JP4677750B2 - Document attribute acquisition method and apparatus, and recording medium recording program - Google Patents

Description

  The present invention relates to a document attribute acquisition method and apparatus for acquiring an attribute of a document from a document image, and a recording medium on which a program for acquiring the document attribute is recorded.

  A document storage system such as an electronic filing system reads a document image with a scanner and stores it in a database. In order to effectively use the stored document, a desired document is quickly and quickly stored in the database. It must be able to be called accurately. For this reason, initially, in this type of system, an operator inputs an appropriate search keyword for each document from the keyboard, and searches are performed based on this keyword when searching for a document. However, it is not only troublesome for an operator to input a keyword for each document when storing the document, but there is also a possibility that a keyword input error may occur. Therefore, a technique for automating keyword creation has been proposed.

  For example, Japanese Patent Application Laid-Open No. 1-106263 discloses a technique relating to a document storage / retrieval apparatus. In this device, an image reading device that reads a document and photoelectrically converts it into a document image, a character recognition device that extracts and recognizes characters in the document image and converts them into a character code string, and words and phrases for classification items are registered in advance. And a keyword group extracting device that extracts keyword data relating to a document by collating a character code string with a word or phrase and attaching it to a document image and storing it in a database. Thus, when the operator inputs a document using the image reading apparatus, the keyword group is automatically assigned a keyword group by the keyword group extracting apparatus described above and stored in the database. At the time of retrieval, the target document can be extracted simply by inputting a general keyword group in the field to which the document belongs.

Japanese Patent Application Laid-Open No. 11-184894 discloses a technique regarding a logical element extraction method for detecting a logical structure of a digital document image or an electronic document. In this technology, when recognizing a logical structure, bibliographic items are accurately extracted using only the layout information without using a character recognition result, and the logical type is automatically recognized by identifying the document type. It is. Specifically, a document is input as a digital image, the document image is divided into predetermined elements, layout characteristics of the document are detected, and a logical structure model created in advance for each of a plurality of document types is detected. A predetermined logical element is extracted in a first unit from the document image using one model, and then a predetermined logical element is extracted in a second unit. The layout feature of the one model and the extracted logic The similarity to the layout feature of the document corresponding to the element is calculated, and the extracted logical element is output when a value obtained by multiplying the similarity by a predetermined value is equal to or greater than a predetermined threshold.
JP-A-1-106263 Japanese Patent Laid-Open No. 11-184894

  Conventionally, as described above, a keyword dictionary is prepared in advance, and if a word registered in the keyword dictionary is used in a document, the word is registered as a keyword of the document, or the layout of the document is used. The attributes (bibliographic items) are extracted by analyzing and extracting logical elements and examining the similarity between the layout features corresponding to the extracted elements and that of the model document. However, the former only extracts the contents of one attribute “keyword”, and the latter only extracts the contents of predetermined attributes such as “title”, “large headline”, and “small headline”. Is. Such attribute setting is inflexible, and this is because the criteria for “what should be attributed in a document” are not clear. For this reason, conventionally, there has been a problem that in structuring a document for a scanned image, the meaning and association of a specific region on the image cannot be performed, and the document cannot be structured or reused.

Accordingly, an object of the present invention is to provide a program and a recording medium storing the for said to solve the problems, possible a document attribute preparative TokuSo location and document attribute acquisition of a high-quality document structuring is there.

An object of the present invention is to provide an attribute dictionary storage means for storing a plurality of attribute dictionaries generated for each attribute , having an attribute type section storing a plurality of attribute types and an attribute entity type section storing a plurality of attribute entity types, and an input Acquisition means for acquiring a plurality of text areas including character information obtained by character recognition from an image, character information of each text area acquired by the acquisition means, and attribute types of attribute type sections of the plurality of attribute dictionaries An attribute entity matching unit that matches the character information of each text area acquired by the acquiring unit and the attribute entity type of the attribute entity type part of the plurality of attribute dictionaries, A selection means for selecting from the plurality of attribute dictionaries an attribute dictionary including the attribute type matched by the attribute matching means and the attribute entity type matched by the attribute entity matching means; Refers to the attribute dictionary where the selection means is selected, said one of the attribute dictionary, the types of types and attributes entity attributes stored in one attribute dictionary, a text including character information that matches the type of the attribute and region extraction means for extracting a text region containing the character information that matches the type of the attribute entities, and attributes of the one attribute dictionary, the one attribute dictionary of the text area extracted by the extracting unit This is achieved by a document attribute acquisition apparatus comprising output means for outputting character information of a text area corresponding to a stored attribute entity type as an attribute and an attribute entity, respectively.

