JP2001167124A - Document classification device and recording medium recording document classification program - Google Patents

Abstract

(57) [Summary] [Problem] To automatically classify a large number of documents into a category that matches topicality or a user's preference without a user consciously determining a category for classification. A document classification device according to the present invention includes a storage unit (not shown) that stores documents of a classification target document and history information, and a keyword extraction unit that extracts a keyword from a document stored in the storage unit. 1 and an analysis for calculating a weight for classifying the keyword based on at least one of the importance of the keyword extracted by the keyword extracting unit 1 and the storage date and time or storage location of the document including the keyword. A classification unit that classifies documents based on the weights calculated by the analysis unit;

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a document classification apparatus, and more particularly to a document classification apparatus for classifying documents from a group of documents in a system for storing information and a recording medium on which a document classification program is recorded.

[0002]

2. Description of the Related Art With the spread of information communication networks such as the Internet, opportunities for classifying document information acquired from information providers and the like as necessary have been increasing.

[0003] As known techniques related to the present invention, Japanese Patent Application Laid-Open Nos. Hei 6-348755 and Hei 11-15848 are known.
There is a document classification method disclosed in Japanese Unexamined Patent Application Publication No. 2000-163,873. The document classification method described in each of the above publications extracts a word (keyword) unique to each field from a group of classified documents, and based on the frequency of appearance of the keyword in the document to be classified,
This determines the classification destination of the document to be classified.

A document classification method disclosed in Japanese Patent Application Laid-Open No. Hei 7-114572 generates a document vector expressing the characteristics of a document from a characteristic vector expressing the characteristics of a word included in a document to be classified, and outputs the document vector. Classification is performed using similarity between vectors.

[0005]

As can be seen from the above publications, in the conventional document classification system, as elements for determining the classification destination of a document to be classified, the document itself such as the document structure of the keyword or the frequency of appearance of the keyword is used. Is not used to automatically classify documents to be classified into a topical category or a category that matches the user's preference.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a document classification apparatus capable of automatically classifying documents to be classified into categories suitable for topicality or user preference, and a recording medium recording a document classification program. It is in.

[0007]

In order to achieve the above object, a document classifying apparatus according to the present invention comprises a storage unit for storing a document, and an extraction unit for extracting a keyword from the document stored in the storage unit. Weight calculating means for calculating a weight of the keyword as a classifying destination of the document based on at least one of the importance of the keyword extracted by the extracting means, a storage date and time or a storage location of a document including the keyword. Classifying means for classifying the document based on the weight calculated by the weight calculating means.

[0008] The weight calculating means changes the weight when the keyword is used in at least one of a heading, a title, a diagram, a hyperlink or highlighting of the document, or a proper noun. This makes it possible to classify documents with emphasis on the importance of keywords.

Further, the weight calculating means calculates a date and time when the keyword appears from the storage date and time of the document including the keyword, and calculates the weight based on a comparison between the calculated date and a predetermined reference date and time. Therefore, the date and time of appearance of the keyword and the like are given as time information, and a topical category can be detected.

[0010] The weight calculating means may be arranged so that the document including the keyword is cached in a predetermined browser, a link destination of a favorite file or bookmark of the browser, a character input string at the time of search, a hierarchical classification structure performed by a user, Alternatively, when the weight is changed in the case of at least one of the automatically classified hierarchical structures, operation history information or the like is added, and a category having a preference can be detected.

Further, the apparatus further comprises control means for controlling a specific gravity with respect to the weight calculated by the weight calculation means, wherein the classifying means includes a specific gravity from the control means and the weight calculated by the calculation means. By classifying the documents based on the above, the documents can be classified in consideration of user needs.

[0012] In another aspect, the present invention provides a computer, comprising: a storage unit for storing a document; an extraction unit for extracting a keyword from the document stored in the storage unit; Weight calculating means for calculating a weight of the keyword as a classification destination of the document based on at least one of the importance, the storage date and time or the storage location of the document including the keyword, and the weight calculated by the weight calculating means. A computer-readable recording medium having recorded thereon a program for functioning as a document classification device including: a classification unit that classifies the document based on a weight.

