JP2015069393A - Document data comparison method, document data comparison apparatus, and document data comparison program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a document data comparison method, a document data comparison apparatus, and a document data comparison program in which conversion into character codes and calculation of metric values are contrived to allow utilization of attribute information other than character codes.SOLUTION: A document data comparison apparatus 1 comprises: a character string/attribute extraction unit 2 which divides document data composed of sets of character string fragments into character string fragments and extracts attribute information representing the status of the character string fragments in the document; a metric calculation unit 3 which converts the characters contained in the character string fragments into the corresponding character codes to extract character code sequences, and calculates the similarity metric of the character string fragments; and a comparison unit 3 which searches the document data for sets of character string fragments having a ratio of similarity metric being within a numerical range A and matched attribute information, and which when the ratio of similarity metric is within a numerical range B, determines that the contents of the sets of character string fragments match and outputs the determination as a comparison result. This allows the contents of the document data to be compared in the form character codes.

Description

  Embodiments described herein relate generally to a document data comparison apparatus using a method for comparing contents of document data with character codes, and a document comparison program for executing the same on a computer.

  When comparing document data, there is a method for comparing the meanings of the documents, not the exact match of the characters constituting the documents. As a method of comparing the meanings of documents, there is a method of comparing using the similarity of a set of character string data.

  A technique for measuring the similarity of a set of character string data can be realized using a character code. This technology is aimed at field matching. A group of data having the same attribute included in tabular data is called a field.

  Field matching is an operation for associating fields in tabular data. For example, two tables have the same fields named “name” and “name”, respectively, and it is determined that they are the same field.

  In the procedure, first, the characters of each element constituting the field are converted into character codes and added for each element. Next, for each field, the average value and variance or deviation value of the added values are calculated. Further, the field is represented as a point on a mapping space of two or more dimensions composed of the variance or deviation value of the average value. Finally, the distance between the two points corresponding to the field to be compared is calculated and the value is used as the similarity. (For example, refer to Patent Document 1).

JP 2004-227037 A

  Conventional techniques for measuring similarity using character codes are intended for field matching and are not suitable for comparing document data. What is necessary for comparing the contents of text data is to search which part of each document is compared, and to compare the contents of the character strings linked in this search. That is, in the comparison of document data, only one character string having the same pattern as a certain character string is searched from a set of character strings having different patterns, and their contents are compared.

  On the other hand, field matching links a set of a plurality of character strings having the same pattern.

  As described above, since comparison of document data is for a single character string, less information is obtained from the character string than field matching. It has been a problem to extract more criteria for determining similarity from one character string.

  Embodiments of the present invention have been proposed to solve the above-described problems of the prior art, and are documents that devise conversion to character codes and calculation of metrics and use attribute information other than character codes. A data comparison method, a document data comparison device, and a document data comparison program are provided.

  In order to achieve the above object, the document data comparison method of the present embodiment is a document data comparison method by a document data comparison device that compares the contents of document data with character codes, and divides the document data into character string fragments. A character string / attribute extraction step for extracting attribute information indicating the position of the character string fragment in the document, and converting a character included in the character string fragment into a corresponding character code to extract a character code number sequence, A metric calculation step for calculating a similarity metric of the character string fragment, and a set of the character string fragments that have a ratio of the similarity metric within a certain numerical range A and match the attribute information, are the document data When the ratio of the similarity metric is within a certain numerical range B, it is determined that the contents of the set of character string fragments match, and the comparison result is output. Characterized in that it comprises a comparison step of the.

  A document data comparison apparatus using the document data comparison method and a document comparison program using the document data comparison apparatus are also one aspect of the present invention.