Here, before Symbol attribute dictionary may further comprise an attribute entity content portion that contains the group name attribute entities are matched to the attribute entity type unit. The attribute entity content part can be shared among the plurality of attribute dictionaries .

In addition, at least one of the attribute type of the attribute type part and the attribute entity type of the attribute entity type part may have a rank for collation.

A recording medium according to the present invention stores a plurality of attribute dictionaries generated for each attribute, the computer having an attribute type section storing a plurality of attribute types and an attribute entity type section storing a plurality of attribute entity types. Attribute dictionary storage means, acquisition means for acquiring a plurality of text areas including character information obtained by character recognition from an input image, the acquired character information of each text area, and attributes of the attribute type section of the plurality of attribute dictionaries Attribute collating means for collating, the attribute entity collating means for collating the acquired character information of each text area and the attribute entity type of the attribute entity type section of the plurality of attribute dictionaries, the attribute type Selection means for selecting from the plurality of attribute dictionaries an attribute dictionary that includes a type of attribute matched by collation of a part and a type of attribute entity matched by collation of the attribute entity type part, Referring to out attribute dictionary, the one attribute dictionary, the types of types and attributes entity attributes stored in one attribute dictionary, and a text area including the text information that matches the type of the attribute, the extracting means for extracting a text region containing the character information that matches the type of the attribute entities, and attributes of the one attribute dictionary, the type of the stored attribute entity to the one attribute dictionary of the extracted text region A computer-readable recording medium storing a program for causing the character information of the corresponding text area to function as an output means for outputting the information as an attribute and an attribute entity, respectively.

The program according to the present invention stores a plurality of attribute dictionaries generated for each attribute, the computer having an attribute type section storing a plurality of attribute types and an attribute entity type section storing a plurality of attribute entity types. Attribute dictionary storage means, acquisition means for acquiring a plurality of text areas including character information obtained by character recognition from an input image, character information of each acquired text area, and attribute type section of the plurality of attribute dictionaries Attribute matching means for matching the attribute type, attribute entity matching means for matching the acquired character information of each text region and the attribute entity type of the attribute entity type section of the plurality of attribute dictionaries, the attribute Selector that selects from the plurality of attribute dictionaries an attribute dictionary that includes the type of attribute matched by matching the type part and the type of attribute entity matched by matching the attribute entity type part , By referring to the selected attribute dictionary, among the attribute dictionary, the types of types and attributes entity attributes stored in one attribute dictionary, and a text area including the text information that matches the type of the attribute extraction means for extracting a text region containing the character information that matches the type of the attribute entities, and attributes of the one attribute dictionary, attributive entity stored in said one of the attribute dictionary of the extracted text region This is to cause the character information of the text area corresponding to the type to function as output means for outputting the attribute and attribute entity respectively.

  According to the present invention, it is possible to obtain a document attribute acquisition method and apparatus that enable high-quality document structuring and a recording medium on which a program is recorded. In the present invention, in a document structure recognition process for a scanned image, high-quality document structuring is enabled by extracting a friend (region) having a high-precision attribute-attribute entity relationship. This effect can be obtained regardless of whether the original is fixed or non-standard, and whether or not a table or ruled line is present. Thereby, at the time of reuse, for example, attribute entity extraction and sorting can be performed with high accuracy.

  FIGS. 1A and 1B are diagrams showing an embodiment of a document attribute acquisition method according to the present invention. In the present embodiment, as shown in the figure, a plurality of attribute dictionaries 1 to 4 having an attribute type section storing attribute types and an attribute entity type section storing attribute entity types are prepared. For example, the illustrated attribute dictionary 1 is a dictionary relating to employee numbers, and the attribute type section includes employee numbers, employee numbers, and the like as attribute types. , No. Are stored in the attribute entity type section as attribute entities such as arithmetic numerals, Chinese numerals, and Roman numerals. A plurality of attribute entity type sections can be provided, which will be described later. The other attribute dictionaries 2 to 4 store other attributes such as the document creator and the storage location.

