JP2018207270A - Non-structured fax document classification system - Google Patents

Abstract

To provide a non-structured FAX document classification system capable of almost certainly performing automatic classification of a non-structured FAX document using an OCR result (including a candidate character and attribute information such as reliability).SOLUTION: A non-structured FAX document classification system comprises: a FAX reception function for receiving a non-structured document that is not a document previously structured for OCR processing, using FAX; a classification rule setting function for setting a classification rule for classifying a received non-structured FAX document in units of a person or in units of an item; an automatic classification function for applying the classification rule set as a method for classification to classify the received non-structured Fax document in units of a person or in units of an item; and a storage and output function for storing or outputting classified unit document group in units of a person or in units of an item.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a classification system that automatically classifies a faxed document (particularly an unstructured document that is not a prestructured document for OCR processing).

Japanese Patent Application Laid-Open No. H10-228688 discloses a FAX document classification apparatus that can accurately and efficiently classify received FAX documents.
More specifically, the apparatus classifies a FAX document by associating a keyword corresponding to identification information included as a part of the FAX document with each group of FAX documents received in one communication. Device. In this apparatus, a group of FAX documents is displayed one by one in the image display area 121 on the display screen 120, and each displayed FAX document is displayed as a keyword in the character input area 122 while being displayed. Is stored, and the correspondence between the input keyword and the image file name that identifies each FAX document to be displayed is stored.

JP 2001-282814 A

In such a FAX document classification apparatus, the user inputs “keyword” and “priority” for each received FAX document before classifying the FAX document. Certainly, such a device may be able to accurately classify received fax documents.
However, it is absolutely necessary for the user to input data such as “keyword” and “priority order”, and humans must specify information related to the FAX document.

If humans are allowed to specify information related to a FAX document, efficiency will be hindered. On the other hand, it is possible to read a document by using an OCR (optical character reader) and specify the contents described in the document.
However, FAX documents have poor image quality and are difficult to read accurately with OCR. Even if it is a FAX document, it may be possible to read it with OCR to some extent if the document is structured in advance so as to perform OCR processing. It is difficult to prove that it is completely accurate.

  On the other hand, the inventor of the present application does not expect the completeness of the OCR result for the received unstructured FAX document, and even if there is a part that could not be read by OCR, it remains as it is (unspecified and remains ambiguous) In order to provide a fully automated unstructured FAX document classification system that does not require human identification, the present invention has been completed.

In order to achieve the above object, the first invention provides a FAX reception function for receiving an unstructured document that is not a previously structured document for OCR processing by FAX, and the received unstructured FAX document is received by a person. A classification rule setting function that sets a classification rule for classifying into units or cases, and an unstructured FAX document received by applying a classification rule set as a method for classification into people or cases An unstructured FAX document classification system comprising an automatic classification function for classification and a storage output function for storing or outputting a unit document group for each classified person unit or project unit. Here, the FAX reception function does not exclude reception of a structured document pre-structured for OCR processing or a semi-structured semi-structured document.
The second invention calculates the similarity between each page of the unstructured FAX document from the recognition result obtained by performing OCR processing on the unstructured FAX document according to the classification rule as a method for classification in the automatic classification function, This unstructured FAX document classification system is characterized in that classification is performed using a classification algorithm based on similarity.
According to a third aspect of the present invention, there is provided the unstructured FAX document classification system in which an unclassified FAX document having different incoming calls is classified into a person unit or a project unit with the automatic classification function as an object.
According to the fourth aspect of the present invention, a unit document group that is missing or missing due to a lack of a necessary unstructured FAX document is stored in a supplementary list in a folder using a list output or storage output function, and later using an automatic classification function. The unstructured FAX document classification system is provided with an incomplete / additional document management function for classifying incoming unstructured FAX documents into unit document groups that are listed or stored in a folder.
In the fifth aspect of the present invention, a unit document group having an unstructured FAX document received at a time difference is stored in a waiting list by a list or a storage output function, and is stored later by using an automatic classification function. The unstructured FAX document classification system includes a standby document management function for classifying structured FAX documents into unit document groups that are listed or stored in a folder. Here, the sender of the previous unstructured FAX document and the sender of the subsequent unstructured FAX document received at a time difference include the same case as well as different cases.

