JP2013152662A - Table classification device, table classification method, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To classify tables included in a document on the basis of similarity in table constitution information.SOLUTION: As a table classification device, an information processing device comprises: a table constitution information extraction unit that extracts information constituting each individual table from a document as table constitution information; and a table classification unit that calculates an inter-table-element distance between respective elements in the individual table, discriminates a manner of continuation between table elements with a short distance, and classifies the individual table on the basis of the manner of continuation.

Description

  The present invention relates to a technique for classifying a table included in a document according to a table configuration pattern.

When a document is analyzed by a computer, how to handle a table included in the document is an issue. The table is generally composed of a row / column as a heading and a cell for storing contents along the heading. In order to correctly analyze the table, it is necessary to make the computer recognize the structure of the table as correctly as possible.
In order to process a large number of tables included in a document at once, it may be desirable to classify the tables according to the closeness of the table structures. For example, when converting a table into a natural sentence, it is desirable to change the processing depending on whether the first column of the table is a heading, the first row is a heading, or both are headings. In the example of the table (T10) in FIG. 10 described below, the first line is assigned to the elements on the second and subsequent lines, such as “Time is 1; subject is national language; content is haiku; teacher is Kozuka.” However, if the first column is assigned to the elements after the second column, such as “Time is subject, 1 is national language,...”, The sentence is not established. However, the reverse conversion should be applied to the table whose heading is the first column. When processing must be changed depending on the structure of the table, it is important to correctly classify the table according to the closeness of the structure.

Patent Documents 1 and 2 are conventional techniques for causing a computer to recognize the configuration of a table.
Patent Document 1 determines an attribute name area that is a heading portion of a table and its direction (whether it is a row direction or a column direction), and generates table extraction information in which attribute names and attribute values are paired. The method recognizes specific expressions and domain terms using information for recognizing specific expressions such as names, time expressions, and monetary expressions such as names of people and places, and a dictionary that defines the terms of each field in a tree structure. , Tagging. Then, the table structure element (cell contents) at the beginning of the row / column is compared with other table structure elements to determine which is the higher level concept based on the tag, and thereby the attribute name area and its direction Determine.
Patent Document 2 is a technique for converting table information into a format that can be displayed on a terminal when the table information is displayed on a terminal that cannot be displayed in a table format. To recognize the structure of the table, display characteristics of the table items such as the background color of the table, the color / thickness of the table, the number of columns in the table, the number of rows, the same values are arranged in the same row (column) The table is classified according to characteristics, and the table is converted according to certain rules.

JP 2001-325284 A JP 2001-331406 A

  However, the technique described in Patent Document 1 has a problem in that the table structure cannot be specified unless specific recognition expressions and field terms are defined in advance. In addition, in the case of a special table having a structure in which the attribute name area extends to a portion other than the first row (or one column), the table structure cannot be specified.

  Moreover, although the technique described in the said patent document 2 can classify | categorize a table along the kind of table defined beforehand, a table which is not defined cannot be classified. Therefore, in order to correctly classify a table, it is necessary to define all the table patterns to be classified.

  As described above, the techniques described in Patent Document 1 and Patent Document 2 have a problem that the tables in the document cannot be classified as necessary.

  An object of the present invention is to provide a table classification device, a table classification method, and a program for automatically classifying a table included in a document according to a table configuration pattern.

  Another object of the present invention is to provide a table classification device, a table classification method, and a program for automatically and efficiently extracting table structure information from a table included in a document.

  The table classification device according to the present invention calculates a table configuration information extraction unit that extracts information constituting each table from the document as table configuration information, and calculates a distance between table elements between each element of each table, It is characterized by comprising at least a table classifying unit that identifies a method of continuation of table elements having a short distance and classifies each table type based on the method of continuation.

  In addition, the table classification device determines the table structure information indicating the data structure of the table in which the attribute and attribute value of one table in each classification are paired according to the extracted table configuration feature or the designation from the user. And a table structure information determining unit that converts the remaining tables in the same classification into data as a table structure based on the same rule.

  According to the present invention, it is possible to provide a table classification device, a table classification method, and a program capable of automatically classifying a table included in a document according to a table configuration pattern without requiring preparation in advance.

  Another object of the present invention is to provide a table classification device, a table classification method, and a program for automatically and efficiently extracting table structure information from a table included in a document.

It is a block diagram which shows the structure of embodiment of 1st invention. It is a block diagram which shows the structure of embodiment of 2nd invention. It is a flowchart which shows operation | movement of embodiment of 1st invention. It is a flowchart which shows operation | movement of embodiment of 2nd invention. It is a figure which shows the process of the distance calculation of a morpheme structure. It is a figure which shows the example of the table | surface to classify | categorize. It is a figure which shows the number of characters of the table element of the table | surface of FIG. It is a figure which shows the morpheme structure of the table element of the table | surface of FIG. It is a figure which shows the character classification structure of the table element of the table | surface of FIG. It is explanatory drawing which shows the calculation process which performs distance calculation of the table element of the table | surface of FIG. 6 in a column direction. It is explanatory drawing which shows the calculation process which performs distance calculation of the table element of the table | surface of FIG. 6 in a line direction. It is a figure which shows the example of table structure information. It is a figure which shows the example of table structure information.