[0013]

Embodiments of the present invention will be described in detail below with reference to the accompanying drawings. FIG. 1 is a block diagram showing a configuration of a document classification device according to an embodiment of the present invention. The document classification device includes a keyword extraction unit (extraction means)
1, control unit (control means) 5, analysis unit (weight calculation means) 1
0, a classification unit (classification means) 13 and a CD-ROM drive 14.

The analysis unit 10 includes a document structure analysis unit 2,
A classifying unit 13 including a time information analyzing unit 3 and a history analyzing unit 4;
Has a category degree calculation unit 6, a category keyword registration unit 7, a categorization unit 8, and a storage unit 9. Also, C
The D-ROM drive 14 reads a program stored in the CD-ROM 15 based on an instruction from the control unit 5.

The document structure analysis unit 2 and the time information analysis unit 3
In addition, the history analysis unit 4 constitutes an analysis unit 10 that calculates a weight for classifying the keyword into a document based on the importance of the keyword, the storage date and time or the storage location of the document including the keyword. .

The document to be classified 11 is a document to be classified by the document classification apparatus. Classification target document 11
May be a document downloaded from the Internet or the like and temporarily stored in a recording device such as an HDD, or may be an element in a set of documents that is the history information 12. The classification target document 11 and the history information 12 are in a state where information such as the storage date and time and storage location of each document can be recognized.
It is stored in a recording device (storage means) such as an HDD.

The history information 12 is a set of documents managed by the document classification device. The history information 12 is, for example, Wo
rld Wide Web browser (hereinafter referred to as Web browser)
Cache documents, Web browser favorite files or bookmark link destination documents, input character strings at the time of a search operation, or a hierarchical classification structure already performed by the user or a hierarchical structure automatically classified.

The keyword extracting unit 1 includes a classification target document 11
Then, a keyword is extracted from the document of the history information 12. When a certain character string in the document of the classification target document 11 and the document of the history information 12 matches a category keyword registered in advance in a keyword dictionary (not shown) of the category keyword registration unit 7, the character string is extracted as a keyword. I do.

In the analysis unit 10, the document structure analysis unit 2
Calculates the weight on the document structure of the keyword extracted by the keyword extraction unit 1 as the structure weight. The greater the structural weight of the keyword, the more important the keyword in the document.

For example, the structural weight is increased because the heading in the classification target document 11 or the history information 12 in the document,
These are keywords used in titles and titles, and keywords used in hyperlinks and the like. In addition, the structural weight of the highlighted keyword, the keyword displayed in a font larger than that of the other keywords, and the keyword used in the proper noun also increases. This structural weight is changed by an instruction from the control unit 5.

The time information analysis unit 3 calculates the weight on the time axis of the keyword extracted by the keyword extraction unit 1 as a time weight. From the information such as the update date and time and the storage date and time of the document of the history information 12 including the keyword extracted from the classification target document 11, the date and time at which the keyword appears most frequently is calculated. In this case, a date and time at which the appearance frequency of the keyword exceeds a predetermined frequency may be calculated as the keyword appearance date and time.

For example, a keyword whose calculation date and time is closer to the reference date and time has a larger time weight, and a keyword whose calculation date and time is farther from the reference date and time has a smaller time weight. The time weight is changed by an instruction from the control unit 5. Further, the history analysis unit 4 calculates, as a history weight, the frequency at which the keyword extracted by the keyword extraction unit 1 appears in the document of the history information 12.

For example, the history weight is large when used for a page recently browsed by a Web browser and stored in a cache, or when used for a character string or the like used in a search operation. Become. This history weight is changed by an instruction from the control unit 5.

