JP2006277413A - Document classification device and document classification method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a document classification device for changing a parameter to be used for classification for every document to be classified. <P>SOLUTION: The document classification device is provided with a comparison object document information storage part 11 for storing the information of a comparison object document to be compared with a document to be classified and the field of the comparison object document as comparison object document information, a word weight information storage part 12 for storing a word and the word weight of the word and a classification processing part 10 for comparing the document to be classified with the comparison object document information, and for extracting a commonly used word commonly used by the document to be classified and the comparison object document, and for generating the commonly used word, the use frequency of the commonly used word, the word weight of the commonly used word read from the word weight information storage part and an adjustment value for adjusting the word weight set for every word weight as common word information, and for calculating the similarity of a plurality of comparison object documents and the document to be classified based on the generated common word information, and for specifying a field based on the calculated similarity, and for specifying a new field based on the comparison object document by varying the adjustment value. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a document classification apparatus and a document classification method for classifying document fields.

  Conventionally, there is a document classification system that identifies a field to which an input document belongs based on a plurality of document information stored in advance in a database. In such a document classification system, first, a comparison target document similar to an input document whose classification is to be specified is extracted from a plurality of comparison target documents whose fields are specified in advance and stored in the database. Thereafter, there is a method for specifying a field to which an input document belongs based on a field previously associated with the extracted comparison target document (for example, Patent Document 1).

Further, in order to increase the efficiency of the classification process, a document classification system generally performs a classification process for a large number of input documents at once. Also, in such a document classification system, the input document is often calculated for similarity with a large number of comparison target documents. Furthermore, in a document classification system that requires high accuracy, the result of classification by a computer is manually checked.
JP 2001-1555025 A

  As described above, the conventional document classification system that collectively classifies a large number of input documents according to the same standard can efficiently process a large number of documents at one time, but at the time when the classification process is performed collectively. The result is confirmed. As a result, the problem of interactivity may arise.

  It is possible to adjust the parameters again in the classification process and execute the classification again, but according to this, by adjusting the parameters, the changed parameters are applied to all the classification results that are classified after that. Is done. According to this, there is a problem that all classification results are classified with the same parameter.

  The present invention has been made to solve the above problem, and provides a document classification apparatus and a document classification method capable of adjusting parameters used for classification for each document to be classified. Objective.

  According to the document classification apparatus according to the first aspect of the present invention, the document classification apparatus classifies the field to which the input classified document belongs, and information of a comparison target document to be compared with the classified document and the comparison The comparison target document information storage unit that stores the field of the target document as comparison target document information, the word, the word weight information storage unit that stores the word weight of the word, and the classified document are compared with the comparison target document information. The common use words that are commonly used in the classified document and the comparison target document are extracted, the common use words, the number of use of the common use words, and the word weights of the common use words read from the word weight information storage unit And an adjustment value that is set for each word weight and adjusts the word weight is generated as common word information, and the similarity between a plurality of comparison target documents and classified documents is calculated based on the generated common word information. Seeking And a classification processing unit that identifies a field based on the obtained similarity and identifies a new field based on a comparison target document by varying an adjustment value based on an instruction from an input device. Yes.

  Further, according to the document classification device according to the second feature, the document classification device classifies the field to which the input classified document belongs, and information on the comparison target document to be compared with the classified document, and the comparison target A comparison target document information storage unit that stores the field of the document as comparison target document information, a word, a word weight information storage unit that stores the word weight of the word, and a classified document compared with the comparison target document information Commonly used words that are commonly used in the classified document and the comparison target document are extracted, the commonly used words, the number of times of use of the commonly used words, and the word weights of the commonly used words read from the word weight information storage unit, A common word information generation unit that generates, as common word information, an adjustment value that is set for each word weight and adjusts the word weight, and the number of times each common word used in the common word information is used and the common word used A similarity calculation unit that calculates a similarity based on the word weight and an adjustment value associated with the word weight, and identifies and classifies the field to which the classified document belongs based on the similarity calculated by the similarity calculation unit When the classification result is confirmed, a classification result including the adjustment value of each word used in the classified document is generated and stored in the storage device. And a classification result storage unit to be stored.