It is a block diagram which shows the structure of the document data comparison apparatus which concerns on 1st Embodiment. It is a block diagram which shows the structure of the character string and attribute extraction part of 1st Embodiment. It is a block diagram which shows the structure of the metric calculation part of 1st Embodiment. It is a block diagram which shows the structure of the comparison part of 1st Embodiment. It is a block diagram which shows the structure of the output part of 1st Embodiment. It is a figure which shows an example of the document data A and B which are compared in 1st Embodiment. It is a figure which shows the attribute of the character string fragment | piece in the document data A of 1st Embodiment. It is a figure which shows an example of A and B for comparison data of 1st Embodiment. It is a figure which shows an example which converted the character string fragment of 1st Embodiment into the character code number sequence. It is a figure which shows an example of A1 and B2 for comparison data of 1st Embodiment. It is a figure which shows an example of the comparison result list | wrist of 1st Embodiment. It is a figure which shows an example of the comparison result reflection document data of 1st Embodiment. It is a figure which shows the character code number sequence created with the number sequence creation method of other embodiments.

[First Embodiment]
Embodiments of a document data comparison apparatus according to the present invention will be described below in detail with reference to the drawings. In addition, the overlapping description is abbreviate | omitted suitably by attaching | subjecting the same code | symbol to the same part through each figure.

In the document data comparison apparatus 1 of this embodiment, the document data 100 is input to the document data comparison apparatus, and the document data A and the document data B in the document data 100 are compared. The document data comparison apparatus 1 compares the target document data A and document data B by the following steps (1) to (5).
(1) A character string / attribute extracting step of dividing the document data A and B into character string fragments and extracting attribute information indicating the position of each character string fragment in the document.
(2) A metric calculation step of extracting a character code number sequence from a character string fragment and extracting a similarity metric for each character string fragment.
(3) A set of character string fragments is detected from the document data A and the document data B based on the ratio of similarity metrics and attribute information, and the comparison is made based on whether the contents of the character string fragment sets match. A comparison process.
(4) An output step of outputting a comparison result list and comparison result reflecting document data reflecting the comparison result in the original document data based on the result of the comparison step.
(5) A display step in which, when a character string fragment on the comparison result list is selected, a corresponding character string fragment is retrieved from document data reflecting the comparison result and displayed.

[1-1. Constitution]
FIG. 1 is a block diagram illustrating a configuration of a document data comparison apparatus 1 according to the present embodiment. In this embodiment, in order to perform comparison in the steps (1) to (5), a character string / attribute extraction unit 2, a metric calculation unit 3, a comparison unit 4, an output unit 5, a display unit 6, a user interface 7 and Is provided. The document data comparison apparatus further includes a comparison data storage unit 101 that stores comparison data necessary for document comparison.

(Character string / attribute extraction unit 2)
The character string / attribute extraction unit 2 reads the document data A and document data B to be compared in the document data 100, converts them into comparison data, and outputs them. As shown in FIG. 2, the character string / attribute extraction unit 2 includes a document data reading unit 20, a document data division unit 21, an attribute information extraction unit 22, and a comparison data output unit 23.

  The document data reading unit 20 reads the document data A and document data B to be compared. Document data A and B are independent document data, each of which is a set of arbitrary character string fragments. In the present embodiment, one of the character strings constituting the document is a character string fragment, and the group of character string fragments constituting the document data A is a character string a1.

  The document data dividing unit 21 divides the document data A read by the document data reading unit 20 into a plurality of character string fragments, and sets the set as a character string a1. In this embodiment, the document data A is divided into delimited character string fragments by chapters, line breaks, and punctuation points, and the set is defined as a character string a1. Similarly, the document data B is divided into a plurality of character string fragments, and the set is defined as a character string b1.

  The attribute information extraction unit 22 extracts attribute information of character string fragments constituting the character strings a1 and b1. In the documents A and B, attribute information indicating attributes of character string fragments constituting the character strings a1 and b1 is stored in addition to the character strings a1 and b1. The attribute information includes, for example, types such as headers, footers, texts, tables, and heading information.

  The comparison data output unit 23 creates comparison data A and B for the character strings a1 and b1 from the attribute information of the character string fragments constituting the character strings a1 and b1, and uses the comparison data A and B for comparison. The data is output to the data storage unit 101.