As described above, in the present embodiment, the document structure analysis unit 2, the time information analysis unit 3, and the history analysis unit 4 are provided, and the keywords extracted from the classification target document 11 The document structure analysis unit 2 can detect an important category as an index indicating whether or not the classification destination is appropriate, and the time information analysis unit 3 assigns the appearance date and time of the keyword as time information. This makes it possible to detect a topical category and to detect a preference category by the history analysis unit 4.

The control unit 5 mainly performs the following four controls. As a first control, a reference value of a weight calculated by the document structure analysis unit 2, the time information analysis unit 3, and the history analysis unit 4 is set. The standard values for document structure analysis include headings,
Set a reference vector with title, caption, hyperlink, highlight, and font as elements. With respect to this reference vector, it is determined which element is to be given a specific gravity and the weight is calculated. In addition, reference values for time information analysis include:
Set the value on the time axis. The time weight of a keyword close to the set reference value increases.

The reference values in the history information analysis include:
A reference vector having the document of the history information 12 as an element is set. With respect to this reference vector, it is determined which element is to be given a specific gravity and the weight is calculated. For example, if a reference vector in which the element “input character string at the time of the search operation” is enlarged is set as the reference value, the keyword used in the “input character string at the time of the search operation” has a large weight, Weights are calculated.

As a second control, when the categorical degree is derived in the categorical degree calculating section 6 from the weights calculated by the document structure analyzing section 2, the time information analyzing section 3, and the history analyzing section 4, the specific weight of each weight is controlled. I do.

As a third control, the number of category keywords is controlled. When the number of classification target documents 11 is large, the number of keywords registered as category keywords is increased, and when the number of classification target documents 11 is small, the number of keywords registered as category keywords is reduced.

As a fourth control, the maximum value and the minimum value of the number of documents to be classified are controlled. If the number of documents to be classified exceeds the maximum value, a new category keyword is registered and re-classified. If the number of documents to be classified is less than the minimum value, the category is not classified and classified into another category.

The category degree calculation unit 6 derives a category vector for each keyword from the document structure analysis unit 2, the time information analysis unit 3, and the history analysis unit 4, and calculates the category degree from the size of the category vector. The category vector is made up of three elements, a structure weight, a time weight, and a history weight, and a specific weight for each weight is given by the control unit 5.

As described above, when deriving the category vector, the specific gravity for each weight can be changed. When the weight of the weight in the document structure is increased, a category vector that emphasizes the importance of the keyword is derived. When the relative weight of the time weight, which is a topical element based on the information on the date and time of storage of the document, is increased, a category vector that emphasizes topicality is derived. When the weight of the history weight, which is a preference element based on the information on the operation history of the user, is increased by using the information on the storage location of the document, a category vector that emphasizes the preference of the user is derived.

The category keyword registration unit 7 registers a category keyword based on the category degree calculated by the category degree calculation unit 6. For example, the keywords extracted from the documents to be classified 11 and the documents of the history information 12 are registered in ascending order of category degree up to the upper limit of the number of category keywords given by the control unit 5.

The categorizing unit 8 categorizes the category keyword having the highest category degree from the category keywords registered by the category keyword registering unit 7 as the classification destination of the document 11 to be classified. Storage unit 9
Stores the classification target document 11 based on a classification destination instruction from the categorizing unit 8.

Next, the operation of the document classification device according to the present embodiment will be described with reference to the flowcharts shown in FIGS. FIG. 2 is a flowchart illustrating a document classification process of the document classification device according to the present embodiment. First, in step S11, a category keyword is registered from the keywords extracted from the classification target document 11, and then, in step S12, the classification target document 11 is categorized.

In step S13, the number of documents to be classified is
It is determined whether or not the value is equal to or less than the set reference minimum value. If the number of documents to be classified is equal to or less than the reference minimum value, the category keyword set as the classification destination is determined to be inappropriate as a category and is deleted, and the process returns to step S12 to execute the categorizing process again. .
If the number of documents to be classified is larger than the reference minimum value, the process proceeds to step S14.