  According to the present invention configured as described above, it is possible to provide a document classification apparatus and a document classification method for adjusting parameters for classifying a document.

  According to the present invention, parameters for classifying a document can be adjusted in a document classification apparatus and a document classification method.

  Hereinafter, a document classification apparatus 1 and a document classification method according to the preferred embodiment of the present invention will be described with reference to the drawings.

[Document Classification Device]
FIG. 1 is a block diagram of a document classification apparatus 1 according to the preferred embodiment of the present invention.

  A document classification apparatus 1 shown in FIG. 1 includes a classification processing unit 10, a comparison target document information storage unit 11, a word weight information storage unit 12, and a classification result storage unit 13.

  The comparison target document information storage unit 11 compares information (text information) of the comparison target document to be compared with the classified document to be classified into the field to which the document belongs, and the comparison target document information in which the field of the comparison target document is associated. Is remembered.

  The word weight information storage unit 12 stores a word and a word weight serving as an index indicating characteristics of the field of the document including the word.

  The classification result storage unit 13 stores the classification result including the information related to the classified document classified by the classification processing unit 10 and the adjustment value of the used word used in the classified document.

  The classification processing unit 10 compares the classification key document with the comparison target document information, extracts common usage words that are commonly used in the classification key document and the comparison target document, and at least the common usage words and The common word information by associating the number of use of the common word, the word weight of the common word read from the word weight information storage unit, and an adjustment value set for each word weight and adjusting the word weight. Generate as Further, the classification processing unit 10 obtains a similarity with a plurality of comparison target documents from the common word information, obtains a classification result specifying the field based on the field of the comparison target document having a high similarity, and obtains a classification result. The data is stored in the storage unit 13. Further, the classification processing unit 10 specifies a new field based on the comparison target document by varying the adjustment value based on an instruction from the input device.

  FIG. 2 shows an example of comparison target document information stored in the comparison target document information storage unit 11. The comparison target document is a document to be compared with a classified document to be classified into fields. The comparison target document information is generated based on the comparison target document.

  Specifically, the comparison target document information 11a illustrated in FIG. 2 includes information on “title”, “field”, “use word”, “use count”, and “adjustment value” of a plurality of comparison target documents. “Title” is a title of the comparison target document, and “Field” is a field defined in the comparison target document. The “used word” is a word used in the comparison target document, and the “use count” is the number of times each used word is used in the comparison target document.

  The “adjustment value” is a value for adjusting the word weight for each used word of each classified document. In the document classification device 1 according to the preferred embodiment of the present invention, the parameter for obtaining the similarity is adjusted by changing the “adjustment value”.

  According to the comparison target document information 11a shown in FIG. 2, the title of the comparison target document 1 is “reduction in database update processing time” and the field is “database update”. Further, the words used in the comparison target document 1 and the number of times of use thereof are 2 for “Large scale”, 5 for “Database”, 8 for “Update processing”, and 3 for “Time”, respectively. , “Shortening” represents 2 times. In addition, the initial value “1.0” is set as the “adjustment value” of each used word.

  FIG. 3 is an example of word weight information 12 a stored in the word weight information storage unit 12. In the word weight information 12a shown in FIG. 3, for example, the word weight of “automatic classification” is “8.5”, and the word weight of “database” is “4.3”.

  For this “word weight”, for example, the reciprocal of the number of times the used word is used in all comparison target documents in the comparison target document information storage unit 11 is used. This is because words that are used frequently are general words and do not represent the characteristics of the document, and conversely, words that are used frequently are considered characteristic words. . In the best mode of the present invention, it is assumed that word weight information 12a as shown in FIG. 3 is created in advance and stored in the word weight information storage unit 12 as word weight information used for classification.