(Metric calculation unit 3)
The metric calculation unit 3 reads the comparison data A and the comparison data B stored in the comparison data storage unit 101, and calculates a similarity metric as a comparison reference. As shown in FIG. 3, the metric calculation unit 3 includes a character code number sequence creation unit 30, a similarity metric calculation unit 31, and a comparison data output unit 23.

  The character code number sequence creating unit 30 converts the character strings a1 and b1 of the comparison data into character code number sequences a2 and b2 composed of character codes. The character code number sequence is generated by converting all characters included in the character string fragments constituting the character strings a1 and b1 into corresponding character codes and extracting the character code number sequence. The character code number sequence is a number sequence based on each character string fragment. Therefore, if the character string a1 is a set of five character string fragments, five character code number sequences a2 are created from the character string a1.

  The similarity metric calculation unit 31 calculates a similarity metric that serves as an evaluation criterion for evaluating the comparison data A and the comparison data B from the character code number sequences a2 and b2 created by the character code number sequence creation unit 30. This similarity metric is the total value of the character codes of each character code number sequence.

  The comparison data output unit 23 compares the character strings a1 and b1 based on the attribute information of the character string fragments constituting the character strings a1 and b1 and the similarity metric of the character string fragments constituting the character strings a1 and b1. Data A1 and B2 are created and output to the comparison data storage unit 101.

(Comparison part 4)
The comparison unit 4 refers to the comparison data storage unit 101 and compares the document data A and the document data B. For comparison in the comparison unit 4, the similarity metric of the comparison data A1 and B2 is used. As illustrated in FIG. 4, the comparison unit 4 includes a comparison target search unit 40, a content determination unit 41, and a comparison result output unit 42.

  The comparison target search unit 40 searches for a character string fragment in the character string b1 of the document data B to be compared with a character string fragment in the character string a1. There are two types of search criteria for comparison: criteria based on a similarity metric that evaluates the contents of character string fragments, and criteria based on attribute information that evaluates positioning and context in the document. A set of character string fragments that satisfy both of the two criteria is used as a comparison target.

  The criterion based on the similarity metric is that the ratio of the similarity metric corresponding to the two character string fragments to be compared is taken and the ratio is within the numerical range A. The numerical value range A is a value close to 1 represented by 1.0 ± X using an arbitrary real number X. X selects an appropriate value according to the document data 100. The criterion based on the attribute information is that the attribute information obtained by the attribute information extraction 22 of the character string / attribute extraction unit 2 matches between the two character string fragments.

  The content determination unit 41 determines whether the contents of the two character string fragments searched by the comparison target search unit 40 to be compared match. The matching criterion is determined based on whether the ratio of the similarity metric is within several ranges B as in the criterion based on the similarity metric. The numerical range B is independent of the numerical range A and is a value close to 1 represented by 1.0 ± Y using an arbitrary real number Y. Y selects an appropriate value according to the document data 100.

  The comparison result output unit outputs the comparison result to the comparison data 101. The comparison result indicates whether a comparison target is found for each character string fragment of the character string of the document data A or whether the contents match.

(Output unit 5)
The output unit 5 refers to the comparison data storage unit 101 and outputs the comparison result obtained by the comparison unit 4 as the comparison result list 102 and the comparison result reflecting document data 103 reflecting the comparison result in the original document data. . As shown in FIG. 5, the output unit 5 includes a comparison result list creation unit 50 and a comparison result reflected document data creation unit 51.

  The comparison result list creation unit 50 creates a list in which the attribute of each character string fragment and the comparison result are written. In addition, the comparison result reflecting document data creating unit 51 outputs comparison result reflecting document data 103 in which the comparison result is reflected in the document data 100. The comparison result reflecting document data 103 based on the document data A is created as comparison result reflecting document data 103A, and the comparison result reflecting document data based on the document data B is created as comparison result reflecting document data 103B.