In step S14, it is determined whether or not the number of documents to be classified is equal to or greater than a set reference maximum value.
If the number of documents to be classified is equal to or greater than the reference maximum value, it is determined that the document can be classified into a smaller unit category, a category keyword is newly registered, and the process returns to step S12 to execute the categorizing process again. I do. Also,
If the number of documents to be classified is smaller than the reference maximum value, the process proceeds to step S15, and the classification target documents 11 are stored in the classification destination.

FIG. 3 is a flowchart for explaining the category keyword registration processing of the document classification device according to the present embodiment, and corresponds to the processing in step S11 in FIG. First, in step S21, keywords are extracted from the documents to be classified 11 and the documents of the history information 12, and then, in step S22, the categorization is calculated for each of the extracted keywords.

Next, in step S23, the calculated category degree is compared with the reference value. Here, if the calculated category degree is equal to or larger than the reference value, the process proceeds to step S24 and is registered as a category keyword. If the calculated category degree is equal to or smaller than the reference value, the process proceeds to step S25.

In step S25, it is determined whether all the keywords have been searched. If it is determined that all the keywords have been searched, the category keyword registration processing is completed. If not all keywords have been searched, the process returns to step S21 and proceeds to the processing of the next keyword.

FIG. 4 is a flowchart for explaining the category degree calculation processing of the document classification device according to the present embodiment, and corresponds to the processing in step S22 in FIG. First, in step S31, a structure weight based on the document structure is calculated. The document structure analysis unit 2 checks the document structure of the keywords included in the classification target document 11 among the keywords extracted by the keyword extraction unit 1, and calculates a time weight for each keyword. As described above, the document structure indicates the place where the keyword is used, the character size, and the like in the documents of the classification target document 11 and the history information 12.

As the reference value of the structure weight, a reference vector having a heading, a title, a title, a hyperlink, a highlight, and a font as elements is set. Weighting is performed by distribution based on the reference vector. For example, in weighting based on a reference vector in which a specific gravity is given to a hyperlink, the structural weight of the keyword used for the hyperlink becomes large. This reference vector is set in advance,
The user can change it appropriately.

Next, in step S32, a time weight based on the date and time information is calculated. The time information analysis unit 3 examines date and time information of a document including the keyword included in the classification target document 11 among the keywords extracted by the keyword extraction unit 1, and calculates a time weight for each keyword. As described above, the date and time information indicates the date and time when the keyword appears. However, when the keyword appears in a plurality of documents, the value calculated from the number of appearances in each document and the storage date and time of each document is used as date and time information. The value on the time code is set as the reference value in the time weight calculation. The time weight of a keyword close to the set reference value increases.

In step S33, a history weight based on the history information 12 is calculated. The history analysis unit 4 includes, among the keywords extracted by the keyword extraction unit 1,
Find the storage location of the document containing the keyword contained in 1,
A history weight is calculated for each keyword.

As the reference value of the history weight, a reference vector having the document of the history information 12 as an element is set. Based on the reference vector, it is determined which element is to be weighted with a specific gravity. For example, if a vector in which the element “input character string at the time of the search operation” is increased is set as the reference value, the keyword used in the “input character string at the time of the search operation” is given a high specific gravity, and the history weight is calculated. Is done.

In step S34, a category vector having the structural weight, time weight, and history weight calculated for each keyword as elements is derived. The magnitude of each weight of the derived category vector is changed in consideration of the specific gravity of each weight from the control unit 5. Next, step S35
Then, the category degree is calculated from the magnitude of the derived category vector.

FIG. 5 is a flowchart showing a categorizing process of the document classification device according to the present embodiment.
Corresponds to the processing of step S12. First, in step S41, a category keyword registered for the classification target document 11 is searched, and then in step S4
In step 2, it is determined whether or not a category keyword has been found in the classification target document 11.

In step S42, classification target document 1
If a category keyword is found from 1, the process proceeds to step S43. In step S43, the category keyword having the highest category degree is searched for, and the category keyword having the highest category degree is classified into categories.