  As shown in FIG. 4, the document classification apparatus 1 according to the preferred embodiment of the present invention includes a central processing control device 101, a ROM (Read Only Memory) 102, a RAM (Random Access Memory) 103, and an input / output interface 109. Are connected via a bus 110. An input device 104, a display device 105, a communication control device 106, a storage device 107, and a removable disk 108 are connected to the input / output interface 109.

  The central processing control device 101 reads and executes a boot program for starting the document classification device 1 from the ROM 102 based on an input signal from the input device 104, and further reads an operating system stored in the storage device 107. Further, the central processing control device 101 controls various devices based on input signals from the input device 104, the communication control device 106, etc., and reads programs and data stored in the RAM 103, the storage device 107, etc., into the RAM 103. A processing device that loads and implements a series of processing described later, such as data calculation or processing, based on a program command read from the RAM 103.

  The input device 104 includes input devices such as a keyboard and a mouse through which an operator inputs various operations. The input device 104 generates an input signal based on the operation of the operator, and inputs via the input / output interface 109 and the bus 110. It is transmitted to the central processing control apparatus 101. The display device 105 is a CRT (Cathode Ray Tube) display, a liquid crystal display, or the like. The display device 105 receives an output signal to be displayed on the display device 105 from the central processing control device 101 via the bus 110 and the input / output interface 109. This is a device that displays the processing results of the processing control device 101. The communication control device 106 is a device such as a LAN card or a modem, and is a device that connects the document classification device 1 to a communication network such as the Internet or a LAN. Data transmitted / received to / from the communication network via the communication control device 106 is transmitted / received to / from the central processing control device 101 via the input / output interface 109 and the bus 110 as an input signal or an output signal.

  The storage device 107 is a semiconductor storage device, a magnetic disk device, or the like, and stores programs and data executed by the central processing control device 101. The removable disk 108 is an optical disk or a flexible disk, and signals read / written by the disk drive are transmitted / received to / from the central processing control apparatus 101 via the input / output interface 109 and the bus 110. The storage device 107 of the document classification device 1 according to the embodiment of the present invention stores a document classification program, and also stores a comparison target document information storage unit 11, a word weight information storage unit 12, and a classification result storage unit 13. Is done. Further, the document classification program is read and executed by the central processing control device 101 of the document classification device 1 to implement the classification processing unit 10.

  The document classification apparatus 1 according to the best embodiment of the present invention may be realized by a single computer or a plurality of computers that can communicate with each other. For example, the configuration for performing batch processing and the configuration for performing interactive processing may be on the same computer system or on different computer systems connected via a network or the like. . Further, each of the classification processing units 10 may be realized by a single computer, or may be realized by a plurality of computers.

  As shown in FIG. 5, the classification processing unit 10 in the document classification apparatus 1 according to the embodiment of the present invention includes a control unit 200 and a memory unit 250.

  The control unit 200 includes an initialization unit 201, an input unit 202, a word weight reading unit 203, a classified document information generation unit 204, a comparison target document information reading unit 205, a common word information generation unit 206, and a comparison target document similarity calculation unit. 207, a field-specific similarity accumulation unit 208, a field identification unit 209, a similarity calculation parameter adjustment unit 210, and a classification result storage unit 211.

  The memory unit 250 includes a word weight buffer unit 251, a classified document information buffer unit 252, a comparison target document information buffer unit 253, a common word information buffer unit 254, a comparison target document similarity buffer unit 255, and a field-specific similarity accumulation unit. A value buffer unit 256 is provided.

  The word weight buffer unit 251 stores word weight information in which a used word and its word weight are associated with each used word that is a word used in the comparison target document.

  The classified document information buffer unit 252 stores classified document information generated from the classified document to be classified.

  The comparison target document information buffer unit 253 stores comparison target document information to be compared with classified document information.

  The common word information buffer unit 254 includes a common use word that is a word used in common in the classified document and the comparison target document, the number of times the common use word is used in the document, and the word weight of the common use word. The common word information associated with the adjustment value that is set for each word weight and adjusts the word weight is stored.