The present invention has the following effects.
(1) Even if it is an unstructured FAX document, by applying the classification rules set in the classification method, it is automatically classified into human units or project units without requiring human identification. be able to.
(2) In other words, a method (Levenstein distance, Hamming distance, etc.) and clustering for calculating the similarity as a distance based on the result of reading an unstructured FAX document by OCR (including attribute information such as candidate characters and reliability) By using (classification) techniques (graph theory, nearest neighbor search, neural network, etc.), even if the recognition result of OCR processing is insufficient (fuzzy), it is almost automatically classified it can.
(3) In particular, when classifying into individual units or project units, “kanji” is important, but in the case of kanji, it is difficult to completely match. Therefore, it is possible to automatically classify by confirming the similarity based on the elements of the unstructured FAX document, the address, the contract number, etc., and classifying it with a predetermined algorithm (clustering).
(4) In addition, the present invention enables classification not only between a group of unstructured FAX documents received at one incoming call but also between unstructured FAX documents with different incoming (time), The range of utilization expands.
(5) One of them is management of incomplete documents. When an incomplete document (missed document) is sent later by FAX, the unit document group can be completed by sorting the incomplete document into unit document groups that have already been classified.
(6) The other is management of standby documents. When a waiting document (for example, a document sent with a time difference such as a document relating to a delivery date or an estimate) is sent later by FAX, the standby document is sorted into unit document groups that have already been classified. A complete set of documents.
(7) The invention of the present application can be classified automatically almost certainly even if the recognition result obtained by OCR processing is insufficient (ambiguous). Even in the case where a part of the information to be written is hidden characters, an unstructured FAX document including the hidden characters can be automatically classified. In this case, by having a master list with no hidden characters, even if the recognition result obtained by the OCR process is insufficient (obvious), it can be automatically and reliably classified.

The system conceptual diagram (1) of this invention. The system conceptual diagram (2) of this invention. Example of use of the present invention (1) Example of use of the present invention (2) Example of use of the present invention (3) Example of use of the present invention (4)

Embodiments of the present invention will be described with reference to the drawings.
1 and 2 are conceptual diagrams of an unstructured FAX document classification system according to the present invention.
In FIG. 1, “automatic classification function” and “classification rule” are mainly illustrated.
First, “automatic classification processing” is based on the recognition result obtained by reading an unstructured FAX document received by the FAX reception function using OCR, and calculates the similarity for each page, and uses a classification method. Classify by unit or project. Here, the similarity is calculated by quantifying the similarity between pages as a distance, and classification is performed using a predetermined algorithm as a classification method.

Next, in order to operate the “automatic classification function”, a “classification rule” for classifying the received unstructured FAX document into a person unit or a project unit is set as a premise (classification rule setting function).
The classification rule includes, for example, definition of identification information (identification information is defined according to the purpose of the document to be classified). Specifically, if it is an item name, a character string such as “name”, “date of birth”, “address”, “subject”, and “contract number” is set, and the character string following these character strings is acquired as a value. To do. It also has a synonym dictionary, where the name is matched to a character string such as (name, good name, surname), and if it is an address (such as a registered address, delivery address), etc. Can do.
The value format can be specified, and if it is a date of birth, the value format such as 9999/99/99/9999/99/99 (where 9 is a number), contract number If so, specify it in a format like AAA-99999999 (three letters of the alphabet followed by a hyphen and eight digits). If the value formats match, the value can be acquired even if the item name is omitted.
In addition, when an attached document such as a certificate or a receipt does not include a defined item name, it may be regarded as an integral part of the page including the immediately preceding item name and classified into the same group.
Further, sender attribute data (such as TSI code or FAX number) received by FAX from the sender may be incorporated into the classification rule.

In order to supplement the “automatic classification function”, a “classification correction process” may be provided. The “classification correction process” is to correct the classification of unit document groups classified by the automatic classification function. Although not shown, an unstructured FAX document that has been classified as “others” by the automatic classification function is classified into a unit document group by this classification correction function. This ensures the classification.
In addition, by providing a FAX document display function that displays automatically classified FAX documents in unit document groups for each person or project unit, it is possible to visually display and efficiently perform confirmation and correction. You may provide the user interface which can ensure the reliability of the classification | category of a FAX document by processing according to the defined input procedure.

FIG. 2 mainly illustrates the “time difference automatic classification function” that enables time difference classification.
On the application date, the unit document group A has all the necessary documents. On the other hand, in the unit document group B, the document of B2 is insufficient. When the document of B2 is received N days after the application date, B2 is classified into unit document group B by the time difference automatic classification function, and unit document group B is completed.

3 to 6 introduce examples of using the unstructured FAX document classification system of the present invention.
First, FIG. 3 and FIG. 4 exemplify ambiguous classification of a person unit or a case unit when a plurality of persons or unstructured FAX documents of a plurality of cases are mixed in one incoming call.
FIG. 3 assumes a case of receiving monthly processing of a care service offer slip. The service provision slips and separate tables of a plurality of care service users (hereinafter referred to as “users”) held by the care manager are paired and transmitted by FAX. It is made a bundle of user units (person unit) by the insured person number or insured person name of the received FAX document. In recent years, there have been cases where part of the insured person name and insured person number is masked with hidden characters from the viewpoint of personal information protection. Even such cases can be automatically classified. In this case, by having an insured master list without obfuscation, even if the recognition result obtained by OCR processing is insufficient (fuzzy), it is more automatically classified. Can do.

