JP2006053612A - Document conversion device and method, and document conversion program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a document processor capable of properly extracting a table structure or a portion of a structure similar to a table in a document, and generating an output document wherein the extracted table structure or protion of the structure similar to the table can be used as a structured document. <P>SOLUTION: This document processor 1 has: a knowledge dictionary part 7 defining attribute impartment information; a conversion rule part 9 defining output restriction information; an attribute impartment part 3 deciding presence/absence of a predetermined character string to a character string appearing in a specified conversion target part, and imparting attribute information corresponding to the predetermined character string on the basis of the attribute impartment information of the knowledge dictionary part; a conversion mapping generation part 4 determining regularity of appearance of the predetermined attribute information defined in the output restriction information from the attribute information imparted by the attribute impartment part, and generating conversion mapping of the output restriction information; and an output document generation part 5 generating and outputting the output document from information about the generated conversion mapping in a form according with the output restriction information. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a document conversion apparatus and method, and a document conversion program, and more particularly to a document conversion apparatus and method for extracting a table structure in a document or a portion having a structure similar to a table, converting the extracted portion into a structured document, and a document Concerning conversion program.

  In recent years, document structuring has attracted attention. When creating a structured document by adding a tag to an electronic text document, a person reviews the description of each input document, and the correspondence between the structure of the input document and the output document, the so-called conversion map, is converted manually. This was realized by writing as a script or conversion program, or by assigning a marking symbol that is a clue to conversion, and then performing batch conversion using a script.

On the other hand, a document may include a so-called tabular part. As a technique for structuring such a tabular portion, there is a technique described in JP-A-2001-325284. According to this technique, each element of the table structure can be recognized.
JP 2001-325284 A

  However, in the case of the technique described in Patent Document 1 described above, the tabular part must be clearly specified and input to the document processing apparatus. Because of the variety of ways to describe a table structure or a structure similar to a table, it is difficult to create a structured document by using the above technology to correctly extract the contents from a document that prioritizes browsing was.

  Therefore, the present invention appropriately outputs an output document that can appropriately extract a table structure or a structure portion similar to the table in the document and use the extracted table structure or structure portion similar to the table as a structured document. An object of the present invention is to provide a document processing apparatus that can be generated.

  The document processing apparatus according to the present invention includes a knowledge dictionary unit that defines attribute assignment information for assigning attribute information to a predetermined character string, and performs conversion on a conversion target portion of an input document to output an output document. A conversion rule part that defines output constraint information for outputting the character string, and the presence / absence of the predetermined character string with respect to the character string that appears at the specified location to be converted, An attribute assigning unit that assigns the attribute information corresponding to the predetermined character string based on the attribute assigning information, and the attribute information given by the attribute assigning unit, is defined in the output constraint information A conversion map generation unit for determining regularity of appearance of predetermined attribute information, and generating a conversion map of the output constraint information based on the determined regularity of appearance, and information of the conversion map The Serial generates output document matches the format in the output restriction information and an output document generator outputting.

  According to the document processing apparatus of the present invention, a table structure in a document or a portion having a structure similar to a table is appropriately extracted, and the extracted table structure or a portion having a structure similar to a table is used as a structured document. An output document that can be generated can be generated.

  Embodiments of the present invention will be described below with reference to the drawings.

  FIG. 1 is a functional block diagram showing the configuration of the document structuring apparatus according to the present embodiment. The document structuring apparatus 1 includes a conversion target location specifying unit 2, an attribute analysis unit 3, a conversion mapping generation unit 4, a conversion execution unit 5, a knowledge dictionary analysis unit 6, a knowledge dictionary unit 7, and a conversion rule. An analysis unit 8 and a conversion rule unit 9 are included. The document structuring apparatus 1 is a computer device having a storage device capable of storing various programs and various data, for example. The programs and necessary data for each part are stored in a storage device of the computer.

  Each “unit” in this specification is a conceptual one corresponding to each function of the embodiment, and does not necessarily correspond to a specific hardware or software routine on a one-to-one basis. Therefore, in the present specification, the embodiment will be described below assuming a virtual circuit block (unit) having each function of the embodiment. In addition, each step of each procedure in the present embodiment may be executed in a different order for each execution by changing the execution order and performing a plurality of steps at the same time, as long as it does not contradict its nature.