(Display unit 6)
The display unit 6 displays the comparison result list 102 obtained by the output unit 5 and the comparison result reflecting document data 103 in conjunction with each other. When a character string fragment in the comparison result list 102 is selected, the corresponding character string fragment is highlighted on the comparison result reflecting document data 103 and displayed.

(User interface part 7)
The user interface unit 7 controls the character string / attribute extraction unit 2, the comparison unit 4, the output unit 5, and the display unit 6. It is assumed that the user selects a means to execute and sets input data and output data necessary to execute the means.

[1-2. Action]
Next, a document comparison method of the document data comparison apparatus 1 according to the present embodiment will be described with reference to FIGS. In the present embodiment, document data is compared through the following steps.

(String / attribute extraction process)
In the character string / attribute extraction step, the document data A and B are divided into character string fragments, and attribute information representing the position of each character string fragment in the document is extracted.

  First, the document data reading unit 20 reads the document data A and document data B to be compared. FIG. 6 is a diagram showing an example of document data A and document data B. In document data A, “Comparison of document data”, “Chapter 1”, “○○○ ●.” There is a character string fragment "". On the other hand, in the document data B, there are character string fragments of “comparison based on similarity metric”, “1. Introduction”, “◆ ◇ ◆ ◇”, “convert data by device A”, “◆ ◇ ◆.”. To do.

  After reading the document data A and the document data B, the document data A and B are divided into character string fragments. In FIG. 6, the document data A is divided into “About comparison of document data”, “Chapter 1”, “○○○ ●.”, “Data is converted by the device A”, and “● ○ ●.”. Let it be a string fragment. Similarly, the document data B is divided into “comparison based on similarity metric”, “1. Introduction”, “◆ ◇ ◆ ◇”, “convert data with device A”, “◆ ◇ ◆”, and character string fragments. And

  Thereafter, attribute information indicating the position of each character string fragment in the document is extracted. The attribute information of the character string fragment in the document data A and B is extracted. In FIG. 7, the attribute information “title” is included in the character string fragment “document data comparison” of the document data A, and the attribute information “heading” is included in the character string fragment “first chapter”. The attribute information “text” is stored in the character string fragment “data is converted by the device A”. This attribute information is extracted to create comparison data.

  FIG. 8 shows comparison data A and B created from document data A and B. FIG. For example, the comparison data A is created by associating the character string a1 and the attribute information of the character string fragments constituting the character string a1. Since the document data A is composed of 5 character strings, each character string fragment is assigned a number #. Therefore, in the comparison data A, the character string fragments constituting the character string a1 are # 1, # 2. , # 3, # 4, and # 5 are assigned.

(Metric calculation process)
In the metric calculation step, a character code number sequence is extracted from the character string fragment, and a similarity metric for each character string fragment is extracted.

  First, the character code number sequence creating unit 30 creates character code number sequences a2 and b2 composed of character codes from the character strings a1 and b1 of the comparison data. FIG. 9 shows a character code number sequence created from the character string cross-section constituting the character string a1. As shown in FIG. 9, the character string fragment a1 # 4 “Convert data with device A” is divided into “device”, “device”, “A”, “de”,. Then, the “pack” is converted into a shift JS code “9195”. Similarly, “place” is converted to “9275”, “A” is converted to “8260”, and “de” is converted to “82C5”. As a result, the character string fragment a1 # 4 “Convert data by apparatus A” is represented by the character code number sequence a2 # 4 “9195” “9275” “8260” “82C5”. In this manner, the character code number sequence a2 # 1 to # 5 is created from the character string fragments a1 # 1 to # 5 by creating the character code number sequence a2 for the other character string fragments constituting the character string a1. Similarly, numerical sequences b2 # 1 to # 5 are created from the character string fragments b1 # 1 to # 5.

  After creating the character code number sequences a2 and b2 from the character strings a1 and b1, the similarity metric for each character string fragment constituting the character strings a1 and b1 is extracted. The similarity metric is a total value of the character code number sequences a2 and b2. For example, the character string fragment a1 # 4 “convert data by device A” is “9195”, “9275”, “8260”, “82C5”.・ The total is "6DBA6". As described above, the similarity metric can be extracted by summing up the corresponding character code number sequence a2 for other character string fragments constituting the character string a1.

