WO2017009900A1 - Document processing system and document processing method - Google Patents

Abstract

In this document processing system: a learning dictionary having registered therein a physical structure of an elements included in a paragraph that should be extracted from document data and an information aggregation present in a region of the element is stored in a storage region; and a processor divides the document data into at least one region, acquires an information aggregation present in the divided region, calculates the level of physical structure agreement indicative of the level of agreement between the region in which the acquired information aggregation is present and the physical structure of the element registered in the learning dictionary, calculates the level of information aggregation agreement indicative of the level of agreement between the acquired information aggregation and the information aggregation registered in the learning dictionary, determines whether or not the divided region is the paragraph that should be extracted on the basis of the calculated level of physical structure agreement and level of information aggregation agreement, and registers the acquired information aggregation in the learning dictionary when the divided region is the paragraph that should be extracted.

Description

Document processing system and document processing method

The present invention relates to a document processing system that extracts information from document data.

Document data has a list with at least one element as one set. For example, a delivery specification or the like includes a plurality of elements such as a code number, a product name, and a specification. There is a demand to extract information described in such a list and create a database for checking inventory of goods or managing defect information.

Here, there are structured documents and unstructured documents in the database. A structured document is a document in which logical elements such as a document title, chapter, body, table, and figure are described in a predetermined structure. In such a structured document, it is only necessary to extract logical element information according to the structure of the logical element, and it is easy to extract elements of the structured document.

On the other hand, unlike structured documents, unstructured documents do not have logical elements described in a predetermined structure. When extracting elements of such an unstructured document, it is necessary to analyze the text of the unstructured document.

Japanese Patent Laid-Open No. 11-250041 (hereinafter referred to as Cited Document 1) is a technique for extracting information on each element of an unstructured document. This publication describes "means 1 for extracting layout objects and structures from document images, and means for extracting logical objects such as paragraphs, lists, mathematical formulas, programs, and annotations based on typography from a text region extracted from a document image. 3, means 5 for extracting a plurality of possible reading orders between objects, and means 4 for extracting a logical structure by applying a pre-defined model to the logical object. Extracting primary information and secondary information from various multi-page documents composed of, graphics, tables, etc., and converting them into various electronic formats enables automatic construction of document management systems and various computer applications Makes it possible to effectively use the system ”(see summary).

Japanese Patent Laid-Open No. 11-250041

The technique described in Patent Document 1 needs to define a layout model for extracting information from a document image, a key word, and the like for each document.

Accordingly, the present invention provides a document processing apparatus that extracts information for each element from an unstructured document without defining a layout model for extracting information from a document image and key words for each document in advance. The purpose is to provide.

A typical example of the present invention is a document processing system that extracts an information set from input document data, and includes a processor and a storage area, and the storage area should be extracted from the document data. A learning dictionary in which a physical structure of an element that is included in a paragraph and where information to be extracted exists and an information set that exists in the area of the element are registered is stored, and the processor receives the input Information structure existing in at least one area of the document data obtained, and a physical structure matching degree indicating a matching degree between the existence area of the acquired information set and the physical structure of the element registered in the learning dictionary. And calculating an information set matching degree indicating a matching degree between the acquired information set and the information set registered in the learning dictionary, and based on the calculated physical structure matching degree and information set matching degree. Determining whether or not the region is the paragraph to be extracted, and when the region is the paragraph to be extracted, the acquired information set is set as the physical structure match degree and the information set match degree. The information set registered in the learning dictionary as an information set existing in the element region corresponding to the calculated information set, and the acquired information set corresponds to the element of the information set for which the physical structure match degree and the information set match degree are calculated. The attached output information is stored in the storage area.

The following is a brief description of the effects obtained by the representative inventions disclosed in the present application. That is, it is possible to provide a document processing apparatus that extracts information for each element from an unstructured document without defining in advance for each document a word or the like that is a key for extracting information from document data.

Problems, configurations, and effects other than those described above will be clarified by the following description of the embodiments.

It is a block diagram of the document processing system of a present Example. It is explanatory drawing of the outline | summary of the paragraph designation | designated reception process of a present Example. It is explanatory drawing of the outline of the document processing of a present Example. It is explanatory drawing of the block dictionary of a present Example. It is explanatory drawing of the repetition table of a present Example. It is a flowchart of the outline of the paragraph designation | designated reception process of this Example, and the outline of document processing. It is a flowchart of the paragraph designation reception process of a present Example. It is a flowchart of the repetition part extraction process of a present Example. It is a flowchart of the repetition candidate part extraction process of a present Example. It is a flowchart of the element matching determination process of a present Example. It is explanatory drawing of the process which divides | segments the area | region of the paragraph designated by the user of a present Example using the XY cut method. It is explanatory drawing of the process which divides | segments the whole area | region of the document data of a present Example using an XY cut method. It is a flowchart of a coincidence degree calculation process of a present Example. It is explanatory drawing of the process which divides | segments the whole area | region of the document data in the document process of a present Example using an XY cut method. It is explanatory drawing of the repetition partial display screen of a present Example. It is explanatory drawing of the process result display screen of a present Example.

FIG. 1 is a block diagram of the document processing system 100 of this embodiment.

The document processing system 100 extracts an information set included in input document data (for example, a document image) for each element. The information set is a concept including a character string and a graphic to be extracted.

The document processing system 100 includes a processor 101, a memory 102, a secondary storage device 103, an input interface (IF) 104, an output interface (IF) 105, and a network interface 106. These are connected to each other via a bus 107.

The processor 101 executes various arithmetic processes. The secondary storage device 103 is a non-volatile non-transitory storage medium, and stores various programs and various data. The memory 102 is a volatile temporary storage medium. The memory 102 is loaded with various programs and various data stored in the secondary storage device 103, and the processor 101 executes the various programs loaded in the memory 102. Then, various data loaded in the memory 102 are read and written.

Input devices such as a keyboard and a mouse are connected to the input IF 104, for example. For example, an output device such as a display is connected to the output IF 105. The network IF 106 is an interface that connects the document processing system 100 to a network.

The processor 101 includes a paragraph designation reception processing unit 111 and a document processing unit 112.

The paragraph designation reception processing unit 111 executes a paragraph designation reception process. In the paragraph designation reception process, in a certain document data, the designation of a paragraph and the designation of an element in the paragraph are accepted from the user, the information about the paragraph and the element for which the designation is accepted is registered in the block dictionary 121, and information existing in the element is registered. Register in the word dictionary 122. Also, in the paragraph designation receiving process, an area (repeated portion) where the designated paragraph is repeated in the document data is extracted, and information regarding the extracted repeated portion and elements included in the repeated portion is registered in the output DB 123. The information set existing in the element is registered in the word dictionary 122. Details of the paragraph designation receiving process will be described with reference to FIGS. 6 and 7 to 12.

The document processing unit 112 executes document processing. In the document processing, referring to the block dictionary 121 and the word dictionary 122, the paragraph designation is accepted from the area corresponding to the paragraph whose designation is accepted in the paragraph designation accepting process from the document data different from the document data on which the paragraph designation accepting process is executed. The element corresponding to the element that has received the designation in the process is extracted, and information existing in the extracted element is extracted. Details of the document processing will be described with reference to FIGS.

It should be noted that programs corresponding to the paragraph designation reception processing unit 111 and the document processing unit 112 are stored in the memory 102, and the processor 101 executes these programs to cause the processor 101 to execute the paragraph designation reception processing unit 111 and the document processing unit. 112 is implemented.

The memory 102 stores a block dictionary 121, a word dictionary 122, an output database (DB) 123, and a repetition table 124.

In the block dictionary 121, a physical structure of a paragraph including an element to be extracted and a physical structure and a logical structure of an element included in the paragraph are registered. The physical structure of the paragraph is information that can specify the area of the paragraph, for example. The physical structure of the element is position information within the paragraph of the element, and the logical structure of the element is, for example, a notation format and attribute of information existing in the element. Details of the block dictionary 121 will be described with reference to FIG.

In the word dictionary 122, an element and an information set (word, figure, etc.) existing in the element are registered in association with each other. In the output DB 123, elements extracted from the document data and information sets existing in the elements are registered in association with each other. In the repetition table 124, information related to the repeated portion of the paragraph whose designation is received in the paragraph designation receiving process is registered. Details of the repetition table 124 will be described with reference to FIG.

