourth embodiment shows the flow of processing when using the search function and constructing a dictionary in the second embodiment, and FIG. 4 shows an example thereof.
[0039]
When the search function is used, the history in which the searcher 10 has searched for information using the search function is stored as search history data 12 as in process 40.
[0040]
On the other hand, when constructing a dictionary, first, as in process 41, search keywords and search attribute data are extracted from the search history data 12, and the search keywords are stored as first registration candidate term data 21 and search attribute data 23.
[0041]
Next, as in a process 42, the pre-editing term data 15 and the first registered candidate term data 21 are compared to extract a registered candidate term that does not exist in the pre-edited term data 15, and as the second registered candidate term data 22, save.
[0042]
Further, at the time of editing the dictionary, the second registration candidate term data 22 and the pre-edit term data 15 are read and displayed as in processing 43, and the dictionary editor 19 accepts the display and gives an instruction to edit the term. Is performed. After the editing, the editing result is saved as the edited term data 100 as in a process 44.
[0043]
The fifth embodiment shows an example of the dictionary editing screen in the first and second embodiments, and FIG. 5 shows an example thereof.
[0044]
The screen 50 is divided into an area 57 (left side of the screen) for displaying and editing the pre-editing term data 15 in a tree and an area (right side of the screen) for selecting a registration candidate term.
[0045]
In the registration candidate term narrowing condition designation area 51, the narrowing condition by the regular expression of the term shown in the first embodiment and the narrowing condition by the search attribute data shown in the second embodiment are input. After inputting the narrowing-down conditions, a registration candidate narrowing-down process is executed by clicking the registration candidate term narrowing down button 52 with a pointing device such as a mouse.
[0046]
Next, in the display condition specification area 53, display conditions such as the display order of the terms narrowed down above are specified. The display conditions include a simple term list display in descending order and ascending order, as well as a method of displaying in a hierarchical manner by character string pattern matching. After inputting the display conditions, the display reflection process is executed by clicking the display condition reflection button 54 with a pointing device such as a mouse, and the result is displayed in the registration candidate term selection area 55.
[0047]
In the registration candidate term selection area 55, the dictionary editor selects a term 56 to be registered in the dictionary with a pointing device such as a mouse, and drags and drops the term 56 to a term 58 considered to be applicable in the dictionary term editing area 57. As a result, the term 56 is added to the dictionary term editing area 57 in the lower hierarchy of the drop destination term 58.
[0048]
In the dictionary term editing area 57, the terms can be arbitrarily moved by dragging and dropping. If there is an unnecessary term, it can be deleted by selecting the term and then clicking the term delete button 59 with a pointing device such as a mouse. It is.
[0049]
When the dictionary editing operation is finally completed, the dictionary registration button 500 is clicked with a pointing device such as a mouse, and is saved as edited term data.
[0050]
Thereby, when registering the dictionary registration terms, it becomes possible to register the dictionary registration terms in association with the terms in the pre-editing term data 15 (existing dictionary).
[0051]
The sixth embodiment shows another example of the dictionary term editing area screen in the fifth embodiment, and FIG. 6 shows an example thereof.
[0052]
In the registration candidate term selection area 55, when the dictionary editor selects a term 56 to be registered in the dictionary with a pointing device such as a mouse, the term 56 is displayed in the center of the dictionary term editing area 60.
[0053]
At the same time, the degree of character string matching is calculated by character string pattern matching between the term 56 and each term in the term data before editing. For example, when the registration candidate term is “XXZ device”, as an example of the character string matching degree with “X device”,
Matched character detection direction: Number of characters whose back position matched: a = 2 characters Other characters matched: b = 1 character (character string matching degree) = w1 × a + w2 × b
Here, w1 and w2 are weights (numerical values of 0 or more), and generally, w1> w2. The distance from the center of the dictionary term editing area 60 is, for example,
(Distance) = 1 / (character string matching degree)
, There is a method using the reciprocal of the character string matching degree.
[0054]
When the arrangement position of each term is determined in the dictionary term editing area 60 by the above method, terms similar to the term 61 are displayed near the center and terms not similar are displayed far away.
[0055]
The dictionary editor drags and drops the term 62 that is considered to be applicable from the terms displayed near the term 61. As a result, the term 61 is added to the dictionary term editing area in a hierarchy lower than the drop destination term 62.
[0056]
Thereby, when registering the dictionary registration terms, it becomes possible to register the dictionary registration terms in association with the terms having a high degree of matching included in the pre-editing term data 15 (existing dictionary).
[0057]
The seventh embodiment is an example of a system configuration when the present invention is applied to a dictionary construction service, and FIG. 7 is an example of the system.
[0058]
The search engine server 72 for performing the search service generates a search index from the information source 73 via a network 74 such as the Internet. The information searcher 76 accesses the search engine server 72 via the network 74 through the search client 75. Here, the system administrator 70 manages the search engine server 72 through the management client 71.
[0059]
The person who receives the dictionary construction service permits the search history data 78 of the information searcher 76 belonging to the same organization to be provided to the dictionary construction service provider 79.
[0060]
The dictionary construction service provider 79 takes in the search history data 78 and the user data 77 into the dictionary construction support system 700 already described in the present invention, and constructs a technical term dictionary.
[0061]
The service fee may be set by ranking according to the number of words in the dictionary as the final product, the number of information searchers 76, and the like.
[0062]
This makes it possible to provide a smooth dictionary construction service to the customer.
[0063]
As described above, various embodiments have been described. A device for realizing this may be configured as a dedicated device. However, as illustrated in FIG. And a general-purpose computer system including a storage unit for storing input data and programs as a database, an operation unit, and the like, and a display 83 and a general-purpose computer system. It can be realized by a processing program.
[0064]
When the processing program is added to such a general-purpose computer system and realized, the processing program is recorded on a medium such as the magnetic disk 91 illustrated in FIG. 9 or the CD-ROM 101 illustrated in FIG. , Stored, mounted, read by a magnetic disk reader or a CD-ROM reader provided in the computer main body 82, and taken into the computer main body 82. When a processing program delivered via a communication network is implemented by input means, the captured processing program is stored and stored in a medium such as a magnetic disk so that it can be used repeatedly.
[0065]
【The invention's effect】
ADVANTAGE OF THE INVENTION According to this invention, the dictionary construction method, system, and screen which reduce the dictionary construction man-hour can be provided.
[Brief description of the drawings]
FIG. 1 is an example of a functional block diagram for implementing a term dictionary construction support using search history data when a user uses a search function in an embodiment of the present invention.
FIG. 2 is an example of a functional block diagram for realizing improvement in dictionary editing efficiency using search attribute data in FIG.
FIG. 3 is an example of a data structure diagram of a database in FIGS. 1 and 2;
FIG. 4 is an example of a flowchart showing a processing flow when a search function is used and a dictionary is constructed in FIG. 2;
FIG. 5 is an example of a dictionary editing screen in the first and second embodiments.
6 is an example of another screen of the dictionary term editing area in FIG. 5;
FIG. 7 is a system configuration example when the present invention is applied to a dictionary construction service.
FIG. 8 is an example of a computer system.
FIG. 9 is an example of a magnetic disk.
FIG. 10 shows an example of a CD-ROM.
[Explanation of symbols]
10, 36 ... searcher, 11 ... search function, 12, 78 ... search history data, 13 ... search keyword extraction and storage processing unit, 14, 21 ... first registration candidate term data, 15 ... term data before editing, 16 ... Term comparison processing unit, 17, 22: second registration candidate term data, 18, 24: Term configuration display / edit processing unit, 19: dictionary editor, 20: search keyword, search attribute data extraction / storage processing unit, 23 ... Search attribute data, 30, 33 ... table, 31, 34 ... registration candidate term ID, 32, 56, 61 ... registration candidate term, 35 ... search date and time, 40 ... processing steps at the time of search in Figs. 42, 43, 44: processing steps for dictionary construction, 50: screen, 51: registration candidate term narrowing condition designation area, 52: registration candidate term narrowing button, 53: display condition designation area, 54: display Item reflection button, 55: registration candidate term selection area, 57, 60: dictionary term editing area, 58, 62: registration destination term (parent), 59: term deletion button, 70: system administrator, 71: management client, 72 search server, 73 information source, 74 network, 75 search client, 76 information searcher, 77 user data, 79 dictionary service provider, 81 keyboard, 82 computer 83, display, 91, magnetic disk, 100, edited term data, 101, CD-ROM, 500, dictionary registration button, 700, dictionary construction support system.