  FIG. 10 shows comparison data A1 and B2 created from document data A and B. For example, the comparison data A1 is created by associating the similarity information with the attribute information of the character string fragments a1 # 1 to # 5 and the character string fragments a1 # 1 to # 5.

(Comparison process)
In the comparison step, a set of character string fragments is detected from the document data A and the document data B based on the similarity metric ratio and the attribute information, and whether or not the contents of the character string fragment sets match. Make a comparison. In the detection of a pair of character string fragments, a similarity metric ratio is taken, and a combination in which the ratio is within the numerical range A and the attribute information is the same is detected. Since the detection is performed using the ratio of similarity metrics, it is possible to search for combinations of character string fragments whose contents in the character string fragment and the position in the document substantially match. For example, when a character string fragment to be compared with the character string fragment a1 # 4 of the document data A is detected from the character string fragment in the document data B, the character string fragment b1 # 4 has the same similarity metric, and the attribute Since the information is also the same, it is detected as a comparison target.

  Then, it is determined whether or not the contents of the combinations of character string fragments detected as comparison targets match. In the determination, it is determined that the contents of the combinations in which the ratio of the metric values is within the numerical range B are the same. As a result, the contents are not completely matched, and the contents can be compared including almost the same. For example, since the character string fragment a1 # 4 and the character string fragment a2 # 4 have the same similarity metric, it is determined that the contents match.

(Output process)
In the output step, the comparison result list 102 and the comparison result reflecting document data 103 reflecting the comparison result in the original document data are output based on the comparison result. FIG. 11 is a diagram illustrating an example of the comparison result list 102. The comparison result list 102 includes a list of file names, page numbers, and character strings that belong to character string fragments of each document, and their comparison results in one line, arranged in order from the beginning of the document. In the comparison result, “O” or “X” is used to indicate a match or mismatch. In FIG. 11, the character string fragment “Convert data by device A” described on page 1 of the file name “01.txt” in document data A is the file name “02.txt” in document data B. It can be seen that it matches the character string fragment “Convert data with device A” described on page 1 of “txt”.

(Display process)
In the display step, the comparison result reflecting document data 103 is displayed. The comparison result reflecting document data 103 is created based on the document data A or B to be compared. The comparison result reflecting document data 103 is linked to the comparison result list 102. When a character string fragment in the comparison result list 102 is selected, the corresponding character string fragment is highlighted on the comparison result reflecting document data 103. Is displayed. For example, FIG. 12 shows the comparison result reflecting document data 103 between the document data A and the document data B created based on the document data A. The comparison result reflecting document data 103 in FIG. 12 is linked to the comparison result list 102 in FIG. 11. When the character string fragment “Convert data with device A” is selected in the comparison result list 102 in FIG. The corresponding character string fragment “Convert data with device A” is highlighted on the comparison result reflecting document data 103 and displayed.

  In the display step, the character string fragment of the comparison result reflecting document data 103 can be color-coded based on the comparison result in advance. For example, a matched character string fragment is filled with cyan, and a mismatched character string fragment is filled with a yellow marker, and a fragment for which no comparison target is found is output without being painted.

[1-3. effect]
The document data comparison apparatus according to the present embodiment as described above can achieve the following effects.

(1) Document Content Comparison Using Character Code Number Sequences By comparing document data using a character code number sequence, it is possible to accurately determine content matches, not complete document matches.

  By converting characters into character codes, the comparison of character string fragments was changed to the comparison of character code sequences. Unlike the characters, the numbers constituting the character code can be combined into one numerical value by calculation. Therefore, the character string fragment can be expressed by a numerical value depending on all characters included in the character string fragment. The numerical value is the similarity metric used in this embodiment.