  The comparison target document similarity buffer unit 255 stores the similarity calculated based on the commonly used word information for each comparison target document obtained with respect to the classified document as the comparison target document similarity. In the best mode of the present invention, the comparison target document similarity calculation unit 207 calculates the comparison target document similarity, and stores the comparison target document similarity information generated based on the comparison target document similarity. Yes.

  The field-specific similarity integrated value buffer unit 256 stores field-specific similarity integrated values totaled for each field to which the comparison target document belongs with respect to the comparison target document similarity.

  The initialization unit 201 initializes the buffer units 251 to 256 of the memory unit 250.

  The input unit 202 inputs a classified document and an operation instruction via an input device.

  The word weight reading unit 203 reads word weight information from the word weight information storage unit 12 into the word weight buffer unit 251.

  The classified document information generation unit 204 decomposes the classified document input to the input unit 202 into words, generates classified document information including each decomposed word and the number of times the word is used, and classifies the classified document. It is stored in the document information buffer unit 252.

  The comparison target document information reading unit 205 reads the comparison target document information from the comparison target document information storage unit 11 into the comparison target document information buffer unit 253.

  The common word information generation unit 206 reads out the classified document information stored in the classified document information buffer unit 252 and the comparison target document information stored in the comparison target document information buffer unit 253, and reads the classified document and the comparison target document. Extract commonly used words commonly used in. Further, the common word information generation unit 206 generates common word information in which the number of times the common used word is used in the document, the word weight of the common used word, and the initial value of the adjustment value, and generates a common word information buffer. Stored in the unit 254.

  The comparison target document similarity calculation unit 207 reads the common word information stored in the common word information buffer unit 254, and based on the number of common words used, the word weight, and the adjustment value included in the read common word information. The comparison target document similarity that is the similarity between the classified document and each comparison target document is calculated. Further, the comparison target document similarity calculation unit 207 generates comparison target document similarity information based on each calculated comparison target document similarity, and stores it in the comparison target document similarity buffer unit 255.

  In the best mode of the present invention, the method for calculating the comparison target document similarity is based on the sum of the number of used words commonly used in the two documents of the classified document and the comparison target document. This will be described using an example in which the product of the weight and the adjustment value is used as the similarity. However, the method of calculating the similarity is not limited to the above method, and may be obtained by other calculation methods.

  When the comparison target document similarity is stored in the comparison target document similarity buffer unit 255, the sector similarity accumulation unit 208 accumulates the comparison target document similarity for each field for each field that matches the comparison target document similarity. The added value is added to the integrated value and is stored in the field-specific similarity integrated value buffer unit 256.

  The field specifying unit 209 reads the field-specific similarity integrated value stored in the field-specific similarity integrated value buffer unit 256 and the common word information 254a stored in the common word information buffer unit 254 and associates them with each other. Field identification results are generated and output to an output device such as a display device.

  The similarity calculation parameter adjustment unit 210 performs reclassification based on the new common word information 254a based on the adjustment value changed to adjust the word weight by the user via the input device. The common word information 254a in the buffer unit 254 is rewritten.

  When the classification result is confirmed, the classification result storage unit 211 reads the common word information 254a stored in the common word information buffer unit 254, and the classification result for registration including the confirmed classification, the used word, and the adjustment value. 13 a is generated and stored in the classification result storage unit 13.

[Document classification processing]
Next, document classification processing in the document classification apparatus 1 according to the embodiment of the present invention will be described with reference to FIGS. The flowcharts shown in FIGS. 6 and 7 show processing in the classification processing unit 10.

  First, as shown in the flowchart of FIG. 6, the initialization unit 201 initializes the buffer units 251 to 256 of the memory unit 250 (S001). Thereafter, the word weight reading unit 203 reads word weight information from the word weight information storage unit 12 into the word weight buffer unit 251 (S002).

  Subsequently, when a classified document is input via the input unit 202, the classified document information generation unit 204 decomposes the input classified document into words. The classified document information generation unit 204 generates classified document information including each decomposed word and the number of times each word is used, and stores the generated classified document information in the classified document information buffer unit 252 ( S003).