  In the document 5 from which the attribute is to be acquired, for example, an employee number, a creator, and a storage location relating to the document are described. If the document 5 is electronic document information input by a word processor or the like, a plurality of text areas exist there. When the document 5 is an image, image information is read for each area by an optical character reader (OCR), and a plurality of text areas 6 are obtained by character recognition from the read input image.

  The attribute of the document 5 is acquired as follows. First, the contents of each text area 6 and the contents of the attribute type parts of the plurality of attribute dictionaries 1 to 4 are collated. Further, the contents of each text area and the contents of the attribute entity type part of the plurality of attribute dictionaries are collated. Then, an attribute dictionary that matches the matching result of the attribute type part and the matching result of the attribute entity type part is selected, and the attribute and attribute entity of the text area are acquired from the attribute type part and attribute entity type part of the selected attribute dictionary. . In this way, matching is performed for each of the attribute type part and the attribute entity type part for each text region 6, and a pair (relevance) is determined and selected.

In the example of FIG. 1A, the attribute dictionary 1 (employee number) is selected as an attribute dictionary in which the matching result of the attribute type part matches the matching result of the attribute entity type part. In this case, “employee number” and “012345” in the text area 6 are the employee number attribute pair 7 and are acquired as attributes and attribute entities of the text area.
In the example of FIG. 1B, the attribute dictionary 2 (creator) is selected as an attribute dictionary in which the matching result of the attribute type part matches the matching result of the attribute entity type part. In this case, “creator” and “atsushi Ito” in the text area 6 are the creator attribute pair 8 and are acquired as attributes and attribute entities of the text area.

  As described above, in this embodiment, in the document structure recognition process, an attribute dictionary is selected for each text area, and a pair or relationship in which an attribute and an attribute entity are related with reference to the selection result, that is, an attribute and an attribute. A companion (region) having an entity relationship is determined and extracted, and association as a document structure is performed.

  2A and 2B are views showing another embodiment of the document attribute acquisition method according to the present invention. The embodiment of FIG. 2A differs from the embodiment of FIG. 1 in that the attribute types 1 and 4 of the attribute dictionaries 1 to 4 are assigned ranks 1, 2, 3,.・ Attached. This order can be provided in at least one of the attribute type part and the attribute entity type part. Further, the embodiment of FIG. 2B differs from the embodiment of FIG. 1 in that the distance between the text areas 6 as indicated by the arrows in the figure is calculated, and the text area is referenced by referring to the calculated distance. Acquires attributes and attribute entities. Not only the distance between the text areas but also the arrangement position of each text area can be referred to, which will be described later. All of these embodiments are effective in quickly and accurately acquiring document attributes.

  3A and 3B are diagrams showing another embodiment of the document attribute acquisition method according to the present invention. FIG. 3A is a block diagram showing an example of a document attribute acquisition apparatus, and FIG. 3B is a diagram for explaining the document attribute acquisition method. FIG. As shown in FIG. 3A, the apparatus includes an image input unit or document input unit 31, a character region extraction unit 32, a character recognition unit (OCR processing unit) 33, an attribute dictionary storage unit 34, an attribute dictionary A collation unit 35 and an attribute information output unit 36 are provided. The image input unit 31 is an image scanner, for example, and inputs a document 14 as shown in FIG. In the case of the document input unit 31, electronic document information that is not attributed is input. The character area extraction unit 32 extracts the character area of the input scan image or document information. A character recognition unit (OCR processing unit) 33 performs character recognition in the character region and converts it into a plurality of text regions 15. In the case of the document input unit 31, the character area extraction unit 32 and the character recognition unit (OCR processing unit) 33 can be omitted.