  Note that the document structuring apparatus 1 can use, for example, a general-purpose computer device as basic hardware. Then, the conversion target location specifying unit 2, the attribute analysis unit 3, the conversion map generation unit 4, the conversion execution unit 5, the knowledge dictionary analysis unit 6, and the conversion rule analysis unit 8 execute a program on a processor mounted on the computer device. This can be realized. At this time, the document structuring apparatus 1 may install the above program in a computer device in advance, or record the program on a removable storage medium such as a CD-ROM or via a network. It may be realized by distributing and installing this program on a computer device as appropriate. The knowledge dictionary unit 7 and the conversion rule unit 9 include a storage device such as a memory and a hard disk device built in the computer device, a storage device such as a memory and a hard disk device externally attached to the computer, and a floppy (registered trademark). It can be realized by appropriately using a removable storage medium such as a disk.

  In the document structuring apparatus 1, input document data is input to the conversion target location specifying unit 2. For example, as shown in FIG. 2, the input document is a document including a table structure or a portion having a structure similar to the table. FIG. 2 is a diagram illustrating an example of an input document. A table structure or a structure similar to a table includes rows and columns, and includes information on cell contents that are elements of the table.

  Hereinafter, in this specification, a part of a table structure or a structure similar to a table is simply referred to as a tabular part. Further, in the following description, the contents of each cell are shown as an example in which character string information is included.

  The conversion target location specifying unit 2 receives the input document and specifies the target location or target area to be structured from the input document. As shown in FIG. 2, the input document includes a tabular part and a non-tabular part. For example, an input document printed on paper or the like shown in FIG. 2 is read and converted by an OCR (optical character reader), and is converted as character string data, that is, text data as shown in FIG. FIG. 3 is a diagram illustrating an example of an input document converted into text data. Note that some OCR software may convert the character string data into a format other than the format shown in FIG. 3, but here the input document is assumed to be converted into text data in the format shown in FIG. explain. Note that the input document is not limited to the output of the OCR software, but has been manually created as text data in advance, for example, data including the table format displayed in the browser is stored as text data by cut & paste But it ’s okay.

  The necessary conditions that such an input document format should have are that each text content (including white space) corresponding to each element of the table is replaced with line breaks, tabs, and other separators in data that is centrally listed as text data. It is assumed that the elements can be clearly distinguished by symbols and the like, and the elements in the table structure may be in the row direction or the column direction, but are listed according to a certain order.

  The attribute analysis unit 3 first extracts a text element from the conversion target part specified by the previous conversion target part specifying unit 2. Note that the text element here indicates each line when it is divided in units of a solid text as shown in FIG. If the input is already a tagged document, for example, a document in HTML (Hyper Text Markup Language) format with a tag of coarse granularity, each character string enclosed by the tag is indicated.

  The knowledge dictionary unit 7 registers keywords and the like to be collated in order to give attribute information in the attribute analysis unit 3 in advance. The knowledge dictionary analysis unit 6 defines the meaning of a vocabulary of a keyword such as a predetermined character string and attribute information to be given to the keyword. The knowledge dictionary unit 7 supplies keywords and attribute information to the attribute analysis unit 3 via the knowledge dictionary analysis unit 6 or via the knowledge dictionary analysis unit 6 and the conversion rule analysis unit 8 as indicated by the dotted line in FIG. .

  The attribute analysis unit 3 refers to the knowledge dictionary unit 7 and assigns an attribute value as attribute information for each text element. The assigned attribute value is output to the conversion map generation unit 4 together with the corresponding original text element as an attribute analysis result. The processing contents of the attribute analysis unit 3 will be described later.

  The conversion rule unit 9 defines output constraint information for outputting a structured document.

The association between the input document and the output document structure, which is the content of the conversion rule unit 9, is defined by the user in order to obtain a structured document in a format desired by the user.
The contents of the conversion rule part 9 are defined by the user in order to obtain a structured document in a format desired by the user.

  The output constraint information is supplied to the conversion map generation unit 4 via the conversion rule analysis unit 8. In the present embodiment, the conversion rule unit 9 includes the attribute information to be attributed, and the information of the conversion rule unit 9 is also supplied to the attribute analysis unit 3 via the conversion rule analysis unit 8. Has been.

  The conversion map generation unit 4 generates a conversion map based on the regularity of appearance of predetermined attribute values from among the attribute values of each text element assigned by the attribute analysis unit 3 in the previous stage. The processing contents of the conversion map generation unit 4 will be described later.

  The conversion execution unit 5 uses the result of the conversion map generation obtained in the process up to the previous stage, and converts the text element obtained in the process so far into a predetermined post-conversion structure defined in the conversion rule. An output document generation unit that obtains a desired structured document, for example, XML (Extensible Markup Language) data. The processing contents of the conversion execution unit 5 will be described later.