FIG. 2 is an explanatory diagram outlining the paragraph designation receiving process of this embodiment.

First, in the paragraph designation receiving process, the paragraph designation receiving processing unit 111 receives a paragraph designation to be extracted from the input document data from the user (201). The paragraph specification is the specification of the area of the paragraph to be extracted. When the paragraph is rectangular, for example, the upper left coordinate and the lower right coordinate are specified, and when the paragraph is a polygon, the coordinates of each vertex of the polygon are specified. Is specified.

When the paragraph designation acceptance processing unit 111 accepts the paragraph designation, it registers the physical structure of the designated paragraph in the block dictionary 121 (202).

The paragraph specification reception processing unit 111 receives specification of elements included in the specified paragraph (203). The element designation is the designation of the area of the element to be extracted. When the element is rectangular, for example, the upper left coordinate and the lower right coordinate are designated, and when the element is a polygon, the coordinates of the polygon vertex are designated. It is specified.

When the paragraph designation reception processing unit 111 accepts the element designation, the paragraph designation reception processing unit 111 registers the physical structure of the designated element in the block dictionary 121. In addition, the paragraph designation reception processing unit 111 extracts an information set existing in the designated element, specifies the notation format and attribute of the element based on the extracted information set, and logically calculates the notation format and attribute of the identified element. The structure is registered in the block dictionary 121. In addition, the paragraph designation reception processing unit 111 registers the extracted information set in the word dictionary 122 and the output DB 123 (204). In FIG. 2, the information set existing in the element is a character string such as “055-3223”, “handbag”, and “in this bag...”. In addition to a character string, there may be a figure or the like, and the character set is a concept including a character string and a figure.

After execution of the process 204, “055-3223” is registered in the element 1 of the word dictionary 122, and “handbag” is registered in the element 2. In addition, “055-3223” is registered in element 1 of the output DB 123, “handbag” is registered in element 2, and “in this bag...” Is registered in element 3.

Next, the paragraph designation reception processing unit 111 extracts a repeated portion of the designated paragraph based on the physical structure and logical structure of the designated paragraph registered in the block dictionary 121 (205). The repetitive part is an area including an element similar to the physical structure of the designated paragraph and similar to the physical structure of the designated element. In FIG. 2, a region relating to “boots” and a region relating to “gloves” located below the designated paragraph are extracted as repeated portions.

The paragraph designation reception processing unit 111 extracts an information set existing in each element included in the extracted repeated portion, and registers the extracted information set in the word dictionary 122 and the output DB 123 (206).

After execution of the process 206, “015-0001” and “015-0149” are added to the element 1 of the word dictionary 122, and “boots” and “gloves” are added to the element 2. In addition, “015-0001” and “015-0149” are added to the element 1 of the output DB 123, “boots” and “gloves” are added to the element 2, and “vinyl made ... "And" Use yarn ... "are added.

This allows the user to specify a single paragraph and specify elements included in the specified paragraph, so that repeated parts of the paragraph are extracted, so that it is not necessary to specify the repeated parts one by one. In addition, since the information of the extracted elements of the repeated portion is automatically registered in the word dictionary 122, the user can omit the work of building the word dictionary 122.

FIG. 3 is a schematic explanatory diagram of document processing of the present embodiment.

In the document processing, an information set existing in the designated element is extracted from document data different from the document data processed in the paragraph designation receiving process.

First, the document processing unit 112 extracts the extraction area candidate portion of the paragraph registered in the block dictionary 121 from the input document data (301). The extraction region candidate portion is a region whose physical structure is similar to the designated paragraph.

Next, the document processing unit 112 refers to the block dictionary 121 and the word dictionary 122 to calculate the total matching degree between the specified paragraph and the extraction area candidate part, and if the calculated total matching degree is equal to or greater than a threshold value. Then, an information set existing in the element included in the extraction area candidate portion is extracted (302). Note that the total matching degree is the degree of matching between the position information in the paragraph of the element included in the specified paragraph and the position information in the extraction region candidate portion of the information set included in the extraction region candidate portion, and is included in the specified paragraph. The degree of coincidence between the information set corresponding to the specified element and the information set included in the extraction area candidate part, the notation format corresponding to the element included in the specified paragraph, and the notation format of the attribute and information set included in the extraction area candidate part And the degree of coincidence with the attribute.

Then, the document processing unit 112 registers the information set extracted in the process 302 in the word dictionary 122 and the output DB 123 (303).

The “folding umbrella” of the document data shown in FIG. 3 is not registered in the word dictionary 122 registered in FIG. 2, but since the total matching degree of the area of “folding umbrella” is equal to or greater than the threshold, “folding umbrella” An information set existing in each element of the area is extracted. “331-0120” is added to element 1 of word dictionary 122, and “folding umbrella” is added to element 2. Further, “331-0120” is added to element 1 of the output DB 123, “folding umbrella” is added to element 2, and “one push...” Is added to element 3.

In this way, the user can extract a set of information existing in elements to be extracted from other document data only by specifying paragraphs and elements. Can be extracted.

FIG. 4 is an explanatory diagram of the block dictionary 121 of this embodiment.

The block dictionary 121 includes a paragraph table 400, an element table 410, and a coordinate table 420.

The paragraph table 400 includes an upper left coordinate 401, a lower right coordinate 402, a partial area number 403, an element number 404, and an element pointer 405.

In the upper left coordinate 401, the upper left coordinate of the paragraph designated by the user is registered. In the lower right coordinate 402, the lower right coordinate of the paragraph specified by the user is registered. In the partial area number 403, the number of areas included in the paragraph designated by the user is registered. In the element number 404, the number of elements specified by the user in the paragraph specified by the user is registered. In the element pointer 405, a pointer to each element table 410 is registered.

When the paragraph specified by the user is a polygon, the paragraph table 400 registers the coordinates of each vertex of the polygon instead of the upper left coordinates 401 and the lower right coordinates 402.

The element table 410 is included in the block dictionary 121 by the number of elements registered in the element number 404 of the paragraph table 400. The element table 410 includes polygon coordinates 411, center coordinates 412, a notation format 413, and attributes 414.

In the polygon coordinates 411, a pointer to the coordinate table 420 in which the coordinates of the vertex of the element designated by the user are registered. Registered in the center coordinate 412 is the center coordinate of the element designated by the user. The center coordinates of this element are coordinates with the center coordinate of the paragraph including the element as the origin. The position is not limited to the center coordinates as long as the position of the element in the paragraph can be specified.

In the notation format 413, the notation format of the information set existing in the element designated by the user is registered. The notation format is expressed as “NNN-NNNN”, “AAAAAAAAA”, “Text” or the like. “N” in “NNN-NNNN” means an arbitrary single-character number, and “-” means a hyphen. “A” in “AAAAAAAAAA” means an arbitrary alphabet. “Text” means that some text exists as an information set.

In the attribute 414, an attribute existing in the element designated by the user is registered. The attribute is expressed by “English capital letter string”, “text”, “image”, and the like. “English capital letter string” means that the information set existing in the element is a character string composed of uppercase letters. “Text” means that the information set existing in the element is some text. “Image” means that the information set existing in the element is an image.

Note that an information set existing in an element registered in the word dictionary 122 may be registered as one entry in the element table 410. That is, the word dictionary 122 may be included in the block dictionary 121, and the block dictionary 121 and the word dictionary 122 are collectively referred to as a learning dictionary.

FIG. 5 is an explanatory diagram of the repetition table 124 of this embodiment.

The repetition table 124 includes a list table 500, a paragraph table 510, an element table 520, and a coordinate table 530.

The list table 500 includes a repetition number 501 and a pointer 502 to a paragraph table.

In the repetition number 501, the number of areas extracted as repetition parts is registered. In the paragraph table pointer 502, a pointer to the paragraph table 510 in which the information of the area extracted as the repeated portion is registered is registered.

The paragraph table 510 is the same as the paragraph table 400 shown in FIG. 4 except that information related to the region extracted as the repeated portion is registered, and thus description thereof is omitted.

The element table 520 is the same as the element table 410 shown in FIG. 4 except that information related to elements included in the region extracted as the repeated portion is registered, and thus description thereof is omitted. In the coordinate table 530, the coordinates of the vertices of the elements included in the region extracted as the repeated portion are registered.