  FIG. 8 shows a classified document 1 that is an example of a classified document input to the classified document information generation unit 204 via the input unit 202. The second half of the document is omitted. Such a plurality of classified documents are input to the classified document information generation unit 204.

  FIG. 9 shows an example of classified document information 252a generated by the classified document information generation unit 204 based on “classified document 1”. As illustrated in FIG. 9, the classified document information 252a includes, for example, “used words” used in the classified document and the number of times the used words are used in the target classified document “ This is information in which “the number of uses” and “adjustment value” are associated. Here, “1.0” is set as the initial value of the “adjustment value”.

  The process in step S003 is performed for all classified documents to be classified (S004). For example, when 2000 documents are input as classified documents, the process of step S003 is repeated 2000 times.

  When classified document information is generated for all classified documents, the comparison target document information reading unit 205 reads the comparison target document information 11a from the comparison target document information storage unit 11 into the comparison target document information buffer unit 253 (S005). ).

  Subsequently, the common word information generation unit 206 reads the classified document information 252 a from the classified document information buffer unit 252, reads the comparison target document information 11 a from the comparison target document information buffer unit 253, and sorts the classified document and the comparison target document. And the common word information 254a is extracted by combining the extracted common use word with the total number of times used in the classified document and the comparison target document. It is generated and stored in the common word information buffer unit 254 for each comparison target document (S006).

  Thereafter, the processes in steps S005 and S006 are repeated until the common word information 254a is generated and stored for all the comparison target documents (S007).

  FIG. 10 shows an example of the common word information 254a. In the common word information 254a, for each combination of each classified document and the comparison target document, the “field”, “used word”, and the used word of the comparison target document are used in the classified document and the comparison target document. “Use count”, “word weight” of the used word, and “adjustment value” for adjusting the word weight are stored in association with each other.

  For example, in the common word information 254a shown in FIG. 10, when the classified document 1 is compared with a document whose title is the comparison target document 1 and whose title is “reduction in database update processing time”, the usage count of the used word “large” is “ This indicates that the usage count of the word “database” is “11” and the usage count of “time” is “5”.

  In the best embodiment of the present invention, the comparison target document similarity calculation unit 207 calculates the similarity by adding the product of the number of times of use of the common word included in the common word information, the word weight, and the adjustment value. . However, in addition to this, it is also possible to calculate the similarity using a vector space method.

  Next, the comparison target document similarity calculation unit 207 reads the common word information 254a stored in the common word information buffer unit 254, calculates the similarity, and compares the calculated similarity as comparison target document similarity information 255a. It is stored in the target document similarity buffer unit 255 (S008).

  In order to calculate the comparison target document similarity, first, the product of the number of times of use and the word weight is calculated for each commonly used word. The total value of the product of the number of times used for each common word and the word weight is used as each comparison target document similarity for each classified document, and the comparison target document similarity is based on each comparison target document similarity. Information 255a is generated.

  For example, in the case of “comparison target document 1” shown in FIG. 10, the comparison target document similarity is 5 × 2.1 + 11 × 4.3 + 5 × 1.7 = 66.3.

  FIG. 11 shows an example of comparison target document similarity information 255 a stored in the comparison target document similarity buffer unit 255.

  After that, when the comparison target document similarity information 255a is stored in the comparison target document similarity buffer unit 255, the field-specific similarity integration unit 208 sets the comparison target document similarity to the field-specific similarity integration value. The comparison target document similarity is added, and the field-specific similarity integrated value information 256a is rewritten and stored in the field-specific similarity integrated value buffer unit 256 (S009).

  When the similarity calculation result is in the state shown in FIG. 11, first, in the field “database update”, the similarity 66.3 calculated for the document whose title is “reduction in database update processing time” and the title is The similarity 43.5 calculated for the “text database update” document is added, and the similarity calculated for the document classified as “database update” in the subsequent classification target document is added to add the similarity by field It is an integrated value.

  FIG. 12 shows an example of the field-specific similarity integrated value information 256a generated from the field-specific similarity integrated values obtained as a result of processing all the comparison target documents in step S105.

  Next, when the comparison target document similarity calculated for all the comparison target documents is added to the field-specific similarity integrated value (YES in S010), the field specifying unit 209 determines the field-specific similarity integrated value buffer unit 256. The field-specific similarity integrated value information 256a is read out, the field-specific similarity integrated values are arranged in descending order, the common word information 254a stored in the common word information buffer unit 254 is read out, and the field of the field-specific similarity integrated value is large. The field identification result associating the common word information 254a with the integrated similarity value in order is generated and output to the output device. This field identification result includes “field name”, “title” and “similarity” of the comparison target document corresponding to the field. The field specifying unit 209 also displays word weights used in calculating similarity with each comparison target document. (S011) At this time, the word weight of each word is displayed in a rewritable state.

  FIG. 13 shows an example of the field identification result display screen 600 displayed in step S011. FIG. 14 shows an example of a display screen 601 of the word weight adjustment screen displayed in step S011. In the example of the display screen 601 in FIG. 14, the word weights are displayed by sorting in descending order of the number of uses.

  The display screen 600 shown in FIG. 13 is provided with a word weight adjustment button 600a, and the display screen 601 is displayed by pressing the word weight adjustment button 600a. However, the word weight adjustment button 600a is indispensable. Instead, the field identification result and the word weight adjustment screen may be displayed at the same time, or may be sequentially displayed after a predetermined time.

Subsequently, the similarity calculation parameter adjustment unit 210 accepts a change in an adjustment value for adjusting the word weight displayed in step S011. (S012)
Specifically, when the user refers to the contents of the displayed classified document and the classification result and determines that the classification result is not correct, the user selects the word that affected the classification included in the classification result and the word weight. It is possible to refer to and adjust the word weight by changing the adjustment value, and to obtain a new classification result using the changed adjustment value. For example, the user does not indicate the characteristics of the field of the classified document among the displayed word weights, but the user has a high weight, or conversely, indicates the characteristics of the field. Nevertheless, when the weight is low, the word weight can be adjusted.

  Referring to the display screen 601 shown in FIG. 14, the “adjustment value” can be freely changed. By pressing the reclassification start button 601 a, the changed “adjustment value” will be described later. The classification is started again based on “ For example, the user determines that “database update” or “document search” is not appropriate as the field belonging to the classified document shown in FIG. 8, and the cause is that the word weight of “database” is high. Assume that you have determined. In this case, the adjustment value of the “database” can be reset to a low value, for example, to 0.2.

  Subsequently, when the similarity calculation parameter adjustment unit 210 determines that the adjustment value has been changed (YES in S013), the similarity calculation parameter adjustment unit 210 uses the changed adjustment value to generate a common word. The common word information 254a in the information buffer unit 254 is rewritten (S014). Specifically, the similarity calculation parameter adjustment unit 210 rewrites the “adjustment value” of the common word information 254a. FIG. 15 shows common word information 254b as an example in which “adjustment value” of “database” is rewritten to 0.2. Thereafter, the processing from step S008 to S013 is executed again.

  FIG. 16 shows comparison target document similarity information 255b generated based on the similarity calculated for the comparison target document 1 when the “adjustment value” of “database” is rewritten to 0.2 as in the above-described example. An example is shown. As a result of such adjustment, the score of the “database update” and “document search” fields is reduced, and the “document classification” field appropriate as the classification destination is raised to the top, which is not suitable as the field to which the classified document belongs. It becomes.

  In the best embodiment of the present invention, the adjustment value for adjusting the common word and its word weight is used as intermediate data for similarity calculation, but is not limited thereto, and based on values calculated by a plurality of methods, An embodiment in which the field is specified by adjusting the ratio of the plurality of values is also conceivable.

  If the adjustment is not accepted in the determination of step S013 (NO in S013), it is determined that the adjustment value is currently set, and the classification result storage unit 211 is stored in the common word information buffer unit 254. Based on the common word information, a weighted registration classification result 13 a is generated and stored in the classification result storage unit 13.

  FIG. 17 shows a registration classification result 13 a stored in the classification result storage unit 13. The classification result 13a shown in FIG. 17 is a registration classification result 13a generated for the classified document 1, and includes an adjustment value associated with the word used together with the contents of the field and the classified document 1.

  The above-described processing of steps S008 to 009 is repeated for all the classified documents to be processed (S016). For example, when 2000 documents are input as classified documents, the process is repeated 2000 times.

  According to the present invention described above, the parameters used for classification can be classified for each classified document that is the classification target. This makes it possible to improve the classification accuracy.

  As described above, the embodiments of the present invention have been described. However, it should not be understood that the descriptions and drawings constituting a part of this disclosure limit the present invention. From this disclosure, various alternative embodiments, examples, and operational techniques will be apparent to those skilled in the art.

  It goes without saying that the present invention includes various embodiments not described herein. Therefore, the technical scope of the present invention is defined only by the matters described in the above description and the invention specific matters according to the obvious claims.

1 is a block diagram for explaining a document classification device according to an embodiment of the present invention. It is an example of the comparison object document information memorize | stored in the comparison object document information storage part which concerns on the best embodiment of this invention. It is an example of the word weight information memorize | stored in the word weight information storage part which concerns on the best embodiment of this invention. It is a figure explaining the document classification device based on the best embodiment of this invention. It is a figure explaining the document classification device based on the best embodiment of this invention. It is a flowchart explaining the process in the document classification device based on the best embodiment of this invention. FIG. 7 is a flowchart for explaining processing in the document classification device according to the preferred embodiment of the present invention, following FIG. 6. It is an example of the classified document input into the document classification device based on the best embodiment of this invention. It is an example of the classified document information produced | generated in the document classification device based on the best embodiment of this invention. It is an example of the common word information produced | generated in the document classification device based on the best embodiment of this invention. It is an example of the comparison object document similarity memorize | stored in the document classification device based on the best embodiment of this invention. It is an example of the similarity integrated value according to field memorize | stored in the document classification device based on the best embodiment of this invention. It is an example of the field identification result calculated | required in the document classification device based on the best embodiment of this invention. It is an example of the word weight adjustment screen displayed in the document classification device according to the preferred embodiment of the present invention. It is an example of the common word information corrected in the document classification device according to the best embodiment of the present invention. It is an example of the comparison object document similarity information produced | generated based on the common word information corrected in the document classification device based on the best embodiment of this invention. It is an example of the classification result for registration produced | generated in the document classification device based on the best embodiment of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Document classification | category apparatus 10 ... Classification process part 11 ... Comparison object document information storage part 11a ... Comparison object document information 12 ... Word weight information storage part 12a ... Word weight information storage part 101 ... Central processing control apparatus 102 ... ROM
103 ... RAM
DESCRIPTION OF SYMBOLS 104 ... Input device 105 ... Display device 106 ... Communication control device 107 ... Memory | storage device 108 ... Removable disk 109 ... Input / output interface 110 ... Bus 200 ... Control part 201 ... Initialization part 202 ... Input part 203 ... Word weight reading part 204 ... Classified document information generation unit 205 ... comparison target document information reading unit 206 ... common word information generation unit 207 ... comparison target document similarity calculation unit 208 ... field-specific similarity accumulation unit 209 ... field specification unit 210 ... similarity calculation parameter adjustment Section 211 ... Classification result storage section 250 ... Memory section 251 ... Word weight buffer section 252 ... Classified document information buffer section 252a ... Classified document information 253 ... Comparison target document information buffer section 254 ... Common word information buffer sections 254a, 254b ... Common word information 255... Comparison target document similarity level Fan unit 255a, 255b ... comparison document similarity information 256 ... sector similarity integration value buffer 256a ... sector similarity integration value information 600 ... screen 600a ... adjustment button 601 ... screen 601a ... reclassification start button

Claims (8)

A document classification device for classifying a field to which an inputted classified document belongs,
A comparison target document information storage unit that stores information of a comparison target document to be compared with the classified document and a field of the comparison target document as comparison target document information;
A word and a word weight information storage unit for storing the word weight of the word;
The classified document is compared with the comparison target document information to extract a common use word that is commonly used in the classified document and the comparison target document, the common use word, and the number of times the common use word is used, The word weight of the commonly used word read from the word weight information storage unit and an adjustment value that is set for each word weight and adjusts the word weight are generated as common word information, and the generated common word A similarity between a plurality of comparison target documents and the classified document is obtained based on information, a field is specified based on the obtained similarity, and the adjustment value is varied based on an instruction from an input device. A classification processing unit that identifies a new field based on the comparison target document;
A document classification apparatus comprising: A document classification device for classifying a field to which an inputted classified document belongs,
A comparison target document information storage unit that stores information of a comparison target document to be compared with the classified document and a field of the comparison target document as comparison target document information;
A word and a word weight information storage unit for storing the word weight of the word;
The classified document is compared with the comparison target document information to extract a common use word that is commonly used in the classified document and the comparison target document, the common use word, and the number of times the common use word is used, A common word information generation unit that generates, as common word information, the word weight of the commonly used word read from the word weight information storage unit and an adjustment value that is set for each word weight and adjusts the word weight;
A similarity calculation unit that calculates a similarity based on the number of times of use of each commonly used word included in the common word information, the word weight of the commonly used word, and an adjustment value associated with the word weight;
A field identification unit that identifies the field to which the classified document belongs based on the similarity obtained by the similarity calculation unit and sets the classification result;
A similarity parameter adjustment unit that varies the adjustment value;
When the classification result is confirmed, a classification result storage unit that generates a classification result including an adjustment value of each used word of the classified document and stores it in a storage device;
A document classification apparatus comprising: The document classification apparatus according to claim 2, wherein the similarity parameter adjustment unit includes:
A document classification apparatus, wherein common word information is generated by varying an adjustment value of a specific common use word. The document classification device according to claim 2, wherein the similarity calculation unit includes:
A document classification apparatus characterized in that a sum of products of the number of times of use of each common word, a word weight, and a common word is calculated and the calculated total value is used as a similarity. A document classification method for classifying a field to which an input classified document belongs,
The comparison target document information for comparing the classified document with the classified document is compared with the comparison target document information associated with the field of the comparison target document. The commonly used words used are extracted, and the commonly used words, the number of times the common used words are used, the word weights of the commonly used words, and adjustment values for adjusting the word weights set for the respective word weights. It is generated as common word information, a similarity between a plurality of comparison target documents and the classified document is obtained from the generated common word information, a field is specified based on the obtained similarity, and an input device A classification processing step of varying the adjustment value based on the instruction and specifying a new field based on the comparison target document;
A document classification method characterized by comprising: A document classification method for classifying a field to which an input classified document belongs,
Comparing the classified document with the information of the comparison target document for comparing the classified document with the classified document, the common used word commonly used in the classified document and the comparison target document is extracted, A common word information generation step for generating, as common word information, the number of times of use of the common use word, the word weight of the common use word, and an adjustment value that is set for each word weight and adjusts the word weight;
A similarity calculation step of calculating a similarity based on the number of times of use of each commonly used word, the word weight of the commonly used word, and an adjustment value associated with the word weight;
A field specifying step of specifying the field to which the classified document belongs based on the similarity obtained for the classified document and making it a classification result;
A similarity parameter adjustment step for varying the adjustment value;
When the classification result is confirmed, a classification result storing step of generating a classification result including an adjustment value of each used word of the classified document and storing it in a storage device;
A document classification method characterized by comprising: The document classification method according to claim 6, wherein the similarity parameter adjustment step includes:
A document classification method, characterized in that common word information is generated by varying an adjustment value of a specific commonly used word. The document classification method according to claim 6 or 7, wherein the similarity calculation step includes:
A document classification method characterized in that a sum of products of the number of times of use of each common word, word weight, and a common word is calculated, and the calculated total value is used as a similarity.

Images