  First, the overall flow of processing in the document structuring apparatus 1 will be described. FIG. 4 is a flowchart showing an example of the overall flow of the structuring process performed on the input document in the document structuring apparatus 1.

  For example, a user who wants to perform a structuring process on a document including a table structure part inputs the document into the document structuring apparatus 1. If the input document is document data including electronic text data, the input document is electronic data. If the original document is a paper document, the input document is electronic data converted from image data to text data by OCR software. is there.

  The conversion target part is specified for the input data (step (hereinafter abbreviated as S) 1). The process of S1 is executed in the conversion target location specifying unit 2. The conversion target location specifying unit 2 registers in advance a word that starts a partial structure such as a “side effect” that is a surface expression in an input document in a predetermined storage area of a computer storage device in advance. The area corresponding to the chapter or section is identified as an area located between the words. Here, it is assumed that the portion after the keyword “4. Side Effects” of the input document in FIG.

  It should be noted that a specific chapter or the like may be designated by a person manually or by specifying which region or range in the input document is a conversion target portion.

  By the processing of the conversion target part specifying unit 2, only a part or a region to be converted (hereinafter referred to as a conversion target part) is specified, that is, extracted from the input document input by the user.

  Then, the attribute analysis unit 3 extracts text elements from the data of the conversion target part extracted by the conversion target part specifying unit 2 (S2), and assigns an attribute value, that is, a class to each of the extracted text elements. Is performed (S3).

  Finally, the conversion map generation unit 4 extracts regularity of class appearance (S4) and determines the conversion map (S5). Finally, the conversion execution unit 5 executes a conversion process for embedding a text element in a predetermined post-conversion structure (S6).

Next, the processing of each unit will be described in detail.
As described above, the attribute analysis unit 3 extracts a text element (S2) from the data of the conversion target portion, and further assigns an attribute (S3). These processes in the attribute analysis unit 3 will be described in more detail with reference to FIG.

FIG. 5 is a flowchart showing an example of the processing flow of the attribute analysis unit 3.
As shown in FIG. 5, first, conversion rules are read (S11). FIG. 6 shows a description example of the conversion rule. In the conversion rule in FIG. 6, a portion 101 surrounded by <extraction target> tags is an extraction target attribute description portion in which attributes extracted and assigned by the attribute analysis unit 3 are described. In other words, designation of which attribute is selected and attributed is performed by reading this conversion rule. In addition, in the conversion rule, when there is no designation of the extraction target, an attribute predetermined as a default is selected for analysis. In FIG. 6, four attributes of “label”, “item name”, “cell contents”, and “comment line” are described as extraction targets to be analyzed.

  Hereinafter, the attribute for each text element is referred to as classification, classification, or simply class.

  Next to S11, it is determined whether or not the part specified from the document at the previous conversion target part, which is the conversion target part, is a tagged document (S12). If the conversion target location can be analyzed as a tagged document (YES in S12), the conversion target location is expanded as a tree structure, for example (S13). Next, text elements are extracted with priority given to the depth of the tree structure. At this time, the parent tag name of each text element and the root of the tree structure in which the text element extracted at this point represents the entire document are started. As for the depth, the value such as the depth is stored as an attribute value extracted from the text element (S14).

  On the other hand, if there is no structure in the conversion target portion and the text is a so-called solid text, NO is determined in S12, each line is read, and each line is stored as one text element (S15). By the processing so far, the text element is stored together with the appearance order in the document regardless of whether the portion to be converted is a tagged document or a solid text document.

  Specifically, a text element list as shown in FIG. 7 is created from the text data of the input document shown in FIG. FIG. 7 is a diagram illustrating an example of a text element list. As shown in FIG. 7, an identifier corresponding to the appearance order, that is, a text identifier (hereinafter referred to as a text element ID) is assigned to each text element. That is, when the processing from S12 to S15 in FIG. 5 is executed on the input document in FIG. 3, a text element list as shown in FIG. 7 is generated. Since the text element ID represents the order of appearance of each text element in the input document, the conversion mapping generation unit 4 applies the predetermined specific attribute values in any order as will be described later. It can be used to determine whether or not a specific set of attribute values appears continuously.

  In response to this result, in the subsequent attribute assigning process (S16), it is checked whether each text element has a predetermined character string length, or whether a predetermined character type and a predetermined character string are included. Then, the attribute value of each line, that is, the attribute value of each text element is determined, and the attribute value is assigned to each text element. As described above, attribute information corresponding to a keyword such as a predetermined character string is defined and stored in advance in the knowledge dictionary unit 7 as shown in FIG. 8, for example.