  As shown in FIG. 3B, the attribute dictionary storage unit 34 stores an attribute dictionary group including a plurality of attribute dictionaries 11 to 13. Each of the attribute dictionaries 11 to 13 includes an attribute type section storing attribute types and a plurality of attribute entity type sections 21 and 22 storing attribute entity types. For example, in the attribute dictionary 13 (person information), name, name, person, name, name, and the like are stored in the attribute type section. The attribute entity type section 21 stores Ito, Suzuki, Takahashi, Yamamoto, Shibata and the like, and the attribute entity type section 22 stores arithmetic numbers, Chinese numerals, Roman numerals, and the like. The attribute collating unit 35 inputs a plurality of text regions 15 obtained from document information or a plurality of text regions 15 obtained by character recognition from an input image, and compares the contents of each text region and the contents of a plurality of attribute dictionaries. Match. The attribute information output unit 36 acquires and outputs the attributes and attribute entities of the text area based on the collation result.

  FIG. 4 shows an example of a flowchart for executing the embodiment of FIG. 3. FIG. 4A shows the case of inputting document information, and FIG. 4B shows the case of inputting image. In the case of inputting document information, as shown in FIG. 4A, in step 41, the document information is input by the information input means. In step 42, the contents of the attribute type sections of the plurality of attribute dictionaries are collated for each text area. In step 43, each text area is collated with the contents of the attribute entity type part of the plurality of attribute dictionaries. If there is a matching attribute dictionary based on the matching result of the attribute type part and the matching result of the attribute entity type part in step 44, the corresponding attribute is set as the attribute and attribute entity of the text area.

  In the case of image input, as shown in FIG. 4B, the image is input by the image input means in step 45. In step 46, the input image is subjected to layout analysis, and OCR processing is performed for each text area. In step 47, the contents of the attribute type sections of the plurality of attribute dictionaries are collated for each text area. In step 48, each text area is collated with the contents of the attribute entity type portion of a plurality of attribute dictionaries. If there is a matching attribute dictionary based on the matching result of the attribute type part and the matching result of the attribute entity type part in step 49, the corresponding attribute is set as the attribute and attribute entity of the text area.

  As described above, in this embodiment, the attribute dictionary has an attribute type part and a plurality of attribute entity type parts, selects an attribute dictionary for each text area, and refers to the selection result to determine and select related friends. The association as a document structure is performed.

  FIG. 5 is a diagram showing another embodiment of the document attribute acquisition method according to the present invention. This embodiment is different from the embodiment of FIG. 4 in that an attribute entity content section 16 storing a group name of attribute entities is further provided. In the present embodiment, as shown in the drawing, each of the attribute dictionaries 11 to 13 includes an attribute type part storing attribute types and a plurality of attribute entity type parts 21 and 22 storing group names as attribute entity types. Have. For example, in the attribute dictionary 13 (person information), the attribute entity type section 21 has a person name group as a group name, and the attribute entity type section 22 has a code group as a group name. For each text area 15, collate with the contents of the attribute entity content part of the plurality of attribute dictionaries, and if they match, collate the group name of the attribute entity content part dictionary with the group name of the attribute dictionary type part of the attribute dictionary To do. If there is a matching attribute dictionary based on the matching result of the attribute type part and the matching result of the attribute entity type part, the corresponding attribute is set as the attribute of the text area and the attribute entity. This attribute entity content portion can be shared among a plurality of attribute dictionaries 12 to 13 as indicated by dotted arrows in the figure.

  FIG. 6 shows an example of a flowchart for executing the embodiment of FIG. 5. FIG. 6A shows a case of inputting document information, and FIG. 6B shows a case of inputting an image. In the case of inputting the document information, as shown in FIG. 6A, the document information is input by the information input means in step 61. In step 62, the contents of the attribute type sections of the plurality of attribute dictionaries are collated for each text area. In step 63, each text area is collated with the contents of each dictionary in the attribute entity contents section of the plurality of attribute dictionaries. If they match, the dictionary group name in the attribute entity contents section and the attribute dictionary type section in the attribute dictionary are matched. Match group names. If there is a matching attribute dictionary based on the matching result of the attribute type part and the matching result of the attribute entity type part in step 64, the corresponding attribute is set as the attribute and attribute entity of the text area.