  The similarity metric other than the number of elements changes in value when even one character is replaced. The number of elements and the total value depend on the length of the character string fragment. The average value, variance, and deviation value represent the character type included in the character string fragment. This is because the numerical range of the character code is delimited by the character type. Therefore, the similarity metric represents the property of the entire character string fragment.

  The similarity metric does not depend on the order of the character string fragments. Since the word order has little influence on the meaning of the character string fragment, the similarity metric of character string fragments having similar meanings has the same value. In general, hiragana is used for particles and endings that connect words and has little effect on the meaning of a document. At the same time, hiragana also has a small character code, so the difference in hiragana contributes little to the similarity metric. Therefore, the comparison using the similarity metric can also determine the content difference.

  When the ratio of useful similarity metrics is taken in this way and this ratio is defined as similarity, a parameter approaching 1 is obtained when the contents of character string fragments are similar.

  In this way, by evaluating the contents of the document with the character code number sequence, the character string fragments between the documents are accurately linked by the comparison target search in the comparison unit 4, and the difference in meaning between the character strings is determined by the content match determination. Can be recognized accurately. Therefore, it is possible to accurately determine the content match rather than the complete match of the document data by the character code.

(2) Easy-to-understand comparison result by output / display function The output unit 5 and the display unit 6 can provide a user-friendly comparison result. First, since the comparison result list 102 has a large amount of information per screen, a large amount of comparison results can be referred to collectively, and the target comparison result can be confirmed earlier. Next, since the result reflection document data 103 can confirm the comparison result in the same format as the original document data, it is easy to visually understand the coincidence / non-coincidence part of the document. Finally, by linking the comparison result list 102 and the result reflection document data 103, the target character string fragment is found in the comparison result list 102 and the result of understanding the comparison result in the result reflection document data 103 is increased. Can do.

(3) Realization of comparison of a long document divided into a plurality of data Since the document data reading unit 20 of the character string / attribute extraction unit 2 can compare a plurality of divided documents, a long document divided into a plurality of data Can also be compared.

(4) Suppression of processing time In the user interface unit 7, by making it possible to select a means to execute and diverting past data, the execution time can be shortened. That is, the attribute information, the metric value, and the comparison result are not obtained every time the document data are compared, but the data at the time of the previous comparison may be stored in the storage unit and the comparison may be performed based on the data. .

(5) In the present embodiment, a specific example has been described. However, the following modifications can be made.

(a) In this embodiment, a set of arbitrary character string fragments is used as document data, and document data A and document data B are used as independent document data. The document data a1 and document data a2 may be composed of a plurality of documents instead of a single segment document. The format of the document data A and B is not particularly limited, but it is possible to use a document object in which each document data is saved as a data object on a computer storage device.

(b) Although the document data A and the document data B are one document data, they may be composed of a plurality of document data. That is, document data a1 in which arbitrary character string fragments are a set and document data a2 in which different character string fragments are a set may be set as one document data A. In this case, when comparing the document data A and the document data B, the document data a1 and the document data a2 are integrated and handled as one document data, so that all of the document data a1 and the document data a2 are all processed. The character string fragment can be compared with the character string fragment in the document data B.

(c) The document data A and B are delimited character string fragments with chapters, line breaks, and punctuation marks. The character string fragments are characters arranged in order, and the division method can be arbitrarily determined. For example, in the case of chapters, line breaks, punctuation marks, and tables, it can be separated by cells.

(d) The character string similarity metric is obtained by converting each character string fragment into a number sequence consisting of character codes and calculating the total value of each number sequence. However, the character string similarity metric can be obtained by a combination of character codes obtained by converting character string fragments. . As an example, there are the number of elements, average value, variance value, and deviation value of the number sequence constituting the character string fragment.

(e) The output unit 5 outputs the comparison result as the comparison result list 102 or the comparison result reflecting document data 103. However, the output unit 5 does not necessarily output these two formats, but the user can select the output format. Also good.

(f) In the character code sequence, all characters of the character string fragment are converted into character codes. Instead of converting all characters into character codes, the following method can be used.