Claims (14)

In the method of building a dictionary,
A dictionary construction method for constructing a dictionary from search keywords or search attribute information extracted from search history information. In claim 1,
Inputting the search history information, extracting a first search term, and storing the first search term in the first registration candidate term data;
Comparing the pre-edited term data with the first registered candidate term data, extracting a second search term that is not included in the pre-edited term data from the first registered candidate term data, Processing to save as
Associating the second search term with a term included in the pre-edited term data;
Adding the associated second search term as the edited term data. In claim 2,
A dictionary construction method comprising a process of displaying the pre-edit term data hierarchically. In claim 2,
A dictionary construction method, wherein the pre-edit term data and the post-edit term data are the same data. The process of associating the second search term with a term included in the pre-edited term data according to claim 2,
A process of selecting a registration candidate term of a registration candidate from the second search terms;
Searching the registered candidate terms from the pre-edited term data and presenting them in the order of terms having the highest similarity. 6. The dictionary construction method according to claim 5, wherein searching the registered candidate term from the pre-edit term data includes searching using a regular expression or search attribute data. The process of associating the second search term with a term included in the pre-edited term data according to claim 2,
A registration candidate term selected from the second search terms is arranged and displayed at the center of the display area, and a match between the selected registration candidate term and each term included in the pre-edit term data is displayed. Calculating a degree, and displaying each term included in the pre-editing term data at a position closer to the center as the degree of matching is higher and closer to the center as the degree of matching is lower. Dictionary construction method. The process of associating the second search term with a term included in the pre-edited term data according to claim 2,
A process of inputting a narrowing condition in order to narrow down the second search term;
Searching for and presenting a third search term that satisfies the narrowing condition;
Selecting a registration candidate term from the third search term. The processing for narrowing down the second search term according to claim 8 is:
A process of inputting a condition for narrowing registration candidate terms using search attribute data or a regular expression;
Searching for and presenting a third search term that satisfies the narrowing condition;
Selecting a registration candidate term from the third search term. A program for causing a computer to execute the dictionary construction process according to claim 2. A computer-readable storage medium storing a program for causing a computer to execute the dictionary construction process according to claim 2. In order to narrow the registration candidate terms from the search history data, a part for inputting conditions for narrowing the registration candidate terms,
A screen displaying the narrowed registration candidate terms. In claim 12,
A screen for inputting display conditions for displaying the registration candidate terms. An apparatus that receives search history information as input, extracts a first search term, and stores the first search term in first registered candidate term data;
Comparing the pre-edited term data with the first registered candidate term data, extracting a second search term that is not included in the pre-edited term data from the first registered candidate term data, A device for storing in
An apparatus for associating the second search term with a term included in the pre-edited term data,
A device for adding the associated second term to the edited term data.

Images