  In the case of image input, as shown in FIG. 6B, in step 65, an image is input by the image input means. In step 66, layout analysis is performed on the input image, and OCR processing is performed for each text region. In step 67, each text area is collated with the contents of the attribute type portion of a plurality of attribute dictionaries. In step 68, each text area is collated with the contents of each dictionary in the attribute entity contents section of the plurality of attribute dictionaries. If they match, the dictionary group name in the attribute entity contents section and the attribute dictionary type section in the attribute dictionary are matched. Match group names. If there is a matching attribute dictionary based on the matching result of the attribute type part and the matching result of the attribute entity type part in step 69, the corresponding attribute is set as the attribute and attribute entity of those text areas.

  In this way, in this embodiment, the attribute dictionary has an attribute type part, an attribute entity type part, and an attribute entity content part, and the attribute entity content part is made independent and can be shared among the attribute dictionaries. If specified, the attribute entity content part can be referred to.

  FIG. 7 is a diagram showing another embodiment of the document attribute acquisition method according to the present invention. FIG. 7A is a block diagram showing an example of a document attribute acquisition device, and FIG. 7B is a diagram for explaining the document attribute acquisition method. FIG. This embodiment differs from the embodiment of FIG. 4 in that a character string arrangement determination unit 37 is further provided as shown in FIG. The character string arrangement determination unit 37 receives information from the character region extraction unit 32, performs character string arrangement determination, and outputs this to the attribute collation unit 35. The character string arrangement | positioning determination part 37 refers to the linearity of arrangement | positioning of each text area | region 15, for example. In this embodiment, as shown in each text area 15 in FIG. 7B, a text area whose arrangement is arranged in a straight line in the direction of the arrow in the figure is selected, and the selected text area is referred to. Get the attributes and attribute entities of the text area.

  FIG. 8 shows an example of a flowchart for executing the embodiment of FIG. 7. FIG. 8A shows a case of inputting document information, and FIG. 8B shows a case of inputting an image. In the case of inputting document information, as shown in FIG. 8A, in step 81, the document information is input by the information input means. In step 82, each text region is compared with the attribute of the attribute type portion of the plurality of attribute dictionaries, and the order of matching attributes is determined. In step 83, each text area is checked against the attribute entities in the attribute entity type part of the plurality of attribute dictionaries, and the order of the matching attribute entities is determined. In step 84, a text area in which the arrangement on the document is arranged on a straight line is selected for each text area. In step 85, the attribute ranking result, the attribute entity ranking result, and the selected text area in the arrangement linearity are referred to, and the attribute corresponding to the higher rank and the high arrangement linearity is set as the attribute and attribute entity of the text area.

  In the case of image input, as shown in FIG. 8B, an image is input by the image input means in step 86. In step 87, layout analysis is performed on the input image, and OCR processing is performed for each text area. In step 88, each text region is compared with the attribute of the attribute type portion of the plurality of attribute dictionaries, and the order of the matched attributes is determined. In step 89, each text area is checked against the attribute entities in the attribute entity type part of the plurality of attribute dictionaries, and the order of the matching attribute entities is determined. In step 90, a text area in which the arrangement on the document is arranged on a straight line is selected for each text area. In step 91, the attribute ranking result, the attribute entity ranking result, and the selected text area in the arrangement linearity are referred to, and the attributes corresponding to the higher ranking and the high arrangement linearity are set as the attributes and attribute entities of the text areas.

  Thus, in this embodiment, the attribute dictionary has an attribute type part and a plurality of attribute entity type parts, and selects an attribute dictionary for each text area. In this case, referring to the arrangement linearity of each text area, This selection result is used to determine and select related friends and perform association as a document structure.

  FIG. 9 is a diagram showing another embodiment of the document attribute acquisition method according to the present invention. This embodiment is different from the embodiment of FIG. 7 in that distance information between text areas is referred to. Also in this embodiment, the character string arrangement determination unit 37 shown in FIG. In the present embodiment, the character string arrangement determination unit 37 calculates the distance between the text areas 15 as indicated by the arrows in FIG. 9, and refers to the calculated distance to attribute and attribute entity of the text area. To get.