  FIG. 4 assumes a case of receiving a response to an overpaid case payment result inquiry of a lawyer office (transmission source: inquiry financial institution, reception source: lawyer office). In order to process daily overpaid claims, we ask each financial institution for payment results and get a reply. In many cases, financial institutions return the results of a plurality of people in one incoming call. For the purpose of sending and receiving personal information, when a call is received, it is necessary to mutually confirm whether the number and number of cases have arrived correctly, and it is necessary to immediately sort (number of pieces transferred per person).

Next, FIGS. 5 and 6 exemplify a case where unstructured FAX documents with different incoming calls are sorted by collating with a complementary list or a storage folder.
Figure 5 (Defective / Additional Document Management Function) assumes the case of additional transmission from the same sender of defective documents at the time of a life insurance / non-life insurance accident notification or payment request (sender: insurance agent, reception) Original: insurance company). When requesting payment with life insurance / non-life insurance, various attached documents will be sent to prove the reason for the claim. However, there are many cases in which certificates are not prepared at a time and are incomplete, and requests for transmission of additional certificates are requested. In recent years, insurance companies have been strongly required to process payments without delay, and it has become essential to improve the efficiency of document reception and speed of confirmation of defects. Therefore, by classifying the unstructured FAX document (document name: medical certificate) additionally received into the unit document group (Group01) stored in the folder waiting for deficiency, the unit document group (Group01) is automatically completed. The

  FIG. 6 (standby document management function) assumes a case in which a time difference transmission from a different sender that links a delivery date reply from a manufacturer to an order form from an end user (sender of standby document: end user) Response sender: manufacturer, receiver: wholesale company). In this case, after processing the order form from the end user, the order is processed with the manufacturer. Wait for the delivery date response from the manufacturer, and send the delivery date to the end user. For this purpose, the delivery date reply from the manufacturer that is sent (replyed) is linked to the corresponding item in the order form that is listed (standby) in the waiting list (answer waiting list). In this case, the sender of the received unstructured FAX document is different, but this standby document management function can also be applied to the case where the sender is the same.

The present invention can be widely used in business using FAX.
For example, it can be used for the following purposes.
(1) Nursing care-Service provision table-Service provision table (separate table)
(2) Non-life insurance (transfer from other companies)
・ Auto insurance application form ・ Current insurance insurance policy ・ Car verification ・ Driver's license (3) Overpayment claim ・ Overpayment return invoice ・ Transaction history (4) Submission of My Number related documents to the company・ My number registration application (cover)
・ Notification card (My number is listed)
・ Driver's license (identification document)

  In addition, due to concerns about information leakage, such as the Personal Information Protection Law, there are many cases where some of the information written in the document is hidden characters (for example, ■ or *), but for unstructured FAX documents that contain such hidden characters, The present invention can be effectively utilized.

Claims (5)

A FAX receiving function for receiving an unstructured document that is not a pre-structured document for OCR processing by FAX;
A classification rule setting function for setting a classification rule for classifying the received unstructured FAX document into a person unit or a project unit;
An automatic classification function that classifies unstructured FAX documents received by applying the classification rules set in the method for classification into human units or project units;
Save output function to save or output the unit document group for each classified person unit or project unit,
An unstructured FAX document classification system characterized by comprising:   As a method for classification in the automatic classification function, the similarity between each page of the unstructured FAX document is calculated from the recognition result obtained by performing OCR processing on the unstructured FAX document in accordance with the classification rule, and the classification is determined from the similarity. 2. The unstructured FAX document classification system according to claim 1, wherein classification is performed using an algorithm.   3. The unstructured FAX document classification system according to claim 1 or 2, wherein the unclassified FAX document having different incoming calls is classified into a person unit or a project unit by the automatic classification function.   For undocumented unit documents with missing / unstructured FAX documents that are missing, add them to the supplement list, save them in a folder using the list output or save output function, and use the automatic classification function to receive unstructured data later 4. The unstructured FAX document classification system according to claim 3, further comprising an incomplete / additional document management function for classifying FAX documents into unit document groups that are listed or stored in a folder.   For unit document groups with unstructured FAX documents that are received with a time difference, list them in the waiting list using the list or save output function, and list the unstructured FAX documents that are received later using the automatic classification function 4. The unstructured FAX document classification system according to claim 3, further comprising a standby document management function for classifying into unit document groups stored up or in folders.

Images