  FIG. 8 is a diagram for explaining an example of information for extracting item names and labels defined in the knowledge dictionary unit 7. That is, as shown in FIG. 8, the character string described on the right side of “;” and the attribute information corresponding to this character string are described on the left side of “;”. It is registered in.

  The process in attribute assignment process S16 is demonstrated. FIG. 9 is a flowchart showing an example of the flow of attribute assignment processing S16. First, the information of the extraction target attribute description unit 101 is read from the description in the conversion rule (S21). This is because each element described in the area surrounded by the <extraction target> tag in FIG. 6, elements of “label”, “item name”, “cell contents”, and “comment line” in the example of FIG. Is equivalent to reading.

  Next, one line (text element) to be processed is read (S22). Then, it is determined whether or not the attribute is the extraction target attribute from the information of the extraction target attribute description unit 101 read in S21 (S23, S25, S27, S29). It is determined whether or not the character string related to the attribute information specified in the knowledge dictionary section 7 shown in FIG. 8 is included in the text element of the line read in S22 (S24, S26, S28, S30).

  First, define words. The “item name” here is a word that generally indicates an element that appears in the table. When the example of FIG. 2 is used as this specific example, “hypersensitivity”, “bleeding tendency”, “ It is assumed that an element appearing in the first column such as “blood” is shown. Further, “item” in the knowledge dictionary is an identifier given in advance to a word or expression corresponding to “item name” in a document in a certain field.

  A specific processing example is shown below. It is determined whether or not a character string corresponding to the item name (character string corresponding to “item” in FIG. 8) appears in the text element of the processing target line read in S22 (S23). In the example of FIG. 6, since the extraction of the item name is described (instructed), YES in S23, referring to the dictionary list shown in FIG. 8, a character string corresponding to the attribute information “item”, for example, If “hypersensitivity”, “blood”, etc. are included in the sentence of the current processing target line (text element), the attribute value of “item name” is assigned to this processing target line (text element). (S24).

  Similarly, it is determined whether or not a character string corresponding to the label name (character string corresponding to “label” in FIG. 8) appears in the text element of the processing target line read in S22 (S25). . In the example of FIG. 6, since the extraction of the label name is described (instructed), YES is obtained in S25, and the character string corresponding to the attribute information “label”, for example, by referring to the dictionary list shown in FIG. If "0.1 to less than 5%", "Frequency unknown", etc. are included in the sentence of the current processing target line (text element), the "label" for this processing target line (text element) Is assigned (S26).

  Similarly, it is determined whether or not the text element of the processing target line read in S22 has cell contents (S27). The cell content refers to an element in which data contents are described, not a label or an item name, among table elements. In the example of FIG. 6, since the extraction of cell contents is described (instructed), the result of S27 is YES, and the processing target line (text element) is a predetermined character string length threshold (for example, 20 characters). If a specific character string such as a keyword, such as “anemia”, is included in the sentence of the current processing target line (text element), this processing target line is determined to be cell contents. Then, an attribute value of “cell contents” is assigned to the processing target line (text element) (S28).

  Furthermore, it is determined whether or not the text element of the processing target line read in S22 is a comment line (S29). In the example of FIG. 6, since extraction of a comment line is described (instructed), the answer to S29 is YES, and the processing target line (text element) is a predetermined character string length threshold (for example, 20 characters). Or a specific character string such as a keyword, such as “Note 1”, is included in the sentence of the current processing target line (text element), this processing target line is determined as a comment line. Then, an attribute value of “comment line” is given to the processing target line (S30).

  In addition, in the dictionary example referred to in this case, a character string including a regular expression as a character string appearing in the processing target line (text element) is described, and an attribute value is given to the processing target line. The determination process is performed.

  Next to S29 or S30, it is determined whether or not the above-described processing has been performed for all the rows in the conversion target portion (S31). If the process has not been completed for all rows in S31 (NO in S31), the process returns to S22. If processing has been completed for all lines in S31 (YES in S31), a temporary attribute of “others” is temporarily assigned to the line to which no attribute is finally given, and the text element of each line and its corresponding In addition to the text ID, the attribute value data given by the attribute analysis unit 3 is passed to the subsequent processing (S32).

  As described above, the attribute analysis unit 3 uses the information in the knowledge dictionary unit 7 to determine whether or not there is a predetermined character string in the character string that appears in the conversion target portion. Assign an attribute value to (text element). The attribute analysis unit 3 stores the processing result as an attribute value for each line (text element) in a storage device in the computer.

  As described above, the attribute analysis unit 3 extracts text elements and assigns attribute values to the extracted text elements with respect to the data of the conversion target portions extracted by the conversion target portion specifying unit 2. The analysis result data is stored in a storage device in the computer.