  FIG. 10 shows an example of a flowchart for executing the embodiment of FIG. 9, where (a) is for inputting document information and (b) is for inputting image. In the case of inputting document information, as shown in FIG. 10A, in step 101, the document information is input by the information input means. In step 102, each text region is compared with the attribute of the attribute type portion of the plurality of attribute dictionaries, and the order of matching attributes is determined. In step 103, each text area is checked against the attribute entities in the attribute entity type part of the plurality of attribute dictionaries, and the order of the matching attribute entities is determined. In step 104, a distance on the document is calculated for each text area. In step 105, the attribute rank result, the attribute entity rank result, and the document distance are referred to, and the attributes corresponding to the high rank and the short distance are set as the attributes and attribute entities of the text areas.

  In the case of image input, as shown in FIG. 10B, in step 106, an image is input by an image input means. In step 107, layout analysis is performed on the input image, and OCR processing is performed for each text area. In step 108, each text region is compared with the attribute of the attribute type portion of the plurality of attribute dictionaries, and the order of the matching attributes is determined. In step 109, each text area is checked against the attribute entities in the attribute entity type part of the plurality of attribute dictionaries, and the order of the matching attribute entities is determined. In step 110, the distance on the document is calculated for each text region. In step 111, the attribute ranking result, the attribute entity ranking result, and the distance on the document are referred to, and the attributes corresponding to the high ranking and the short distance are set as the attributes and attribute entities of the text regions.

  Thus, in this embodiment, the attribute dictionary has an attribute type part and a plurality of attribute entity type parts, and selects an attribute dictionary for each text area. In this case, referring to distance information between the text areas, This selection result is used to determine and select related friends and perform association as a document structure.

  The above document attribute acquisition method can be executed by a computer program. That is, a function for providing a plurality of attribute dictionaries having an attribute type section storing attribute types and an attribute entity type section storing attribute entity types, characters from a plurality of text areas or input images obtained from document information A function of inputting a plurality of text areas obtained by recognition, a function of collating the contents of each text area with the contents of the attribute type part of the plurality of attribute dictionaries, the contents of each text area and the plurality of attribute dictionaries A function for matching the contents of the attribute entity type part, a function for selecting an attribute dictionary in which the matching result of the attribute type part matches the matching result of the attribute entity type part, and an attribute type part of the selected attribute dictionary And a program for realizing the function of acquiring the attribute of the text area and the attribute entity from the attribute entity type part. This program can be uploaded or downloaded via a telecommunication line such as the Internet, and can be provided as a computer-readable recording medium on which the program is recorded.

  Thus, it is possible to provide a document attribute acquisition method, a document attribute acquisition device, a computer program therefor, and a recording medium recording the program that enable high-quality document structuring.

  The present invention relates to a document attribute acquisition method and apparatus for acquiring an attribute of a document from a document image, and a recording medium on which a program for acquiring the document attribute is recorded, and has industrial applicability.

(A), (b) is a figure which shows one Example of the document attribute acquisition method based on this invention. (A), (b) is a figure which shows the other Example of the document attribute acquisition method based on this invention. FIG. 7 is a diagram illustrating another embodiment of a document attribute acquisition method according to the present invention, in which FIG. 5A is a block diagram illustrating an example of a document attribute acquisition apparatus, and FIG. 5B is a diagram for describing the document attribute acquisition method; is there. FIG. 4 shows an example of a flowchart for executing the embodiment of FIG. 3, in which (a) is a case of inputting document information and (b) is a case of inputting an image. It is a figure which shows the other Example of the document attribute acquisition method based on this invention. FIG. 6 shows an example of a flowchart for executing the embodiment of FIG. 5, where (a) is a case of inputting document information, and (b) is a case of inputting an image. FIG. 7 is a diagram illustrating another embodiment of a document attribute acquisition method according to the present invention, in which FIG. 5A is a block diagram illustrating an example of a document attribute acquisition apparatus, and FIG. 5B is a diagram for describing the document attribute acquisition method; is there. FIG. 8 shows an example of a flowchart for executing the embodiment of FIG. 7, where (a) is a case of inputting document information, and (b) is a case of inputting an image. It is a figure which shows the other Example of the document attribute acquisition method based on this invention. FIG. 10 shows an example of a flowchart for executing the embodiment of FIG. 9, where (a) is a case of inputting document information, and (b) is a case of inputting an image.