  If only the character type that affects the contents of the document is converted into a number sequence as in the sequence creating method 71 in FIG. 13, the dependence of the document content on the similarity metric increases, and the accuracy of the evaluation is improved. For example, “the data is converted by the device A”, “the data is converted by the device A”, and “the data is converted by the device A” are synonymous with different expressions. For these character string fragments, if character types other than hiragana and symbols are digitized and compared, all of these character string fragments are converted to a numerical sequence consisting of “device A data conversion”, and content matching of the character string fragments can be easily performed. I can judge.

  In the numerical sequence creation method 72, weights having different sizes can be added depending on the character type, so that the difference between the character types having a large influence on the document content can be emphasized. For example, in general, katakana and numbers often contribute to the meaning of a document, while the character code is small and the dependence on the similarity metric is small. Therefore, if a large weight is set for Katakana and numbers, the similarity can be obtained with higher accuracy.

  In the numerical sequence creation method 73, since the characters are separated for each consecutive character of the same character type, the numerical sequence can be extracted in units of words. Therefore, the difference in words is more remarkably reflected in the similarity metric, and the contents of the character string fragments can be accurately compared.

  The degree of similarity of character string fragments can be more accurately evaluated by devising how to create a number sequence as in the number sequence creation methods 71-73.

(g) In the comparison target search unit 40, a character string fragment in the character string a1 is compared with a character string fragment in the character string b1 of the document data B to be compared. At this time, not only the character string fragments are directly compared but also the similarity metric of the character string fragments before or after the fragments can be used for the comparison target search. As a result, when the number of characters included in the fragment is small, it is possible to avoid making an erroneous determination only with the similarity metric of the fragment, and to improve the accuracy of the comparison target search.

(h) In addition, when searching for a comparison target in the comparison target search unit 40, not only the character string fragments are directly compared with each other, but the fragment and the preceding or following character string break are combined, and the combinations are similar. It can also be used to search for metric comparison targets. For example, when comparing document data A and document data B that originally consist of the same character string fragment “AABBCC”, the character string fragment of document data A is “AABB” and the character of document data B There is a possibility that the column fragment is divided into “AA” and “BB”. In this case, by combining the character string fragments “AA” and “BB” of the document data B and treating them as one character string fragment “AABB”, it is possible to compare the character string fragment of the document data A with “AABB”. It becomes. Thereby, it is possible to avoid making an erroneous determination due to the division of the character string, and to improve the accuracy of the comparison target search.

(i) Although the similarity metric calculation unit 31 calculates one similarity metric from one character string fragment, it can also calculate a plurality of similarity metrics from one character string fragment. For example, in a certain character string fragment, the total value of the character code number sequence is the similarity metric X, the variance value of the character code number sequence is the similarity metric Y, and the deviation value of the character code number sequence is the similarity metric Z. In this way, three similarity metrics X, Y, Z can be calculated from one character string fragment. Then, when extracting a set of character string fragments or comparing extracted character string fragments, a ratio of the respective similarity metrics is obtained, and an average value of the ratios is used. That is, in the comparison between the character string fragment a1 # 3 and the character string fragment b1 # 3, the ratio of the similarity metric X of each character string fragment, the ratio of the similarity metric Y, and the ratio of the similarity metric Z are obtained. The average value of the ratios of the three similarity metrics is calculated and used.

[2. Other Embodiments]
In the present specification, a plurality of embodiments according to the present invention have been described. However, these embodiments are presented as examples and are not intended to limit the scope of the invention. Specifically, the first embodiment can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the scope of the invention. These embodiments and modifications thereof are included in the invention described in the claims and equivalents thereof as well as included in the scope and gist of the invention.

  Moreover, although the specific example was given and demonstrated about 1st Embodiment, each condition etc. are not limited to the numerical value described. Moreover, even if it changes within the range described in the problem-solving means mentioned above, the same effect is acquired.