FIG. 6 is a flowchart of the outline of the paragraph designation receiving process and the outline of the document process of this embodiment.

First, the outline of the paragraph designation receiving process will be described.

The paragraph designation reception processing unit 111 receives input of document data (601). The paragraph designation acceptance processing unit 111 accepts designation of a paragraph by the user from the area of the document data that has been accepted, and accepts designation of an element included in the paragraph that has received the designation (602).

Next, the paragraph designation reception processing unit 111 analyzes the physical structure of the paragraph that has received the designation and the physical structure of the element that has accepted the designation (603). Specifically, the paragraph designation reception processing unit 111 identifies the upper left coordinates and the lower right coordinates of the paragraph for which the designation is received. In addition, the paragraph designation reception processing unit 111 specifies the coordinates of the vertices and the center coordinates of the element of the element that has received the designation.

Next, the paragraph designation reception processing unit 111 analyzes the logical structure of the information set existing in the element that has received the designation (604). Specifically, the paragraph designation reception processing unit 111 identifies the notation format and attributes of the information set existing in the element that has received the designation.

Next, the paragraph designation reception processing unit 111 registers the paragraph and element physical structure analyzed in step 603 and the element logical structure analyzed in step 604 in the block dictionary 121 (605). Specifically, the paragraph designation reception processing unit 111 registers the upper left coordinate of the paragraph specified in the process of step 603 in the upper left coordinate 401 of the paragraph table 400 of the block dictionary 121 and the lower right coordinate 402 of step 603. Register the lower right coordinate of the paragraph specified in the process. In addition, the paragraph designation reception processing unit 111 registers the number of elements whose designation is accepted in the element number 404 of the paragraph table 400, and registers a pointer to each element table 410 in the element pointer 405.

In addition, the paragraph designation reception processing unit 111 registers the coordinates of the vertex of the element specified in the process of step 603 in the polygon coordinates 411 of the element table 410, and the element specified in the process of step 603 in the center coordinate 412. Register center coordinates. The paragraph designation reception processing unit 111 registers the notation format of the element specified in the process of step 604 in the notation format 413 of the element table 410, and registers the attribute of the element specified in the process of step 604 in the attribute 414.

Next, the paragraph designation reception processing unit 111 executes the processing of steps 607 to 610 for all repeated portions (606). First, the paragraph designation reception processing unit 111 extracts an information set of each element of the repeated portion (607). Then, the paragraph designation reception processing unit 111 registers the information set extracted in step 607 in the word dictionary 122 (608), and registers the information set extracted in step 607 in the output DB 123 (609). If the processes of steps 607 to 609 are executed for all the repeated parts, the paragraph designation receiving process is terminated. If the processes of steps 607 to 609 are not executed for all the repeated parts, The processing returns to step 606.

Next, document processing will be described.

First, the document processing unit 112 accepts input of document data (611), and moves to the first page of the input document data (612). Next, the document processing unit 112 extracts character and line segment information from the input document data (613). Next, the document processing unit 112 divides the entire area of the input document data into at least one area using the XY cut method (614). Details of region division using the XY cut method will be described with reference to FIG.

Next, the document processing unit 112 refers to the block dictionary 121 and executes extraction region candidate extraction processing (615). The extraction area candidate extraction process is a process of extracting an area similar to a paragraph registered in the block dictionary 121 as an extraction area candidate.

Next, the document processing unit 112 refers to the block dictionary 121 and the word dictionary 122 and executes a matching degree calculation process (616). The degree of coincidence calculation process calculates the degree of coincidence between the element of the repetition candidate part and the element registered in the block dictionary 121. If the degree of coincidence is equal to or greater than a threshold, the information set existing in the element of the repetition candidate part is extracted. This is a process for registering the extracted information set in the output DB 123 for each element, and details will be described with reference to FIG.

Next, the document processing unit 112 displays a processing result display screen 1600 (see FIG. 16) including the extracted area and the information set area of the extracted element (617), and ends the document processing. Details of the processing result display screen 1600 will be described with reference to FIG.

Next, details of the paragraph designation receiving process will be described with reference to FIGS.

FIG. 7 is a flowchart of the paragraph designation receiving process of this embodiment.

First, the paragraph designation acceptance processing unit 111 accepts an input of an execution request for a paragraph designation acceptance process from a user, activates a user interface for the paragraph designation acceptance process (701), and accepts input of document data (702). Then, the paragraph designation reception processing unit 111 initializes the block dictionary 121 (703).

Next, the paragraph designation reception processing unit 111 receives an input from the user as to whether or not the input document data includes a repeated portion, and whether or not the input received from the user indicates that the document data includes a repeated portion. Is determined (704).

If it is determined in step 704 that the input received from the user indicates that there is no repetitive portion in the document data (704: No), the paragraph designation acceptance processing unit 111 ends the paragraph designation acceptance process.

On the other hand, if it is determined in step 704 that the input received from the user indicates that there is a repeated portion in the document data (704: Yes), the paragraph designation reception processing unit 111 reads the area of the input image data. In step 705, the user designates a paragraph area desired to be extracted. For example, the user designates a paragraph area using a mouse or the like.

Next, the paragraph designation reception processing unit 111 registers the upper left coordinates and lower right coordinates of the paragraph for which the designation has been received in step 705 as the upper left coordinates 401 and the lower right coordinates 402 of the paragraph table 400 (706).

Next, the paragraph designation acceptance processing unit 111 accepts designation of an area of an element that is included in the paragraph area that has been designated in step 705 and that the user desires to extract (707). For example, the user designates an element area using a mouse or the like.

Next, the paragraph designation reception processing unit 111 registers the coordinates of the vertices of the area of the element whose designation is received in the process of step 707 in the coordinate table 420, and sets a pointer to the coordinate table 420 in which the coordinates are registered as the element table. The polygonal coordinates 411 of 410 are registered (708).

Next, the paragraph designation reception processing unit 111 calculates the center coordinates of the area of the element that has been specified in step 707, and registers the calculated center coordinates in the center coordinates 412 of the element table 410 (709).

Next, the paragraph designation reception processing unit 111 extracts an information set existing in the element area for which the designation has been received in the process of step 707, and registers the notation format of the extracted information set in the notation format 413 of the element table 410. Then, the attribute of the extracted information set is registered in the attribute 414 of the element table 410, and the extracted information set is registered in the word dictionary 122 (710).

Next, the paragraph designation reception processing unit 111 divides the paragraph area for which the designation has been received in step 705 into partial areas using the XY cut method, and calculates the number of partial areas (711). The XY cut method is a method in which, for example, an information set continuous in the horizontal direction is extracted from a predetermined area, and the predetermined area is divided based on a vertical distance between two extracted information sets continuous in the horizontal direction. is there. In addition, when extracting the information set which continues in the vertical direction, it divides | segments based on the distance of the horizontal direction of two information sets continuous in the vertical direction. In step 711, the number of areas that can be divided into paragraph areas is calculated as the number of partial areas. A specific example of the processing in step 711 will be described with reference to FIG.

Next, the paragraph designation reception processing unit 111 registers the number of partial areas calculated in the process of step 711 in the number of partial areas 403 of the paragraph table 400 (712).

Next, the paragraph designation reception processing unit 111 registers the pointer to the element table 410 in the pointer 405 to the element of the paragraph table 400 (713).

Next, the paragraph designation acceptance processing unit 111 executes a repeated part extraction process for extracting a repeated part of a paragraph for which designation is accepted from the input document data (714), and ends the paragraph acceptance process. Details of the repeated partial extraction processing will be described with reference to FIGS.

FIG. 8 is a flowchart of the repeated partial extraction process of this embodiment.

First, the paragraph designation reception processing unit 111 moves to the first page of the input document data (801), and initializes the repetition table 124 (802). Next, the paragraph designation reception processing unit 111 extracts information such as characters and line segments from the input document data (803), and uses the XY cut method to extract the region of the input document data using the XY cut method. The area is divided (804). A specific example of the processing in step 804 will be described with reference to FIG.