Explanation of symbols

1-4 Attribute dictionary 5 Document 6 Text area 7, 8 Attribute pair

Claims (6)

An attribute dictionary storage means for storing a plurality of attribute dictionaries generated for each attribute , having an attribute type part storing a plurality of attribute types and an attribute entity type part storing a plurality of attribute entity types;
Obtaining means for obtaining a plurality of text regions including character information obtained by character recognition from an input image;
Attribute collation means for collating the character information of each text area acquired by the acquisition means with the attribute type of the attribute type portion of the plurality of attribute dictionaries;
Attribute entity matching means for matching the character information of each text area acquired by the acquisition means with the attribute entity type of the attribute entity type portion of the plurality of attribute dictionaries;
A selection means for selecting, from the plurality of attribute dictionaries, an attribute dictionary including the attribute type matched by the attribute matching means and the attribute entity type matched by the attribute entity matching means;
Text including character information that matches the attribute type with respect to the attribute type and the attribute entity type stored in one attribute dictionary of the attribute dictionary with reference to the attribute dictionary selected by the selection means Extraction means for extracting a region and a text region including character information matching the type of the attribute entity ;
And attributes of the one attribute dictionary, outputs the character information of the text area corresponding to the type of the stored attribute entity to the one attribute dictionary of the text area extracted by the extracting unit, the attribute and attribute entity respectively Output means for
A document attribute acquisition apparatus comprising:   2. The document attribute acquisition apparatus according to claim 1, further comprising an attribute entity content part in which the attribute dictionary stores a group name of attribute entities to be collated with the attribute entity type part.   3. The document attribute acquisition apparatus according to claim 2, wherein the attribute entity content part is shared among the plurality of attribute dictionaries.   4. The document attribute acquisition apparatus according to claim 1, wherein at least one of the attribute type of the attribute type part and the attribute entity type of the attribute entity type part has a rank for collation. . Computer
An attribute dictionary storage means for storing a plurality of attribute dictionaries generated for each attribute , having an attribute type part storing a plurality of attribute types and an attribute entity type part storing a plurality of attribute entity types;
Acquisition means for acquiring a plurality of text regions including character information obtained by character recognition from an input image;
Attribute collation means for collating the acquired character information of each text region and the attribute type of the attribute type portion of the plurality of attribute dictionaries;
Attribute entity collating means for collating the acquired character information of each text area with the attribute entity type of the attribute entity type section of the plurality of attribute dictionaries;
A selection means for selecting, from the plurality of attribute dictionaries, an attribute dictionary including the attribute type matched by the attribute type part matching and the attribute entity type matched by the attribute entity type part matching;
With reference to the selected attribute dictionary, for the attribute type and the attribute entity type stored in one attribute dictionary of the attribute dictionary, a text area including character information that matches the attribute type, extracting means for extracting a text region containing the character information that matches the type of the attribute entity,
And attributes of the one attribute dictionary, output means for outputting the character information of the text area corresponding to the type of attribute entities stored in the one attribute dictionary, as the attribute and attribute entity each of the extracted text region ,
A computer-readable recording medium in which a program for functioning as a computer is recorded. Computer
An attribute dictionary storage means for storing a plurality of attribute dictionaries generated for each attribute , having an attribute type part storing a plurality of attribute types and an attribute entity type part storing a plurality of attribute entity types;
Acquisition means for acquiring a plurality of text regions including character information obtained by character recognition from an input image;
Attribute collation means for collating the acquired character information of each text region and the attribute type of the attribute type portion of the plurality of attribute dictionaries;
Attribute entity collating means for collating the acquired character information of each text area with the attribute entity type of the attribute entity type section of the plurality of attribute dictionaries;
A selection means for selecting, from the plurality of attribute dictionaries, an attribute dictionary including the attribute type matched by the attribute type part matching and the attribute entity type matched by the attribute entity type part matching;
With reference to the selected attribute dictionary, for the attribute type and the attribute entity type stored in one attribute dictionary of the attribute dictionary, a text area including character information that matches the attribute type, extracting means for extracting a text region containing the character information that matches the type of the attribute entity,
And attributes of the one attribute dictionary, output means for outputting the character information of the text area corresponding to the type of attribute entities stored in the one attribute dictionary, as the attribute and attribute entity each of the extracted text region ,
Program to function as.

Images