  In addition, in acquiring the attribute value when the attribute value is given to the text element, the morpheme analysis is performed on the text element using the morpheme analysis unit 10 shown by the dotted line in FIG. The relationship between the attribute value to be assigned to the morpheme analysis result is prepared in advance as a morpheme analysis rule, and whether or not a specific combination of parts of speech appears in the text element using the morpheme analysis result An attribute value may be assigned based on the attribute value.

  After the processing as described above is performed on all the text elements in the conversion target portion, the processing is terminated, and the finally obtained analysis result is supplied to the subsequent conversion map generation unit 4.

  Next, processing in the conversion map generation unit 4 will be described.

  First, the processing for obtaining regularity of class appearance (S5) in FIG. 4 will be described. Specifically, the results of classification (attribute assignment) associated with each text element are viewed in the order in which the text elements appear, and the regularity of the text elements classified into similar classes appears. Judgment is made. Here, as a result of the classification, it is assumed that the attributes of the text element IDs 3, 4, 5 and 19, 20, 21 are classified as “labels” as shown in FIG. FIG. 10 is a diagram showing a specific example of assigned attribute values and text element IDs.

  According to this example, it is understood that the text element ID of the “label” that first appears in the document is ID3, and continues from here to ID5. Since the “label” continues three elements in succession, it is assumed that three “cell contents” corresponding to the three elements appear repeatedly in this subsequent text element, and the repeated structure. To extract.

  As a result, the text elements after ID 6 are each divided into [[item name], [cell contents]], and a specific example of the text element ID is [[6], [7, 8]. , 9]], [[10], [11, 12, 13]], [[14], [15, 16, 17]], [[18], [19, 20, 21]]... Become. This repeated structure is the same until a text element with a different classification such as the appearance of a "comment line" or a text element with a different "label" appears, or until the location to be converted ends. Assuming that the repeating structure continues, an attribute value of “cell contents” is assigned to the text element. In order to determine the end point of this example, when looking at FIG. 10 to which the attribute value that is the result of classification is given, it is determined that “label” is ID19-21 in addition to ID3-5 There is a text element.

  For this reason, the extraction target part of the repeated structure based on the ID 3-5 of the label is the structure ([[14], [15, 16, 17] in which the repeated structure is correctly acquired immediately before this except for the text elements after ID 19). ]) Is assumed to be the end of this extraction processing target.

  Subsequently, a text element classified as “label” is searched for as a repeated structure to be extracted next. According to the example of FIG. 10, it can be seen that a text element corresponding to the “label” appears in the text element of ID 19-21. Therefore, it is assumed that the repeated structure of the three cell contents corresponding to the three elements appears in the text element subsequent to the ID 19-21, and the repeated structure is extracted. . Although the end of the extraction process in this case, there is a text element to which “comment element” is assigned as an attribute value in ID38-39, the text element after ID38 is excluded from the extraction target. Therefore, the combinations based on the IDs 19-21 are finally [[22] [23, 24, 25]], [[26], [27, 28, 29]], [[30], [31, 32, 33]], [[34], [35, 36, 37]].

  Since there is no element corresponding to “label” for subsequent text elements, the “label” extraction process ends.

  As described above, the conversion map generation unit 4 estimates the appearance of the text element from the attribute and outputs the result by applying the result to the output constraint. Thus, even when information cannot be easily extracted or the information extraction result is inappropriate, the tabular portion can be structured.

  Finally, the text element corresponding to “cell contents” is extracted. This can be obtained, for example, by removing elements whose class has been determined in the process so far from all text elements in the conversion target portion. For example, it is determined by a defining formula of “cell contents” = (all text elements) − “label” − “item name” − “comment line” − (text element before label appearance). A text element obtained on the basis of this definition formula is defined as “cell content”. The portion described as “cell contents” in FIG. 10 shows this result, and the text element ID information of “cell contents” is shown in this result.

  As described above, the conversion map generation unit 4 outputs from the attribute information (in the example, “item name”, “label”, “cell contents”, “comment line”) given by the attribute analysis unit 3. For the predetermined attribute information (for example, “label”) defined in the constraint information, the regularity of appearance is determined, and a conversion map of the output constraint information is generated based on the determined regularity of appearance. The conversion map generation unit 4 supplies the information of the conversion map to the subsequent conversion execution unit 5.

  As shown by a dotted line in FIG. 1, the user confirms the conversion map generated in the conversion map generation unit 4 using the output unit 11 such as a monitor connected to the computer or the input unit 12 such as a keyboard. Then, a conversion map may be presented to the user so that the verification can be performed, or a user input for enabling the user to correct or edit the mapping for the presented conversion map may be accepted.