The paragraph designation reception processing unit 111 executes a repetition candidate part extraction process for extracting a repetition candidate part having a physical structure similar to that of the paragraph area registered in the block dictionary 121 (805). Details of the repetition candidate part extraction processing will be described with reference to FIG.

Next, when the element included in the extracted repetition candidate part matches the element registered in the block dictionary 121, the paragraph designation reception processing unit 111 sets information regarding the repetition part as the repetition part. Element matching determination processing registered in the repetition table 124 is executed (806). Details of the element matching determination processing will be described with reference to FIG.

Next, the paragraph designation reception processing unit 111 displays a repeated part display screen 1500 (see FIG. 15) that displays the extracted repeated part (807), and two or more repeated parts are registered in the repetition table 124. It is determined whether or not (808). Details of the repeated portion display screen 1500 will be described with reference to FIG.

If it is determined in the process of step 808 that two or more repetition parts are registered in the repetition table 124 (808: Yes), the paragraph designation reception processing unit 111 stores the information set of each element included in the repetition part. The extracted information set of each element is registered in the output DB 123, the extracted information set of each element is registered in the word dictionary 122 (809), and the repeated partial extraction process is terminated.

On the other hand, when it is determined in the process of step 808 that two or more repeated parts are not registered in the repetition table 124 (808: No), the paragraph designation reception processing unit 111 ends the repeated part extraction process.

FIG. 9 is a flowchart of the repetition candidate part extraction process of the present embodiment.

First, the paragraph designation reception processing unit 111 selects one area to be processed from the areas divided in the process of Step 804 (901).

Next, the paragraph designation reception processing unit 111 extracts all information sets existing in the area selected in the process of step 901, and calculates the total value of the areas of the extracted information sets as the area of the area (902). .

Next, the paragraph designation reception processing unit 111 determines whether or not the value registered in the partial area number 403 of the paragraph table 400 is 2 or more (903).

When it is determined in step 903 that the value registered in the partial area number 403 of the paragraph table 400 is not 2 or more, that is, the value registered in the partial area number 403 of the paragraph table 400 is 1 (903) : No), the paragraph designation reception processing unit 111 calculates the area of the paragraph area designated by the user based on the coordinates registered in the upper left coordinates 401 and the coordinates registered in the lower right coordinates 402 of the paragraph table 400. Then, it is determined whether or not the area area calculated in step 902 is within a predetermined range from the calculated area area of the paragraph (904).

If it is determined in step 904 that the area calculated in step 902 is not within the predetermined range from the calculated area of the paragraph (904: No), the area selected in step 901 is the user. Since the physical structure is not similar to the area of the paragraph specified by step 901 and the area selected in the process of step 901 does not become a repeat candidate part, the paragraph specification reception processing unit 111 returns to the process of step 901 and returns to steps 901 to 901. An area where the process 912 has not been executed is selected.

On the other hand, if it is determined in step 904 that the area of the region calculated in step 902 is within a predetermined range from the calculated area of the paragraph (904: Yes), the paragraph designation reception processing unit 111 The region selected in the process of step 901 is repeatedly extracted as a candidate part (905).

Next, the paragraph designation reception processing unit 111 determines whether or not the processing in steps 901 to 912 has been executed for all the divided areas of the input image data (906).

If it is determined in step 906 that the processes in steps 901 to 912 have been performed on all the divided areas of the input image data (906: Yes), the paragraph designation reception processing unit 111 The repetition candidate part extraction process is terminated.

On the other hand, if it is determined in step 906 that the processes in steps 901 to 912 have not been executed for all divided areas of the input image data (906: No), the paragraph designation receiving process The unit 111 returns to the process of step 901 and selects an area where the processes of steps 901 to 912 have not been executed yet.

If it is determined in step 903 that the value registered in the number of partial areas 403 of the paragraph table 400 is 2 or more, the paragraph specified by the user includes a plurality of areas. Calculates the area of the region corresponding to the paragraph designated by the user by summing the area of the region selected in the processing of step 901 and the area of the region adjacent to the region. Specifically, first, the paragraph designation reception processing unit 111 selects one area adjacent to the area selected in step 901 as an adjacent area (907). For example, in the process of step 901, assuming that areas are selected in order from the area above the document data, the area adjacent to the area selected in the process of step 901 is selected as the adjacent area in the process of step 907. Is done.

Next, the paragraph designation reception processing unit 111 extracts all the information sets existing in the adjacent area selected in the process of step 907, and calculates the total value of the areas of the extracted information sets as the area of the adjacent area ( 908).

Next, the paragraph designation reception processing unit 111 calculates the total value of the area area calculated in step 902 and the adjacent area area calculated in step 908 (909).

Next, the paragraph designation reception processing unit 111 calculates the area of the paragraph area designated by the user based on the coordinates registered in the upper left coordinates 401 and the coordinates registered in the lower right coordinates 402 of the paragraph table 400, It is determined whether the total value calculated in step 909 is within a predetermined range from the calculated area of the paragraph (910).

When it is determined in step 910 that the area of the region calculated in step 902 is within a predetermined range from the calculated area of the paragraph (910: Yes), the paragraph designation reception processing unit 111 performs step Proceeding to step 905, the region selected in step 901 and the adjacent region selected in step 907 are repeatedly extracted as candidate portions.

On the other hand, if it is determined in step 910 that the area calculated in step 902 is not within the predetermined range from the calculated area of the paragraph (910: No), the paragraph designation reception processing unit 111 It is determined whether the total value calculated in step 909 is smaller than the area of the paragraph area (911).

If it is determined in step 911 that the total value calculated in step 909 is smaller than the area of the paragraph area (911: Yes), the paragraph designation reception processing unit 111 selects the adjacent area selected in step 907. An area adjacent to the area is selected as an adjacent area (912), the process returns to step 908, the area selected in step 901, the area selected in step 907, and the area selected in step 912 are selected. The total value of the area and the area of the paragraph are compared.

On the other hand, if it is determined in step 911 that the total value calculated in step 909 is equal to or larger than the area of the paragraph area (912: No), the paragraph designation reception processing unit 111 performs the process in step 906. Then, it is determined whether or not the processing in steps 901 to 912 has been executed for all divided areas of the input image data.

FIG. 10 is a flowchart of the element matching determination process according to this embodiment.

First, the paragraph designation reception processing unit 111 selects one repetition candidate part to be processed from the repetition candidate parts extracted in the repetition candidate part extraction process (1001).

Next, the paragraph designation reception processing unit 111 selects one information set to be processed from the information set included in the repetition candidate part selected in the process of Step 1001 (1002).

Next, the paragraph designation reception processing unit 111 selects the element table 410 for comparison with the information set selected in the processing of Step 1002 (1003).

Next, the paragraph designation reception processing unit 111 uses the attribute set registered in the notation format and attribute 414 registered in the notation format 413 of the element table 410 selected in the process in step 1003 as the information set selected in the process in step 1002. It is determined whether or not the condition is satisfied (1004).

If it is determined in step 1004 that the information set satisfies the notation format and attributes of the element table 410 (1004: Yes), the paragraph designation reception processing unit 111 determines the center coordinates of the information set selected in step 1002 Is calculated (1005). The center coordinates of the information set are coordinates in the area where the information set exists, and are calculated using, for example, the center coordinates of the area where the information set exists as the origin.

Next, the paragraph designation reception processing unit 111 calculates the distance between the center coordinates of the information set calculated in the process of step 1005 and the coordinates registered in the center coordinates 412 of the element table 410 calculated in the process of step 1003. Calculate (1006).

Next, the paragraph designation reception processing unit 111 determines whether or not the distance calculated in the process of Step 1006 is less than or equal to a predetermined value (1007).

If it is determined in step 1007 that the distance calculated in step 1006 is equal to or less than a predetermined value (1007: Yes), the paragraph designation reception processing unit 111 selects the information set selected in step 1002 and the step The element table selected in the processing of 1003 is stored in association with it (1008).

Next, the paragraph designation reception processing unit 111 determines whether or not all information sets of the repetition candidate part selected in the process of Step 1001 have been selected in the process of Step 1002 (1009).