1: Document data comparison device 2: Character string / attribute extraction unit 20: Document data reading unit 21: Document data division unit 22: Attribute information extraction unit 23: Comparison data output unit 3: Metric calculation unit 30: Character code sequence Creation unit 31: Similarity metric calculation unit 32: Comparison data output unit 4: Comparison unit 40: Comparison target detection unit 41: Content determination unit 42: Comparison result output unit 5: Output unit 51: Comparison result list creation unit 52 : Comparison reflection result creation unit 6: Display unit 7: User interface 100: Document data,
100A: Comparison result reflected document data 100B: Comparison result reflected document data 101: Comparison data storage unit 102: Comparison result list 103: Comparison result reflected document data

Claims (9)

A document data comparison method by a document data comparison device for comparing the contents of document data with character codes,
A character string / attribute extraction step for dividing the document data into character string fragments and extracting attribute information representing the position of the character string fragments in the document;
A metric calculation step of converting a character included in the character string fragment into a corresponding character code, extracting a character code number sequence, and calculating a similarity metric of the character string fragment;
A search is made for the set of character string fragments in which the ratio of the similarity metric is within a certain numerical range A and the attribute information matches, and the range of the similarity metric is within a numerical range B having the ratio of the similarity metric. A comparison step of determining that the contents of the set of character string fragments match and outputting as a comparison result;
A document data comparison method comprising: The metric calculation step includes
2. The document data comparison method according to claim 1, wherein a character code number sequence obtained by converting only a specific character into a character code among characters included in the character string fragment is extracted. The metric calculation step includes
3. The document data comparison method according to claim 1, wherein a character code number sequence obtained by multiplying the converted character code by a weight is extracted. The metric calculation step includes
4. The character code number sequence having a total value of character codes of characters belonging to the same character type among the characters included in the character string fragment as an element is extracted. 5. The document data comparison method described. The character string / attribute extraction step includes:
5. The document data comparison method according to claim 1, wherein the attribute data is extracted by integrating the document data divided into a plurality of parts and extracted. The document data comparison device includes a storage unit that stores attribute information for each character string fragment, the similarity metric for each character string fragment, and the comparison result.
The comparison step refers to any one of the attribute information, the similarity metric, and the comparison result stored in the storage unit when comparing document data. Document data comparison method described in 1. A comparison result list in which the comparison result, the character string fragment and the attribute information are paralleled, and an output step of outputting comparison result reflecting document data reflecting the comparison result in the original document data;
When the character string fragment on the comparison result list is selected, a character string fragment corresponding to the selected character string fragment is searched from the comparison result reflecting document data and displayed.
The document data comparison method according to claim 7, further comprising: A document data comparison device for comparing the contents of document data with character codes,
A character string / attribute extraction unit that divides document data into character string fragments and extracts attribute information indicating the position of the character string fragments in the document;
A metric calculation unit for converting a character included in the character string fragment into a corresponding character code, extracting a character code number sequence, and calculating a similarity metric of the character string fragment;
A search is made for the set of character string fragments in which the ratio of the similarity metric is within a certain numerical range A and the attribute information matches, and the range of the similarity metric is within a numerical range B having the ratio of the similarity metric. A comparison unit that determines that the contents of the set of character string fragments match, and outputs a comparison result;
A document data comparison device comprising: A document data comparison program by a document data comparison device for comparing the contents of document data with character codes,
The program is stored in a computer
A character string / attribute extraction step for dividing the document data into character string fragments and extracting attribute information representing the position of the character string fragments in the document;
A metric calculation step of converting a character included in the character string fragment into a corresponding character code, extracting a character code number sequence, and calculating a similarity metric of the character string fragment;
A search is made for the set of character string fragments in which the ratio of the similarity metric is within a certain numerical range A and the attribute information matches, and the range of the similarity metric is within a numerical range B having the ratio of the similarity metric. A comparison step of determining that the contents of the set of character string fragments match and outputting as a comparison result;
A document data comparison program characterized in that