  Next, the processing procedure of the conversion execution unit 5 as the output document generation unit will be described with reference to the flowchart of FIG. FIG. 11 is a flowchart illustrating an example of a process flow of the conversion execution unit 5. First, an attribute analysis result including a conversion map generated up to the previous stage and a conversion rule are read (S31). At this time, various output constraint information 102 and 103 shown in FIG. 6 are read from the conversion rules prepared in advance.

  In the output constraint information given as an example in FIG. 6, the structure after conversion is specified inside the “<structure example>” tag. First, if you look inside the 102 elements enclosed in “<Structure Example =“ 1 ”>”, there are 4 tags <serialno>, <item>, <frequency>, and <description> respectively. Furthermore, each has a text node (each value is “$ count”, ““ item name ””, ““ label ””, ““ cell contents ””) as its child node.

  In this way, when the classification result is included in the text node, it is verified whether the classification result used in the text node and the classification classification notation in the attribute analysis result are the same. Is done. When the text node does not include a class classification notation, the post-conversion structure described in the conversion rule is replaced with an existing structure as a new structure. As a result of the verification, if the class classification result of the text node is the same as the class classification result included in the attribute analysis result, the text node is replaced with the corresponding text element in the attribute analysis result. In addition, when the notation “$ count” exists in the text node in the conversion rule, it is replaced with a numerical value that increases each time this structure is referenced. This process continues as long as there is an unprocessed text element in the corresponding attribute analysis result.

  First, the value described in the “example” attribute of the “<structure example>” tag is read (S32). Then, it is determined whether or not the “example” attribute is “1” (S33). In the case of FIG. 6, since “1” is set as the “example” attribute of the “<structure example>” tag of the output constraint information 102, YES is obtained in S 33 and the process proceeds to S 34. In S34, every time the final text attribute “cell contents” appearing in the structure example tag appears in the document, the output document is repeatedly output. In other words, the output count of the output document is the number of times the text attribute “cell contents” appears in the document. In this case, the output document is information on the structure of the output document, and the output document output from the conversion execution unit 5 is a document that can be converted into a structured document.

  At this time, the number of occurrences of “cell contents”, which is the number of times of repeated output, does not match the number of occurrences of “item name” and “label”. For “Name”, the content used immediately before is repeatedly applied. In addition, since it is known that “Label” consists of 3 sets of repetitions, the contents of “Cell contents” (including empty elements) are multiplied by each element of 3 sets in order, ie, cell Set other attribute values for each content as the output structure of the converted structure.

  For example, in FIG. 7, the text element of ID6 is an item name, and the text elements of ID3 to ID5 immediately before this have continuous labels, so the horizontal direction of the table format is ID3, ID4, and ID5. It is assumed that the three labels of the text element correspond to each other, and the item name of the ID6 text element corresponds to the vertical direction of the table format. Since the attribute of ID7 and the text contents are the cell contents “rash, painful feeling, etc.”, as shown in FIG. 12, the next output document is generated in a format that matches the output constraint information 102 of Structural Example 1. And output. FIG. 12 is a diagram showing a specific example of the final output document structure for the conversion target portion.

  Depending on the attribute assigned by the attribute analysis unit in the previous stage, there is a case where there is no designation in the conversion rule (in this example, a text element assigned with the “other” attribute corresponds). As described above, when there is no designation of the corresponding attribute in the conversion rule, the corresponding text element cannot appear in the final output document, and is therefore removed as a result.

<serialno> 1 </ serialno>
<item> Hypersensitivity </ item>
<frequency><0.1-5%</frequency>
<description> Rash, painful sensation, etc. </ description> ・ ・ ・ Output (1)
Next, since the cell contents of ID8, ID9, and ID11 are empty elements, after the next output document is generated in a format that matches the output constraint information 102 of Structural Example 1, it is rejected and not output.

  Next, since the ID 12 is the cell content “subcutaneous bleeding, hematuria, etc.”, the next output document is generated and output in a format that matches the output constraint information 102 of the structural example 1.

<serialno> 2 </ serialno>
<item> Bleeding tendency </ item>
<frequency><0.1%</frequency>
<description> Subcutaneous hemorrhage, hematuria, etc. </ description> ・ ・ ・ Output (2)
Similarly, since ID13 is an empty element, it is rejected and not output after the next output document is generated in a format that matches the output constraint information 102 of Structural Example 1.

  Similarly, since the ID 15 is the cell content “anemia etc.”, the next output document is generated and output in a format that matches the output constraint information 102 of the structural example 1.