If no category keyword is found in the classification target document 11, step S44
To perform a similar category search. In step S44,
The category keyword to which the classification target document 11 is most similar is set as the classification destination.

FIG. 6 is a flowchart showing a similar category search process of the document classification device according to the present embodiment.
This corresponds to the process of step S44 in FIG. Since there is no keyword serving as a category keyword in the classification target document 11, first in step S51, the classification target document 11
The average of the category vectors corresponding to the respective keywords extracted from is calculated, and the calculated average is used as the category vector of the classification target document itself.

Next, in step S52, step S5
The similarity between the category vector of the document calculated in step 1 and the category vector of the category keyword is calculated. Next, in a step S53, it is determined whether or not the calculated similarity exceeds a maximum value.

If the calculated similarity exceeds the maximum value in step S53, the maximum similarity value and the similar category are updated in steps S54 and S55. If the calculated similarity does not exceed the maximum value, the process proceeds to step S56.

In step S56, it is determined whether the comparison between the category vector of the document and the category vectors of all category keywords has been completed. Here, when the comparison with the category vectors of all the category keywords is completed, the similar category search processing is completed. Also,
If it is determined that the comparison with the category vectors of all the category keywords has not been completed, step S5
Returning to the process of step 2, a similar category search for the next category keyword is performed. As described above, the category keyword having the highest similarity is the classification destination of the classification target document 11.

As described above, the document classifying apparatus according to the present embodiment uses the keyword extracting unit 1 for extracting a keyword from a set of manageable documents, A document structure analysis unit 2 for calculating a structure weight based on the structure of the document, a time information analysis unit 3 for calculating a temporal position of a keyword appearance from the document creation date / time information as a time weight, and a history weight for user operation history information. A history analysis unit 4 which calculates a category vector of a keyword from the structure weight, the time weight and the history weight, and a categorical degree calculating unit 6 which sets the magnitude of the vector as a categorical degree; A category keyword registration unit 7 for selecting and registering a category keyword as a category from the keywords possessed by the user, It includes a categorization unit 8 to determine, a storage unit 9 for storing the classifying target document 11 on the classification destination, the control unit 5 for controlling the classification by the document number of grouping destinations in the set,
Not only keywords extracted from the classification target document 11, but also
Since the document classification destination is determined using the time information such as the document creation date and the accumulated operation history information of the user, the categorization considering the topicality and the categorization matching the user's preference are realized. be able to.

In the present embodiment, the analysis unit 10 calculates the structure weight, the time weight, and the history weight, respectively, and the classification unit 13 uses the weights as elements based on the weight specific instructions from the control unit 5. Categorical vector is derived. The present invention is not limited to the above-described embodiment. As a special case in which the specific gravity is instructed by the control unit 5, at least the keyword importance (structure weight), topicality (time weight), or user preference (history weight) is used. Any configuration may be used as long as one is emphasized and a category vector is derived in which only the emphasized weight is used as an element to perform document classification processing.

In the present embodiment, the classifying unit 13 is divided into the category degree calculating unit 6, the category keyword registering unit 7, the categorizing unit 8 and the storing unit 9, but the present invention is not limited thereto. Any configuration may be used as long as the classification target document 11 is automatically classified based on each weight calculated by the analysis unit 10.

As described above, the document classification device of the present invention is also realized by a program for causing the document classification device to function. This program is stored in a computer-readable recording medium. In the present invention, a ROM (not shown) itself may be a program medium as the recording medium, or a program reading device such as a CD-ROM drive 14 is provided as an external storage device, and the recording medium is stored therein. It may be a program medium such as a CD-ROM 15 which can be read by being inserted. In any case, the stored program may be accessed and executed by the control unit 5, or the program may be read, and the read program may be downloaded to a program storage area (not shown). Alternatively, a method in which the program is executed may be used. It is assumed that this download program is stored in the main unit in advance.