If it is determined in step 1009 that all information sets of the repetition candidate portion selected in step 1001 have been selected in step 1002 (1009: Yes), the paragraph designation reception processing unit 111 Is determined to correspond to the element table on a one-to-one basis (1010). For example, the first information set, the second information set, and the third information set exist in the repetition candidate portion selected in the process of step 1001, and the first element table, the second element table, and the third element are stored in the block dictionary 121. If a table is included, the first information set and the first element table are stored in association with each other in step 1008, the second information set and the second element table are stored in association with each other, and the third information set is stored. And the third element table are stored in association with each other, it is determined Yes in the process of step 1010. When one element table is associated with a plurality of information sets in an overlapping manner, the information set does not correspond to the element table on a one-to-one basis, and it is determined No in the processing of step 1010.

If it is determined in step 1010 that the information set corresponds to the element table on a one-to-one basis (1010: Yes), the paragraph designation reception processing unit 111 repeats the repetition candidate portion selected in step 1001. It extracts as a part and registers information on the candidate repeat part in the repeat table 124 (1011).

Specifically, the paragraph designation reception processing unit 111 adds 1 to the value registered in the number of repetitions 501 of the list table 500 included in the repetition table 124, and the pointer to the paragraph table 510 is added to the paragraph table pointer 502. Register. The paragraph designation reception processing unit 111 registers the upper left coordinates of the repetition candidate part selected in the process of step 1002 in the upper left coordinates 401 of the paragraph table 510, and registers the lower right coordinates of the repetition candidate parts in the lower right coordinates 402. In addition, the paragraph designation reception processing unit 111 registers the number of areas included in the repetition candidate part extracted as the repetition part in the partial area number 403 of the paragraph table 510. In addition, the paragraph designation reception processing unit 111 registers the number of information sets included in the repetition candidate part extracted as the repetition part in the element number 404 of the paragraph table 510 and the element table in the pointer 405 to the element of the paragraph table 510. A pointer to 520 is registered.

The paragraph designation reception processing unit 111 registers the coordinates of each vertex of the information set area included in the repetition candidate portion extracted as the repetition portion in the polygon coordinates 411 of the element table 520 included in the repetition table 124. A pointer to is registered. In addition, the paragraph designation reception processing unit 111 registers the center coordinates of the area of the information set included in the repeat candidate part extracted as the repeat part in the center coordinate 412 of the element table 520. In addition, the paragraph designation reception processing unit 111 registers the notation format of the information set included in the repetition candidate portion extracted as the repetition portion in the notation format 413 of the element table 520, and the information set is registered in the attribute 414 of the element table 520. Register the attributes.

After the execution of the process of step 1011, the paragraph designation reception processing unit 111 determines whether all the repetition candidate parts extracted by the repetition candidate part extraction process have been selected by the process of step 1001 (1012).

If it is determined in step 1012 that all the repetition candidate parts extracted in the repetition candidate part extraction process have been selected in the process of step 1001 (1012: Yes), the paragraph designation reception processing unit 111 performs element matching determination. The process ends. On the other hand, if it is determined in step 1012 that all the repetition candidate parts extracted in the repetition candidate part extraction process are not selected in the process of step 1001 (1012: No), the paragraph designation reception processing unit 111 Returning to the processing of step 1001.

When it is determined in step 1004 that the information set does not satisfy the notation format and attributes of the element table 410 (1004: No), and in step 1007, the distance calculated in step 1006 is a predetermined value. If it is determined that it is the following (1007: No), the information set selected in the process of step 1002 does not match the physical structure or logical structure of the element registered in the element table 410 selected in the process of step 1003. The paragraph designation reception processing unit 111 determines whether all the element tables 410 have been selected in the processing of Step 1003 (1013). If it is determined in step 1013 that all the element tables 410 have not been selected in step 1003 (1013: No), the paragraph designation reception processing unit 111 returns to step 1003 and is still selected. The element table 410 that is not selected is selected.

On the other hand, if it is determined in step 1013 that all the element tables 410 have been selected in step 1003 (1013: Yes), the information set selected in step 1002 is the physical structure or logic of any element. Since the repetition candidate part selected in the process of step 1001 does not match the structure and includes an information set that does not match the physical structure or logical structure of any element, the paragraph designation reception processing unit 111 selected in the process of step 1001 The repetition candidate part is determined not to be a repetition part (1014), and the process proceeds to step 1012.

If it is determined in the process of step 1010 that the information set does not correspond to the element table on a one-to-one basis (1010: No), the paragraph designation reception processing unit 111 proceeds to the process of step 1014. It is determined that the repetition candidate part selected in the process is not a repetition part.

FIG. 11 is an explanatory diagram of the process of dividing the paragraph area designated by the user of this embodiment using the XY cut method.

In the document data shown in FIG. 2, the paragraph area designated by the user is divided using the XY cut method. In FIG. 11, since the information set continuous in the horizontal direction is extracted, and the vertical distance between the two extracted information sets continuous in the horizontal direction is equal to or smaller than the threshold value, the paragraph region is not divided. For this reason, the number of partial areas of the area of the paragraph is 1.

FIG. 12 is an explanatory diagram of a process of dividing the entire area of the document data according to the present embodiment using the XY cut method.

2) The entire area of the document data shown in FIG. 2 is divided using the XY cut method. In FIG. 12, the entire area of the document data is divided into seven areas, a first area to a seventh area.

The first area includes the character string “Document 1 Sport Book”. The second area includes the character string “There are three types of this product”. The third area includes three figures and a character string related to each figure. The fourth area includes a character string “A product list is shown below”. The fifth area includes a character string related to the handbag. The sixth area includes a character string related to boots. The seventh region includes a character string related to gloves.

Next, the details of document processing will be described.

First, the extraction area candidate extraction process shown in FIG. 6 will be described. In the extraction area candidate extraction process, an area similar to the physical structure of the paragraph registered in the paragraph table 400 included in the block dictionary 121 is generated as an extraction area candidate from the area of the document data input by the document processing unit 112. . The extraction area candidate extraction process can be realized by the same flowchart as the repetition candidate part extraction process shown in FIG. The extraction area candidate extraction process is different from the repeated candidate part extraction process in that the document processing unit 112 extracts the area selected in the process of step 901 as an extraction area candidate in the process of step 905.

Next, details of the degree-of-match calculation processing will be described with reference to FIG. FIG. 13 is a flowchart of the coincidence degree calculation process according to this embodiment.

First, the document processing unit 112 selects one extraction area candidate to be processed from the extraction area candidates extracted in the extraction area candidate extraction process (1301).

Next, the document processing unit 112 selects one paragraph table 400 to be processed from the paragraph tables 400 registered in the block dictionary 121 (1302). For example, in the paragraph designation receiving process, when a plurality of paragraph areas are designated by the user, a plurality of paragraph tables 400 are registered in the block dictionary 121.

Next, the document processing unit 112 selects one information set to be processed from the information set included in the extraction area candidate selected in the processing of step 1301 (1303).

Next, the document processing unit 112 selects, from the element table 410 associated with the paragraph table 400 selected in step 1302, an element table 410 for comparison with the information set selected in step 1303 ( 1304).

Next, the document processing unit 112 executes a distance coincidence calculation process between center coordinates (1305). The center coordinate distance coincidence calculation processing is performed between center coordinates based on the distance between the center coordinates of the information set selected in step 1303 and the coordinates registered in the center coordinates 412 of the element table 410 selected in step 1304. This is a process for calculating the distance matching degree. Note that the smaller the distance between the center coordinates of the information set and the center coordinates of the elements, the higher the distance coincidence between the center coordinates.

Next, the document processing unit 112 executes a word matching degree calculation process (1306). In the word matching degree calculation process, the degree of matching between the characters included in the information set selected in step 1303 and the characters registered in the word dictionary 122 associated with the element table 410 selected in step 1304 is calculated. This is a calculation process. Of the characters included in the information set, the greater the number of characters that are the same as the characters registered in the word dictionary 122, the higher the word matching degree.

Next, the document processing unit 112 executes a notation format and attribute matching degree calculation process (1307). In the notation format and attribute matching degree calculation process, whether or not the information set selected in step 1303 satisfies the notation format registered in the notation format 413 of the element table 410 selected in step 1304, and step 1303. The notation format and attribute matching degree are calculated based on whether or not the information set selected in the process of step 1304 satisfies the attribute registered in the attribute 414 of the element table 410 selected in step 1304. When the information set satisfies the notation format and attribute, the notation format and attribute matching degree are the highest, and when the information set satisfies either the notation format and attribute, the notation format and attribute match degree are the second highest, and the information set Is the lowest in notation format and attribute matching degree when neither of the notation format nor attribute is satisfied.

Next, the document processing unit 112 is based on the distance between the center coordinates calculated in the process of step 1305, the word match calculated in the process of step 1306, and the notation format and attribute match calculated in the process of step 1307. The total matching degree is calculated (1308). Specifically, the document processing unit 112 calculates a total matching degree by weighting the center coordinate distance matching degree, the word matching degree, the notation format, and the attribute matching degree. For example, the weight coefficient for the center coordinate distance coincidence is set to the highest, the weighting coefficient for the notation format and the attribute coincidence is set to the next highest, and the weight coefficient for the word match is set to the lowest. As a result, when a character string that has not yet been extracted is extracted, since the character string is not registered in the word dictionary 122, the word matching degree is low, but the weight coefficient of the word matching degree is the lowest. If the distance coincidence between center coordinates or the notation format and the attribute coincidence are high, the character string can be extracted as an element.

Next, the document processing unit 112 determines whether or not all the element tables 410 associated with the paragraph table 400 selected in step 1302 have been selected in step 1304 (1309).

If it is determined in step 1309 that all element tables 410 associated with the paragraph table 400 selected in step 1302 have not been selected in step 1304 (1309: No), the document processing unit 112 returns to the process of step 1304 and selects a new element table 410.

On the other hand, if it is determined in step 1309 that all element tables 410 associated with the paragraph table 400 selected in step 1302 have been selected in step 1304 (1309: Yes), the document processing unit 112 determines whether all information sets existing in the extraction area candidate selected in the process of step 1301 have been selected in the process of step 1304 (1310).

When it is determined in step 1310 that all information sets existing in the extraction region candidate selected in step 1301 have not been selected in step 1304 (1310: No), the document processing unit 112 Returning to the processing in step 1303, another information set existing in the extraction region candidate selected in the processing in step 1301 is selected.

On the other hand, if it is determined in step 1310 that all information sets existing in the extraction region candidate selected in step 1301 are selected in step 1304 (1310: Yes), selection is performed in step 1301. The total matching degree with all the elements of all the information sets existing in the extracted region candidate has been calculated. In this case, the document processing unit 112 specifies an element that maximizes the total matching degree of each information set, and determines whether or not the specified elements of each information set overlap (1311).

When it is determined in step 1311 that the elements of each information set do not overlap, the document processing unit 112 determines whether or not the maximum total matching degree of all information sets is equal to or greater than a predetermined value (1312). ).

If it is determined in step 1312 that the maximum total matching degree of all information sets is greater than or equal to a predetermined value (1312: Yes), the document processing unit 112 converts each information set to the total of each information set. An element having the highest matching degree is extracted (1313), and the extracted information set is registered in the word dictionary 122 and the output DB 123 in association with the element having the highest total matching degree (1314). Then, the document processing unit 1313 determines whether all extraction region candidates extracted in the extraction region candidate extraction processing have been selected in the processing of Step 1301 (1315).

When it is determined in step 1315 that all extraction region candidates extracted in the extraction region candidate extraction processing are not selected in step 1301 (1315: No), the document processing unit 112 performs step 1301. Returning to the processing, one extraction region candidate to be processed is newly selected from the extraction region candidates extracted in the extraction region candidate extraction processing.

On the other hand, when it is determined in step 1315 that all extraction region candidates extracted in the extraction region candidate extraction processing are selected in step 1301 (1315: Yes), the document processing unit 112 calculates the degree of coincidence. The process ends.

When it is determined in the processing of step 1311 that the elements having the maximum total matching degree of each information set overlap (1311: Yes), or in the processing of step 1312, the maximum total matching degree is smaller than a predetermined value. If it is determined that a set exists (1312: No), the information set selected in the process of step 1303 does not match any of the element tables 410 in the paragraph table 400 selected in the process of step 1302. Therefore, the document processing unit 112 determines whether all the paragraph tables 400 have been selected in the process of step 1302 (1316).

If it is determined in step 1316 that all the paragraph tables 400 have not been selected in step 1302 (1316: No), the document processing unit 112 selects a paragraph table 400 that has not yet been selected ( 1317), the process returns to step 1304, and the element table 410 associated with the paragraph table selected in step 1317 is selected.

On the other hand, if it is determined in step 1316 that all the paragraph tables 400 have been selected in step 1302 (1316: Yes), they are associated with the element table 410 that matches the information set selected in step 1303. The extracted paragraph table 400 does not exist, and the extraction area candidate including the information set is not similar to any paragraph table 400. Therefore, the document processing unit 112 proceeds to the process of step 1314, and determines whether all the extraction area candidates extracted in the extraction area candidate extraction process have been selected in the process of step 1301.

FIG. 14 is an explanatory diagram of processing for dividing the entire area of the document data using the XY cut method in the document processing of this embodiment.

In the processing of step 614 shown in FIG. 6, the entire area of the document data shown in FIG. 3 is divided using the XY cut method. In FIG. 14, the entire area of the document data is divided into seven areas, a first area to a sixth area.

The first area includes the character string “Document 2 Sport Book”. The second area includes a character string “There are three types of this product. A list of products is shown below”. The third region includes a character string related to the handbag. The fourth area includes a character string related to boots. The fifth area includes a character string related to the glove. The sixth area includes a character string related to the folding umbrella.

FIG. 15 is an explanatory diagram of a repeated partial display screen 1500 according to the present embodiment.

The repeated partial display screen 1500 displays document data on which a paragraph designation receiving process has been executed. In the document data, a paragraph area (1501) designated by the user and an element area (1502) designated by the user. Is displayed in a rectangle. In the document data displayed on the repeated part display screen 1500, the repeated part area (1511) extracted in the paragraph designation receiving process is displayed surrounded by a rectangle and displayed in the repeated part extracted in the paragraph designation receiving process. A region (1512) of the information set extracted corresponding to the included element is also displayed surrounded by a rectangle.

In addition, in the repeated part display screen 1500, the designation of the area of the repeated part not extracted (1521) and the designation of the area where the information set to be extracted as an element included in the area of the repeated part exists (1522) are added. Can be accepted.

In FIG. 15, the region related to the handbag is designated as a paragraph, and the region related to boots is extracted as a repeated portion. Even though the region relating to the glove should be extracted as a repeated part, it is not extracted as a repeated part. For this reason, the user designates an area related to gloves as a repetitive part, and also designates an information set to be extracted as an element of the area.

When the repetition part display screen 1500 receives the designation of the area of the repetition part and the area where the information set to be extracted as an element included in the area exists, the paragraph designation reception processing unit 111 selects these areas. Is repeatedly registered in the table 124. Specifically, the paragraph designation reception processing unit 111 adds 1 to the value registered in the number of repetitions 501 of the list table 500 of the repetition table 124, and designates the area of the specified repetition portion in the pointer 502 to the paragraph table. Is registered as a pointer to the paragraph table 510 to be registered. In addition, the paragraph designation reception processing unit 111 registers the upper left coordinates of the repeated portion area designated by the upper left coordinates 401 of the paragraph table 510 and registers the lower right coordinates of the repeated portion area designated by the lower right coordinates 402. To do. In addition, the paragraph designation reception processing unit 111 divides the designated repetitive portion area using the XY cut method, and registers the number of divided areas as the partial area number 403. Also, the paragraph designation reception processing unit 111 registers the number of information sets to be extracted as the designated element in the element number 404 of the paragraph table 510. In addition, the paragraph designation reception processing unit 111 registers a pointer to the element table 520 in which the information set extracted as the designated element is registered in the pointer 405 to the element of the paragraph table 510. In the paragraph table 510, information indicating that the repetitive part is extracted by accepting designation by the user may be registered.

Next, the paragraph designation reception processing unit 111 registers a pointer to the coordinate table 530 in which the coordinates of each vertex of the designated information set area are registered in the polygon coordinates 411 of the element table 520. In addition, the paragraph designation reception processing unit 111 registers the center coordinates of the designated information set area in the center coordinates 412 of the element table 520. The paragraph designation reception processing unit 111 registers the notation format of the designated information set in the notation format 413 of the element table 520 and registers the attribute of the information set in the attribute 414 of the element table 520. In the element table 520, information indicating that the area of the information set is extracted by accepting designation by the user may be registered.

By the above, the user can confirm whether or not the region to be extracted as the repeated portion has been extracted. If the region to be extracted as the repeated portion has not been extracted, the region is additionally repeated. Can be specified as

Note that the repeated partial display screen 1500 may include a word dictionary display screen that displays an information set registered in the word dictionary 122. On the word dictionary display screen, the information set registered by the user's designation and the information set registered by being extracted as a repeated part may be displayed in a distinguishable manner.

FIG. 16 is an explanatory diagram of a processing result display screen 1600 according to the present embodiment.

The processing result display screen 1600 displays document data on which document processing has been executed, and in the document data, an extraction area extracted by document processing and an area of an information set extracted as an element are displayed.

In the processing result display screen 1600 shown in FIG. 16, the first extraction area 1610, the second extraction area 1620, the third extraction area 1630, and the fourth extraction area 1640 are displayed surrounded by a rectangle. In each extraction area, information set areas 1611 to 1613 extracted as each element designated by the user are displayed surrounded by a rectangle.

For example, in the paragraph designation receiving process, as shown in FIG. 2, “055-3223”, “015-0001”, and “015-0149” are registered in the element 1 of the word dictionary 122, and “handbag”, When “boots” and “gloves” are registered, “331-0120” located at the position of element 1 in the fourth extraction region 1640 is not registered in element 1 of the word dictionary 122 and The “folding umbrella” located at the position is not registered in the element 2 of the word dictionary 122. For this reason, the total matching degree of the fourth extraction region 1640 is lower than the total matching degree of the first extraction region 1610 to the third extraction region 1630.

The document processing unit 112 includes a rectangle surrounding the fourth extraction region 1640 having a lower degree of overall matching than the first extraction region 1610 to the third extraction region 1630 and information collection regions 1611 to 1613 included in the fourth extraction region 1640. May be displayed by a dotted line to notify the user that the total matching degree of the fourth extraction region 1640 is lower than that of the first extraction region 1610 to the third extraction region 1630. Thus, by changing the display mode of the rectangle surrounding the extraction area and the information set according to the total matching degree, the user can immediately confirm the total matching degree of each extraction area. Note that the change in the display mode of the total matching degree is not limited to the dotted line, and for example, the color may be changed according to the total matching degree, or the thickness of the line may be changed.

In the processing result display screen 1600, when an area corresponding to the paragraph to be extracted is not extracted, designation of the area corresponding to the paragraph and an information set to be extracted included in the area corresponding to the paragraph The designation of the area may be received from the user. When the document processing unit 112 registers the information set extracted from the region of the information set in the word dictionary 122 and the output DB 123, the document processing unit 112 displays information indicating that the information set is extracted by accepting designation by the user. It may be registered in the output DB 123.

It should be noted that the processing result display screen 1600 may include a word dictionary display screen that displays an information set registered in the word dictionary 122. On the word dictionary display screen, the information set registered by the user's designation and the information set registered by being extracted as a repeated part may be displayed in a distinguishable manner.

In this embodiment, the XY cut method is used as a method of dividing the document data area into at least one area, but other methods may be used. For example, the text existing in the document data area may be regarded as one sentence, the morpheme analysis may be performed, and the document data area may be divided into at least one area based on the result of the morpheme analysis.

In this embodiment, the storage area of the document processing system includes the physical structure of elements (the central coordinates 412 of the element table 410) that are included in the paragraph to be extracted from the document data and have information to be extracted. The information set (word dictionary 122) existing in the area of the element and the learning dictionary (block dictionary 121 and word dictionary 122) registered are stored, and the processor stores the input document data in at least one area. The information set existing in the area is acquired (FIG. 13, 1002), and the physical structure matching degree indicating the matching degree between the acquired area of the information set and the physical structure of the element registered in the learning dictionary is calculated (FIG. 13, 1305). ), The information set matching degree indicating the matching degree between the acquired information set and the information set registered in the learning dictionary is calculated (FIG. 13, 1306), and the calculated physical structure matching degree is calculated. And whether or not the region is a paragraph to be extracted based on the information set matching degree (FIGS. 1311 and 1312), and if the region is a paragraph to be extracted, the acquired information set is matched with the physical structure Is registered in the learning dictionary as an information set existing in the element area corresponding to the information set for which the degree and the information set match degree are calculated (FIG. 131314), and the acquired information set is the physical structure match degree and the information set match degree. The output information associated with the elements of the information set for which is calculated is stored in the storage area (FIG. 13, 1314).

Thus, each time an information set is extracted from the document data based on the physical structure matching degree and the information set matching degree of the element, and the information set is extracted from the document data, the element information set is stored in the learning dictionary. Therefore, the information set can be extracted for each element without defining in advance an information set such as a word that is a key for extracting the information set for each document.

In addition, the document processing system calculates a total matching score that is a sum of the calculated physical structure matching score and the calculated information set matching score (1308 in FIG. 13), and based on the calculated total matching score, It is characterized by determining whether or not it is a paragraph to be extracted (FIG. 13, 1312). As a result, even if one of the matching degrees is low, it can be determined that the region is a paragraph to be extracted if the other matching degree is high.

In addition, the document processing system displays a processing result display screen (FIG. 16) that displays the area that is the paragraph to be extracted and the area of the information set acquired from the area in the input document data. Screen data for output is output, and the display mode of the area that is assumed to be a paragraph to be extracted and the area of the information set is changed according to the total matching degree. As a result, the user can confirm the processing result display screen and immediately confirm the area extracted from the document data and the area of the information set, as well as the overall matching degree.

Further, the processor is characterized in that it can accept designation of a paragraph area and an element area to be extracted via a processing result display screen. Thus, when the paragraph to be extracted and the information set to be extracted are not extracted, the user can reliably extract the information set by designating these areas.

The physical structure of the element registered in the learning dictionary includes the position of the element in the paragraph. In the learning dictionary, at least one of the notation format and attribute of the information set existing in the element area is registered, and the processor Calculates the position of the area of the acquired information set in the area, calculates the distance between the position of the information set and the position of the element in the paragraph registered in the learning dictionary, The physical structure matching degree is calculated based on the information (FIG. 13, 1305), and at least one of the acquired information set notation format and attribute is the notation format and attribute of the information set existing in the element area registered in the learning dictionary. The notation format and attribute matching degree are calculated based on whether or not they match at least one (FIG. 13, 1307), the calculated physical structure matching degree and the calculated information set matching degree, and the calculated notation format and attribute one Calculating the total matching score is the sum of weighted and time (Figure 13 1308) it is characterized.

Thus, based on the distance between the position of the information set existing area and the position of the element in the paragraph registered in the learning dictionary, the physical structure match degree, the information set match degree, the notation format and the attribute match degree The degree of coincidence can be calculated, and the extraction accuracy can be improved.

Further, the physical structure of the paragraph to be extracted from the document data is registered in the learning dictionary (paragraph table 400 in FIG. 4), and the processor registers the paragraph registered in the learning dictionary from at least one area of the input document data. An area similar to the physical structure is selected (FIG. 6, 615), and an information set existing in the selected area is acquired. As a result, an information set is extracted only from an area having a physical structure similar to that of the paragraph to be extracted, so that the processing load of the document processing system can be reduced.

Further, the processor receives the learning dictionary generation document data for generating the learning dictionary (FIG. 7, 702), and the user specifies the paragraph area to be extracted from the input learning dictionary creation document data area. (Fig. 7, 705), the specification of the element area within the paragraph area to be extracted is accepted from the user (Fig. 7, 707), and the paragraph area to be extracted that has received the designation is to be extracted. Is registered in the learning dictionary as the physical structure (Fig. 7, 706), and the region of the element that has received the designation is registered in the learning dictionary as the physical structure of the element (Fig. 7, 708 and 709) and exists in the region of the element that has received the designation. The information set to be registered is registered in the learning dictionary as an element information set (710 in FIG. 7).

Thus, the user can generate a learning dictionary only by specifying one paragraph area to be extracted and an element area in the paragraph area.