<serialno> 3 </ serialno>
<item> Blood </ item>
<frequency><0.1-5%</frequency>
<description> Anemia etc. </ description> ・ ・ ・ Output (3)
Similarly, since ID16 and ID17 are empty elements, after the next output document is generated in a format that matches the output constraint information 102 of Structural Example 1, it is rejected and not output.

  As described above, the above three outputs are obtained based on the label ID 3-5. Similarly, the following output is obtained based on the label ID 19-21.

<serialno> 4 </ serialno>
<item> Digestive </ item>
<frequency><0.1-5%</frequency>
<description> Nausea, stomach discomfort, diarrhea </ description> ・ ・ ・ Output (4)
<serialno> 5 </ serialno>
<item> Digestive </ item>
<frequency><0.1%</frequency>
<description> Vomiting, loss of appetite, constipation, stomatitis, etc. </ description> ・ ・ ・ Output (5)
<serialno> 6 </ serialno>
<item> Liver </ item>
<frequency><0.1-5%</frequency>
<description> Liver dysfunction </ description> ・ ・ ・ Output (6)
<serialno> 7 </ serialno>
<item> Liver </ item>
<frequency> Unknown frequency </ frequency>
<description> Twilight </ description> ・ ・ ・ Output (7)
<serialno> 8 </ serialno>
<item> Kidney </ item>
<frequency><0.1%</frequency>
<description> BUN-creatine rise </ description> ・ ・ ・ Output (8)
<serialno> 9 </ serialno>
<item> Other </ item>
<frequency><0.1-5%</frequency>
<description> Increase of CK (CPK) </ description> ... Output (9)
<serialno> 10 </ serialno>
<item> Other </ item>
<frequency><0.1%</frequency>
<description> Headache, headache, etc. </ description> ・ ・ ・ Output (10)
<serialno> 11 </ serialno>
<item> Other </ item>
<frequency> Unknown frequency </ frequency>
<description> Gynecomastia </ description> ・ ・ ・ Output (11)
As described above, for output constraint information in which “1” is set as the “example” attribute of the “<structure example>” tag of the output constraint information 102, an output document is generated in a format that matches the output constraint information. . This process is repeatedly applied to the location where the cell contents appear in the input document. As a result, the output document is output in a format that matches the output constraint information.

  Next, it is determined whether or not the “example” attribute is “2” (S35). In the case of FIG. 6, since “2” is set as the “example” attribute of the “<structure example>” tag of the output constraint information 103, YES is obtained in S 35 and the process proceeds to S 36.

  In S36, when the internal element 103 surrounded by “<example of structure =“ 2 ”>” is seen, this tag has two tags <serialno> and <note>, each of which is a child element thereof. It has text nodes (each value is “$ count”, ““ comment line ””).

  Therefore, for the output constraint information in which “2” is set as the “example” attribute of the “<structure example>” tag of the output constraint information 103, an output document is generated in a format that matches the output constraint information. Specifically, as long as the text attribute at the end described in the structure example, that is, a comment line, appears in the input document, the output restriction information as a template of the structure example is applied, and the output restriction using the text element is applied. Output documents in a format that matches the information. An example of the output document is shown in FIG.

  Similarly, for the next “example” attribute defined in the conversion rule, the conversion execution unit 5 performs the conversion process and sets “n” as the “example” attribute of the last “<structure example>” tag. When the conversion process is performed (S37, S38), the document having the existing structure is replaced with the document having the converted structure to obtain an output document (S39).

  As a result, the output document shown in FIG. 12 is generated. The output document shown in FIG. 12 is in a format in which a portion having a class classification notation in a text node is replaced with a text element and output. For the replacement of the text element, the class classification (attribute value) given by the conversion map generation unit 4 is referred to.

  As described above, according to the present embodiment, even when variations occur in the format of the input document, the user does not need to create a rule for each input document as in the past, and the cost for document conversion can be reduced. Can be reduced.

  Further, for example, the table format portion read by the OCR software has different output forms depending on the OCR software, such as being separated by commas or being output in one line. As described above, even if there are variations in the format of the input document, the document processing apparatus according to the present embodiment can appropriately output a structured document such as XML.

  Furthermore, according to the document processing apparatus of the present embodiment, various existing text documents can be converted semi-automatically into a structured document (XML document) based on an application standard. Search can be realized easily.

  As described above, according to the present embodiment, it is possible to appropriately specify and extract a predetermined range in a tabular format from an input document, and create a desired output document for the extracted range. Further, since detailed restrictions such as the number of appearances, the order, and the appearance position of the same tag name are taken into account, complicated conversion is possible.

  The present invention is not limited to the above-described embodiment, and various changes and modifications can be made without departing from the scope of the present invention.

It is a block diagram which shows the block diagram of the document structure processing apparatus which concerns on embodiment of this invention. It is a figure which shows the example of the input document which concerns on embodiment of this invention. It is a figure which shows the example of the input document converted into the text data based on embodiment of this invention. It is a flowchart which shows the example of the whole flow of the structured process which concerns on embodiment of this invention. It is a flowchart which shows the example of the flow of a process of the attribute analysis part which concerns on embodiment of this invention. A description example of the conversion rule is shown. It is a figure which shows the example of a text element list. It is a figure for demonstrating the example of information for extracting the item name and label defined in the knowledge dictionary part. It is a flowchart which shows the example of the flow of an attribute provision process. It is a figure which shows the specific example of the provided attribute value and text ID. It is a flowchart which shows the example of the flow of a process of a conversion execution part. It is the figure which showed the specific example of the final output document structure with respect to the conversion object location. It is the figure which showed the other specific example of the final output document structure with respect to the conversion object location.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 ... Document structuring processing apparatus, 2 ... Conversion object location identification part, 3 ... Attribute analysis part, 4 ... Conversion mapping production | generation part, 5 ... Conversion execution part, 6 ... Knowledge dictionary part, 7 ... Knowledge dictionary part, 8 ... Conversion Rule analysis unit, 9 ... conversion rule unit, 10 ... morpheme analysis unit.

Agent Patent Attorney Susumu Ito

Claims (7)

A knowledge dictionary part that defines attribute assignment information for assigning attribute information to a predetermined character string;
A conversion rule part that defines output constraint information for converting the conversion target part of the input document and outputting the output document;
The presence or absence of the predetermined character string is determined for the character string that appears at the specified location to be converted, and the predetermined character string is supported based on the attribute assignment information of the knowledge dictionary unit An attribute assigning unit for assigning the attribute information;
From the attribute information provided by the attribute assigning unit, determine the regularity of appearance of predetermined attribute information defined in the output constraint information, and based on the determined regularity of appearance , A conversion map generation unit for generating a conversion map of the output constraint information;
A document processing apparatus, comprising: an output document generation unit configured to generate and output an output document in a format that matches the output constraint information with respect to the information of the conversion map.   The document processing according to claim 1, further comprising: a conversion target location specifying unit that characterizes the conversion target location of the input document, wherein the conversion target location is specified by the conversion target location specifying unit. apparatus. The conversion target portion includes a character string having a tabular structure,
The conversion map generation unit sets the number of repeated appearances of the predetermined attribute information as the number of repeated appearances of the cell contents in the table format, and according to the number of repeated appearances, the character string that appears in the conversion target location 3. The document processing apparatus according to claim 1, wherein the predetermined character string is determined to indicate the cell contents, and the conversion map is generated.   The document processing apparatus according to claim 3, wherein the output document generation unit does not output the output document including the cell content when the cell content is an empty element.   The attribute assignment unit includes a morpheme analysis unit that performs a morpheme analysis on a character string that appears in the conversion target portion, and assigns the attribute information using an analysis result of the morpheme analysis unit. The document processing apparatus according to any one of claims 1 to 4. Define attribute assignment information for assigning attribute information to a predetermined character string,
Define output constraint information to convert the conversion target part of the input document and output the output document,
The presence or absence of the predetermined character string is determined for the character element appearing in the specified location to be converted, and the attribute information corresponding to the predetermined character string is determined based on the attribute assignment information. Grant,
The regularity of appearance of predetermined attribute information defined in the output constraint information is determined from the assigned attribute information, and the conversion mapping of the output constraint information is based on the determined regularity Produces
A document processing method comprising: generating and outputting an output document in a format that matches the generated output mapping information with the output constraint information. A computer program for document conversion,
A function for defining attribute assignment information for assigning attribute information to a predetermined character string;
A function for defining output constraint information for performing conversion on the conversion target part of the input document and outputting the output document;
The presence or absence of the predetermined character string is determined for the character element appearing in the specified location to be converted, and the attribute information corresponding to the predetermined character string is determined based on the attribute assignment information. The functions to grant,
The regularity of appearance of predetermined attribute information defined in the output constraint information is determined from the assigned attribute information, and the conversion mapping of the output constraint information is based on the determined regularity With the ability to generate
A document conversion program that causes the computer to realize a function of generating and outputting an output document in a format that matches the output constraint information with respect to the generated conversion mapping information.

Images