Here, the program medium is a recording medium configured to be separable from the main body, such as a tape system such as a magnetic tape or a cassette tape, a magnetic disk such as a floppy disk or a hard disk, or a CD-ROM / MO / M.
It carries a fixed program including a disk system of an optical disk such as a D / DVD, a card system such as an IC card (including a memory card) / optical card, or a semiconductor memory such as a mask ROM, EPROM, EEPROM, flash ROM, or the like. It may be a medium.

Further, the medium may carry a program in a fluid manner such that the program is downloaded from a communication network (not shown) via a transmitting / receiving means (not shown). When the program is downloaded from the communication network in this manner, the download program is stored in advance in the apparatus main body,
Alternatively, it may be installed from another recording medium. The content stored in the recording medium is not limited to a program, but may be data.

[0059]

According to the present invention, it is possible to automatically classify documents to be classified by extracting a category that matches the degree of importance, topicality, or user preference of the keyword itself. In addition, a system suitable for user needs can be realized by freely changing the specific gravity of the weight corresponding to each of the importance of the keyword itself, topicality, and user preference.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of a document classification device according to an embodiment of the present invention.

FIG. 2 is a flowchart illustrating a document classification process of the document classification device according to the embodiment of the present invention.

FIG. 3 is a flowchart illustrating a category keyword registration process of the document classification device according to the embodiment of the present invention.

FIG. 4 is a flowchart illustrating a category degree calculation process of the document classification device according to the embodiment of the present invention.

FIG. 5 is a flowchart illustrating a document categorizing process of the document classification device according to the embodiment of the present invention.

FIG. 6 is a flowchart illustrating a similar category search process of the document classification device according to the embodiment of the present invention.

[Explanation of symbols]

 1: Keyword extraction unit (extraction means) 2: Document structure analysis unit 3: Time information analysis unit 4: History analysis unit 5: Control unit (control means) 6: Category degree calculation unit 7: Category keyword registration unit 8: Categorization unit 9: storage unit 10: analysis unit (weight calculation unit) 11: classification target document 12: history information 13: classification unit (classification unit) 14: CD-ROM drive 15: CD-ROM (recording medium)

Claims (6)

[Claims] 1. A storage unit for storing a document, an extraction unit for extracting a keyword from the document stored in the storage unit, a degree of importance of the keyword extracted by the extraction unit, Weight calculating means for calculating a weight of the keyword as a classification destination of the document based on at least one of a storage date and time and a storage location; and classifying the document based on the weight calculated by the weight calculating means. A document classification device comprising: 2. The method according to claim 1, wherein the weight calculation unit changes the weight when the keyword is used in at least one of a heading, a title, a title, a hyperlink or highlighting of the document, or a proper noun. Claim 1 characterized by the following:
Document classification device as described. 3. The weight calculating means calculates a date and time at which the keyword appears from the storage date and time of the document including the keyword, and calculates the weight based on a comparison between the calculated date and time and a predetermined reference date and time. 2. The document classification device according to claim 1, wherein: 4. The weight calculation means, wherein the document including the keyword is a cache of a predetermined browser, a link destination of a favorite file or a bookmark of the browser,
2. The document classification apparatus according to claim 1, wherein the weight is changed in a case where at least one of a character input string at the time of search, a hierarchical classification structure performed by a user, and an automatically classified hierarchical structure. 5. A control unit for controlling a specific gravity of the weight calculated by the weight calculating unit, wherein the classifying unit is configured to determine a specific gravity based on the specific gravity from the control unit and the weight calculated by the calculating unit. 2. The document classification device according to claim 1, wherein the document is classified. 6. A computer, comprising: storage means for storing a document; extraction means for extracting a keyword from the document stored in the storage means; importance of the keyword extracted by the extraction means; Based on at least one of the storage date and time or storage location of the containing document,
Weight calculating means for calculating a weight of the keyword as a classification destination of the document, and a classifying means for classifying the document based on the weight calculated by the weight calculating means,
A computer-readable recording medium on which a program for causing a computer to function as a document classification device comprising: