JP2005234731A - Document sorting apparatus and method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To monitor user's actions and adjust criteria of classification category after the fact. <P>SOLUTION: A document reception section 10 receives documents continuously. A document classification section 11 classifies the documents received in the document reception section 10 based on preliminarily set classification category criteria. A document display list of a classification category is displayed by input operation of a U/I section 15, and the document is opened by same document designation. A user action monitoring section 17 analyzes summary text of an opened document and records frequencies of important phrases. A classification criterion adjustment section 18 records continuously frequencies of important phrases and add the keyword with high importance for a long term, and the one increasing the importance with the lapse of time to the keywords to be used as criteria of the classification category. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a document classification technique for classifying documents based on a predetermined classification standard, and in particular, optimizes the classification standard.

  A document classification technique is known in which a keyword that characterizes a classification is set for each classification frame, and a document highly related to the keyword is allocated to the classification frame. Weights may be assigned between keywords. The user can browse a document by specifying a desired classification frame and narrowing down topics and themes. By the way, it is difficult to set keywords in consideration of various viewpoints after accurately grasping the registered documents and user needs from the beginning, and the planned documents are not properly classified with the initially set keywords. There are many cases.

  It is desirable to tune the keywords of the classification frame after the fact. Note that when a word vector is used instead of a keyword, a word vector is generated from a sample document set. In this case as well, a similar problem occurs because the sample document set deviates from the actually desired document set. There is.

  Patent documents related to this invention include the following.

  Patent Document 1 automatically trains a text classifier through learning. It discloses that after classifying a document based on the classification criteria of the classification frame, training the classification criteria using the classified documents (word vectors contained therein).

  Patent document 2 discloses changing the classification conditions manually.

  Patent Document 3 discloses that a feature vector is also obtained for an unclassifiable category, and the user is prompted to create a category when the unclassifiable category estimate exceeds a certain threshold.

However, in such document classification of patent documents, the classification format is fixed to the setting of the classification frame set at the beginning, and is not changed until the user resets it again. In order to respond, the user needs to set again.
JP-A-9-26963 JP 2000-305948 A JP 2002-117046 A

  The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a document classification technique that automatically adjusts the classification criteria according to changes in the environment and user preferences.

  In the configuration example document classification apparatus according to the present invention, a keyword that characterizes the classification is set in the classification frame, and a document highly related to the keyword is allocated to the classification frame. The user selects a document to be opened by looking at the title and summary of the document from a document list classified in an arbitrary classification frame. Then, such user actions are monitored, words included in the contents of the opened document (especially titles and summaries) are recorded, and the words characterizing the classification frame are corrected. The change of the user's preference is reflected in the classification from the next time.

  The present invention will be further described.

  According to one aspect of the present invention, in order to achieve the above object, the document classification apparatus includes: a document reception unit that receives a document; a document classification unit that classifies the document received by the document reception unit according to a classification standard; Classification result display means for displaying a classification result by the classification means; User action monitoring means for monitoring a user action inputted based on the classification result displayed on the classification result display means; Monitoring by the user action monitoring means And a classification reference adjusting means for adjusting the classification reference based on the user action.

  In this configuration, the classification criteria can be adjusted afterwards using a user action when the user uses the document classification device. All adjustments of classification criteria may be performed automatically without informing the user, or when adjustments are ready, the user or administrator is notified and adjustments are made with the confirmation of the user or administrator. May be.

  The classification standard is, for example, a keyword or word vector registered for classification. Keywords may be weighted.

  The user action monitored by the user action monitoring unit is, for example, a browsing operation for a classified document. For example, when viewing a document, the user selects an arbitrary document from a document list and displays the text.

  The user action monitoring means may monitor browsing time.

  Other user actions to be monitored include, for example, copy storage, commenting (annotation) for classified documents, and recommendations to other users, but are not limited thereto. Document copying and storage may be performed only for document attributes and document addresses. In this case, the document body is not duplicated and stored.

  The classification standard adjusting means adjusts the classification standard by adding or deleting keywords, for example. For example, a keyword determined based on a phrase extracted from a document related to a user action monitored by the user action monitoring unit is added, or the weight attached to the keyword is increased or decreased.

  According to another aspect of the present invention, the document classification apparatus includes: a document reception unit that receives a document; a document classification unit that classifies the document received by the document reception unit into a plurality of classification frames according to a classification criterion; Classification result display means for displaying the classification result of the classified frame by the classification means; user action monitoring means for monitoring a user action for each classification frame; and based on the user action monitored by the user action monitoring means Classification reference adjusting means for adjusting the classification reference is provided, and the display order of the classification frames is changed based on the user action monitored by the user action monitoring means.

  In this configuration, the classification standard is adjusted based on the user action for each classification frame, and it is possible to make adjustments afterwards so that the classification frame with less activity is less likely to be classified. For example, the weight of a keyword set as a classification criterion is reduced, or the threshold for classification determination is increased. In addition, the display order of the classification frames can be changed in the order of activity.

  The present invention can be realized not only as an apparatus or a system but also as a method. Of course, a part of the invention can be configured as software. Of course, software products used to cause a computer to execute such software are also included in the technical scope of the present invention.

  These and other aspects of the invention are set forth in the appended claims and will be described in detail below with reference to examples.

  According to the present invention, the classification criteria can be adjusted afterwards using the user action when the user uses the document classification device.

  Examples of the present invention will be described below.

  FIG. 1 shows the overall configuration of a document classification apparatus according to an embodiment of the present invention, and FIGS. 2 and 3 show examples of document classification operation and user action monitoring operation. In this example, the document management apparatus is mounted on the computer 100, but the present invention is not limited to this.

  In FIG. 1, a document classification apparatus includes a document reception unit 10, a document classification unit 11, a keyword storage unit 12, a classified document storage unit 13, a document management unit 14, a U / I (user interface) unit 15, a display unit 16, It includes a user action monitoring unit 17, a classification reference adjustment unit 18, and the like.

  The document reception unit 10 continuously receives documents or collects documents regularly. The document classification unit 11 classifies the document received by the document reception unit 10 based on a preset classification frame criterion (a list of keywords, which may be weighted). The keywords are stored in the keyword storage unit 12 for each classification frame. The classification frame is as shown in FIG. 4, for example, and a plurality of keywords are registered in each classification frame. For example, in the classification frame “baseball”, keywords of “home run”, “double hit”, “hit”, and “home run” are registered. The classified documents are stored in the classified document storage unit 13.

  There may be one classification frame, and there may be one classification frame keyword. Classification may be performed using word vectors instead of keywords.

  The display unit 16 displays a list of documents classified into one classification frame, a text display of a selected document, and the like. The U / I unit 15 receives a user instruction input via a pointing device or a keyboard. This instruction is transmitted to the document management unit 14 and the user action monitoring unit 17.

  Based on an instruction input from the user, the document management unit 14 displays the document list of the designated classification frame and the content of the designated document on the display unit 16. FIG. 5 shows a display example of the document list (classification by the classification frame of “China market” in FIG. 4). The user can determine the document to be opened by looking at the title, summary, etc. from the list displayed. The document to be opened can be specified using a pointing device or the like.

  For example, the user action monitoring unit 17 analyzes the content text of the opened document (or its summary text) and records the frequency of the important phrase. For example, when the documents [1] and [3] in FIG. 5 are opened, as shown in FIG. 6, the important phrases are extracted and their frequencies are registered. In the example of FIG. 6, when the appearance words / phrases are divided and registered, the frequency is apportioned. Note that the feature amount (vector space) of the content text may be extracted and accumulated.

The classification standard adjusting unit 18 continuously records the frequency of important words and phrases, and keywords that have high importance over a long period of time and keywords that become high in importance over time are used as criteria for determining the classification frame. Add to. On the other hand, keywords included in documents that have not been opened are less important, and if they are not important in the long run, they are removed from the criteria. Specifically, the degree of fixation of the appearance frequency of the extracted important words is determined. For example, when the long-term behavior is monitored, the following criterion is adopted, where t (elapsed time after starting monitoring)> T (minimum monitoring time), k is a constant, and f is an appearance frequency. It can be registered as a keyword when the F of the user exceeds a predetermined threshold.

  In the example of FIG. 7, the threshold value is exceeded at the time of the triangle mark, and “bird flu” is adopted as a keyword.

  In addition, keywords that have already been used in the classification frame are also subject to excessive user interest by, for example, performing the long-term monitoring described above, and lowering the weight assigned to the keywords when they are below the threshold, compared to the time of adoption. If it does not exist, it may be excluded from the classification criteria.

  Reference is now made to FIGS.

  FIG. 2 shows an example of operation during document classification. In FIG. 2, the document receiving unit 10 registers the document (S10), and the sentence disposal unit 1 classifies the document with reference to the keyword list for each classification frame and stores it in the classified document storage unit 13 (S11).

  FIG. 3 shows an example of a user action monitoring operation. In FIG. 3, a classification frame is designated and a document list of the classification frame is displayed (S20). If there is an instruction to close the document list, the process is terminated, and if there is no instruction, it is monitored whether the document is opened (S21, S22). When the document is opened, the appearance frequency of the phrase is recorded (S23). Thereafter, an adoption standard for determining whether or not to adopt as a keyword is determined, and if the adoption standard is satisfied, it is added as a keyword for the classification frame (S24, S25).

  According to this embodiment, keywords can be added afterwards based on the user action, and the classification criteria can be adjusted.

  The present invention is not limited to the above-described embodiments, and various modifications can be made without departing from the spirit of the invention. For example, as shown in FIG. 8, the document classification apparatus may be configured as a server / client system. In FIG. 8, portions corresponding to those in FIG. In the example of FIG. 8, an operation request is sent from the U / I unit 21 of the client computer 300 to the request receiving unit 19 of the server computer 200, and based on this, data for document list display and document body display is displayed. 20 to the client computer 300. The display unit 22 of the client computer 300 displays a document list based on this display data. Also in this example, the user action monitoring unit 17 monitors the operation request of the request receiving unit 19 and records the user action.

  In the above example, the classification criteria based on keywords have been mainly described. However, the classification criteria can be adjusted in the same manner even when word vectors are used as the classification criteria. The word vector may be adjusted by recording. tf * IDF may be used.

  The user action to be monitored is not limited to the document open process, and may be a copy storage, a recommendation process for other users, a comment or annotation addition process, and the like. You may consider the time the document is open.

  Further, not only the classification standard (keyword) but also the view of the activity of the classification frame reflects the user's action. A classification frame having a low activity level can be used to make it difficult to classify a document during document classification, or can be used for the ranking of a classification frame ranking list. The degree of activity can be statistically obtained based on, for example, the frequency of displaying the document list of the classification frame, the open processing of the document being displayed in the classification list, and the like.

It is a figure which shows the structure of the Example of this invention as a whole. It is a flowchart explaining the example of document classification operation | movement of the said Example. It is a flowchart explaining the user action monitoring operation example of the above-mentioned embodiment. It is a figure explaining the example of the classification frame of the above-mentioned Example, and a keyword. It is a figure explaining the example of a display of the document list of the above-mentioned Example. It is a figure explaining the example of a recording of the user action of the above-mentioned Example. It is a figure explaining the example of monitoring of the long-term behavior of the above-mentioned example. It is a figure explaining the modification of the above-mentioned Example.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Document reception part 11 Document classification | category part 12 Keyword storage part 13 Document storage part 14 Document management part 15 U / I part 16 Display part 17 User action monitoring part 18 Classification reference adjustment part 19 Request reception part 20 Display data transmission part 21 U / I unit 22 display unit 100 computer 200 server computer 300 client computer

Claims (14)

A document receiving means for receiving a document;
Document classification means for classifying the documents received by the document reception means according to the classification criteria;
Classification result display means for displaying the classification result by the classification means;
User action monitoring means for monitoring a user action input based on the classification result displayed on the classification result display means;
A document classification apparatus comprising: a classification reference adjusting unit that adjusts the classification reference based on a user action monitored by the user action monitoring unit.   2. The document classification apparatus according to claim 1, wherein the classification standard is a keyword registered for classification and a weight attached to the keyword.   3. The document classification apparatus according to claim 2, wherein the user action monitored by the user action monitoring means is a browsing operation for the classified document.   4. The document classification apparatus according to claim 3, wherein the user action monitoring means monitors browsing time.   2. The document classification apparatus according to claim 1, wherein the user action monitored by the user action monitoring means is a copy storage for the classified document.   4. The document classification apparatus according to claim 2, wherein the classification standard adjusting unit adjusts the classification standard by adding or deleting keywords or increasing or decreasing weights assigned to the keywords.   7. The document classification apparatus according to claim 6, wherein a keyword to be added or an increase / decrease in a weight attached to the keyword is determined based on a word / phrase extracted from the document related to the user action monitored by the user action monitoring means.   The document classification apparatus according to claim 1, wherein the user action is addition of a comment to the classified document.   The document classification apparatus according to claim 1, wherein the classification standard is a word vector set for each classification. A step of receiving a document by the document receiving means;
Classifying the document received by the document receiving means by the document receiving means according to the classification criteria;
A step in which the classification result display means displays the classification result by the classification means;
A step in which the user action monitoring means monitors a user action input based on the classification result displayed on the classification result display means;
And a step of adjusting the classification standard based on a user action monitored by the user action monitoring unit. A step of receiving a document by the document receiving means;
Classifying the document received by the document receiving means by the document receiving means according to the classification criteria;
A step in which the classification result display means displays the classification result by the classification means;
A step in which the user action monitoring means monitors a user action input based on the classification result displayed on the classification result display means;
A computer program for classifying a document, characterized in that a classification reference adjusting unit is used to cause a computer to execute a step of adjusting the classification criterion based on a user action monitored by the user action monitoring unit. A document receiving means for receiving a document;
Document classification means for classifying the documents received by the document reception means into a plurality of classification frames according to classification criteria;
A classification result display means for displaying a classification result by the classification means for the specified classification frame;
User action monitoring means for monitoring user actions for each classification frame;
Classification criteria adjusting means for adjusting the classification criteria based on the user action monitored by the user action monitoring means,
Furthermore, the document classification apparatus characterized in that the display order of the classification frame is changed based on the user action monitored by the user action monitoring means. A step of receiving a document by the document receiving means;
The document classification means classifying the document received by the document reception means into a plurality of classification frames according to the classification criteria;
A step in which the classification result display means displays the classification result of the designated classification frame by the classification means;
A user action monitoring means for monitoring a user action for each classification frame;
A classification reference adjusting means adjusting the classification reference based on the user action monitored by the user action monitoring means;
Furthermore, the display order of the said classification frame is changed based on the user action monitored by the said user action monitoring means, The document classification method characterized by the above-mentioned. A step of receiving a document by the document receiving means;
The document classification means classifying the document received by the document reception means into a plurality of classification frames according to the classification criteria;
A step in which the classification result display means displays the classification result of the designated classification frame by the classification means;
A user action monitoring means for monitoring a user action for each classification frame;
And a step of adjusting the classification criteria based on the user action monitored by the user action monitoring means, and the user action monitored by the user action monitoring means. A computer program for classifying a document, which has a function of changing the display order of the classification frame based on the above.

Images