In addition, the processor selects, from the at least one area, the input learning information creation document data, an area similar to the physical structure of the paragraph registered in the learning dictionary (FIG. 10, 1001), and exists in the selected area. The information set is acquired (FIG. 10, 1002), the existence area of the acquired information set is compared with the physical structure of the element registered in the learning dictionary, and the selected area is the area of the paragraph to be extracted that has received the designation. It is determined whether or not it is a repetitive part (FIG. 10, 1005 to 1007, 1011), and if the selected area is a repetitive part, the information set acquired from the extracted area is associated with the element used for comparison, It is characterized in that output information associated with elements used for comparison is stored in a storage area, which is registered in a learning dictionary and acquired from the extracted area (FIG. 8, 809). That.

This allows the user to extract the repeated part of the paragraph area and specify the information set of the repeated part simply by specifying one paragraph area to be extracted and the element area in the paragraph area. Since the number of information sets that can be registered in the learning dictionary and registered in the learning dictionary increases, the extraction accuracy can be improved.

In the input learning dictionary generation document data, the processor is acquired from the area of the paragraph that received the specification, the area of the element that received the specification, the area extracted as the repeated part, and the area extracted as the repeated part Outputs screen data for displaying the repeated part display screen that displays the area of the information set (FIG. 8, 807), specifies the area to be extracted as the repeated part, and as the repeated part via the repeated part display screen It is possible to accept the specification of the area of the element from which the information set is extracted in the area to be extracted.

Thus, when the repetitive part to be extracted and the area of the information set to be extracted from the repetitive part are not extracted, the user can reliably extract the information set by designating these areas.

In addition, this invention is not limited to the above-mentioned Example, Various modifications are included. For example, the above-described embodiments have been described in detail for easy understanding of the present invention, and are not necessarily limited to those having all the configurations described. Further, a part of the configuration of a certain embodiment can be replaced with the configuration of another embodiment, and the configuration of another embodiment can be added to the configuration of a certain embodiment. Further, it is possible to add, delete, and replace other configurations for a part of the configuration of each embodiment.

Also, it is possible to add, delete, and replace other configurations for a part of the configuration of each embodiment. Each of the above-described configurations, functions, processing units, processing means, and the like may be realized by hardware by designing a part or all of them with, for example, an integrated circuit.

Further, each of the above-described configurations and functions may be realized by software by interpreting and executing a program that realizes each function by the processor.

Information such as programs, tables, and files that realize each function is recorded in memory, hard disk, SSD (Solid State Drive), or IC (Integrated Circuit) card, SD card, DVD (Digital Versatile Disc), etc. Can be placed on any recording medium.

Also, the control lines and information lines indicate what is considered necessary for the explanation, and not all the control lines and information lines on the product are necessarily shown. Actually, it may be considered that almost all the components are connected to each other.

Claims (10)

A document processing system that extracts an information set from input document data,
A processor and a storage area;
In the storage area,
A learning dictionary in which a physical structure of an element which is included in a paragraph to be extracted from the document data and where information to be extracted exists and an information set existing in the area of the element are registered is stored,
The processor is
Obtaining an information set present in at least one area of the input document data;
Calculating the physical structure coincidence indicating the coincidence between the acquired region of the information set and the physical structure of the element registered in the learning dictionary;
Calculating an information set coincidence indicating a degree of coincidence between the acquired information set and the information set registered in the learning dictionary;
Based on the calculated physical structure coincidence and information set coincidence, determine whether the region is the paragraph to be extracted,
When the region is the paragraph to be extracted, the learning dictionary is used as the information set existing in the region of the element corresponding to the information set for which the physical structure matching degree and the information set matching degree are calculated. Registered with
The document processing system, wherein output information in which the acquired information set is associated with an element of the information set for which the physical structure matching degree and the information set matching degree are calculated is stored in the storage area. The document processing system according to claim 1,
The processor is
Calculating a total matching degree that is a weighted sum of the calculated physical structure matching degree and the calculated information set matching degree;
A document processing system that determines whether or not the region is the paragraph to be extracted based on the calculated total matching degree. The document processing system according to claim 2,
The processor is
In the input document data, output screen data for displaying a region to be extracted and a processing result display screen for displaying a region of an information set acquired from the region,
A document processing system, wherein a display mode of an area determined to be a paragraph to be extracted and an area of an information set acquired from the area is changed according to the total matching degree. The document processing system according to claim 3,
The document processing system, wherein the processor is capable of accepting designation of the paragraph area to be extracted and the element area via the processing result display screen. The document processing system according to claim 2,
The physical structure of the element registered in the learning dictionary includes the position of the element in the paragraph,
In the learning dictionary, at least one of a notation format and an attribute of an information set existing in the area of the element is registered,
The processor is
Calculating the position of the acquired region of the acquired information set in the region;
Calculating the distance between the position of the region of the information set and the position of the element in the paragraph registered in the learning dictionary;
Calculate the physical structure coincidence based on the calculated distance,
Based on whether at least one of the notation format and attribute of the acquired information set matches at least one of the notation format and attribute of the information set existing in the area of the element registered in the learning dictionary, and Calculate attribute match,
A document processing system that calculates a total matching degree that is a weighted sum of the calculated physical structure matching degree, the calculated information set matching degree, and the calculated notation format and attribute matching degree. The document processing system according to claim 1,
In the learning dictionary, a physical structure of a paragraph to be extracted from the document data is registered,
The processor is
Selecting a region similar to the physical structure of the paragraph registered in the learning dictionary from at least one region of the input document data;
A document processing system for acquiring an information set existing in the selected area. The document processing system according to claim 6,
The processor is
Learning dictionary generation document data for generating the learning dictionary is input,
Accepting from the user the specification of the area of the paragraph to be extracted among the areas of the input learning dictionary creation document data,
Accepting designation of the area of the element from the user within the area of the paragraph to be extracted that has accepted the designation,
Registering the region of the paragraph to be extracted that has received the designation as the physical structure of the paragraph to be extracted in the learning dictionary,
Register the region of the element that has received the designation as the physical structure of the element in the learning dictionary,
A document processing system, wherein an information set existing in a region of an element that has received the designation is registered in the learning dictionary as the information set of the element. The document processing system according to claim 7,
The processor is
Select the region similar to the physical structure of the paragraph registered in the learning dictionary from at least one region of the input learning information creation document data,
Obtaining an information set present in the selected area;
Comparing the acquired region of the information set with the physical structure of the element registered in the learning dictionary, whether or not the selected region is a repeated part of the region of the paragraph to be extracted that has received the designation Determine
When the selected area is the repeated portion, the information set acquired from the extracted area is registered in the learning dictionary in association with the element used for the comparison,
The document processing system, wherein the output information in which the information set acquired from the extracted area is associated with the element used for the comparison is stored in the storage area. The document processing system according to claim 8,
The processor is
Acquired from the input document data for generating the learning dictionary from the paragraph region that has received the specification, the region of the element that has received the specification, the region extracted as the repeated portion, and the region extracted as the repeated portion Output screen data for displaying the repeated partial display screen that displays the area of the information set
It is possible to accept specification of an area to be extracted as the repeated part and specification of an element area for extracting an information set in the area to be extracted as the repeated part via the repeated part display screen. A document processing system. A document processing method for extracting an information set from document data input using a computer having a processor and a storage area,
In the storage area, the physical structure of an element that is included in the paragraph to be extracted from the document data and in which the information to be extracted exists, and the information set existing in the area of the element are registered. Learning dictionary is remembered,
The document processing method includes:
The processor obtains an information set existing in at least one area of the input document data,
The processor calculates a physical structure coincidence indicating a degree of coincidence between the acquired region of the information set and the physical structure of the element registered in the learning dictionary;
The processor calculates an information set coincidence indicating a degree of coincidence between the acquired information set and the information set registered in the learning dictionary;
The processor determines whether the divided area is the paragraph to be extracted based on the calculated physical structure matching degree and information set matching degree,
When the divided area is the paragraph to be extracted, the processor has the acquired information set in an area of an element corresponding to the information set for which the physical structure matching degree and the information set matching degree are calculated. Register as an information set in the learning dictionary,
The document processing method, wherein the processor stores, in the storage area, output information in which the acquired information set is associated with elements of the information set for which the physical structure matching degree and the information set matching degree are calculated. .

Images