  Next, embodiments for carrying out the invention will be described in detail with reference to the drawings.

  Referring to FIG. 1, a table classification device 10 according to the first exemplary embodiment of the present invention includes at least a table configuration information extraction unit 101, a storage unit 110, a table classification unit 102, and a table structure information determination unit 103. It has. The storage unit 110 has at least table configuration information 111 and table configuration information 112.

  Further, as the second embodiment of the present invention, in addition to the configuration of the first embodiment, as shown in FIG. 2, a table header specifying unit 104 and a table reclassifying unit 105 may be provided. In that case, the storage unit 110 includes a headline specifying rule 113 and a table classification rule 114 in addition to the configuration of the first embodiment.

The table configuration information extraction unit 101 extracts information constituting each table from the document. That is, a table portion is read from a document described in a specific format, information indicating what kind of character string is arranged in what row and column of the table is acquired, and the acquired information is converted into table configuration information 111. As a means for storing in the storage unit 110.
The table configuration information includes a row number and a column number indicating positions in the table, connection information indicating whether or not cells are connected, and table elements indicating character strings in the cells.
The table structure information can be acquired by reading the contents of the <TABLE> tag from a document written in Hyper Text Markup Language (HTML) or Extensible Markup Language (XML) and converting it to table structure information. Alternatively, a method may be used in which a character string is read from a paper medium with a scanner and the ruled line portion is regarded as a table. Alternatively, a method may be used in which the contents of a ruled line portion are regarded as a table by reading the contents from a document recorded in an image format by image recognition processing.
Further, as the table configuration information, information read from an HTML tag or the like, such as a table background color, a character color, a character thickness, and a ruled line type may be stored.

  The table classification unit 102 uses the number of characters of the table elements, the configuration of the morphemes, and the configuration of the character types for each table collected in the storage unit 110 as the characteristics of the table elements, and the differences between cells (between each element) (Calculated value difference) is calculated as the distance between table elements (distance between table elements), and it is identified how the cells with close distances are continuous (this feature is referred to as the table element feature hereinafter) A means for classifying individual tables according to table element features. By referring to the table element features that show differences between adjacent cells, it is possible to grasp the overall connection between table elements and identify which table is the same as which table. Is possible.

The table structure information determination unit 103 is a means for determining the table structure for each table classification classified by the table classification unit 102 or the table reclassification unit 105 described later. The determined table structure is converted into data as table structure information 112 and recorded in the storage unit 110. Further, the tables having the same classification are converted into data based on the same rules as the table structure determined as the representative of the classification.
The table structure information is information that associates table headings with table elements. By determining the table structure, it is possible to convert the table into a document, to give meaning to the table element, and to use the contents of the table for various information processing.
The method for determining the table structure may be automatically determined using the table element feature calculated by the table classification unit 102, or the table heading and the table element are determined by looking at the representative of the table into which the user has been classified. The association rules may be set, or both of them may be used together.

The table heading specifying unit 104 is means for specifying which row / column the heading part of the table is in accordance with a preset heading specifying rule 113.
The table heading specifying unit 104 uses a row / column whose table element attributes such as background color, character color, ruled line type, character thickness, etc. are different from other rows / columns as headings. As the headline specifying rule 113, it is set what value the attribute of the table element takes is determined as a headline.

The table reclassification unit 105 is a means for classifying the table classified by the table classification unit 102 more finely according to a preset table classification rule 114, or collecting the classified items into a new classification.
The table classification rule 114 is a rule for dividing the classification with reference to the table configuration information, for example, dividing the classification according to the number of rows or the number of columns, or dividing the classification according to how the cells are connected. Alternatively, it is possible to set a table classification rule for reclassifying the table according to the number of rows and columns of the headline specified by the headline specifying unit 104.
After the table structure information is converted into data, the table structure information converted into data is searched under the conditions specified by the user to extract a table that satisfies the required conditions, and is reproduced based on the specified table classification rule. Classification may be performed.

  The storage unit 110 is means for storing various types of information necessary for table classification, and includes table configuration information 111, table structure information 112, a headline specifying rule 113, and a table classification rule 114.

Next, the operation of the table classification apparatus 10 will be described with reference to the flowcharts of FIGS.
First, with reference to FIG. 3, the operation | movement which classify | categorizes a table using the table classification | category apparatus 10 of 1st Embodiment (structure of FIG. 1) is demonstrated. In this operation explanation, after classifying a table from one document file, the table structure of one table is extracted, and a table similar to the table structure is identified by applying the same table structure to collective identification. Increase processing efficiency.

First, the table configuration information extraction unit 101 extracts table configuration information of individual tables from the document (step S100).
Next, the table classification | category part 102 calculates the distance of the table elements of the same table (step S101). A method for calculating the distance will be described later.
Further, the table classification unit 102 specifies how the table elements are continuous in the table (table element characteristics) from the distance information between the table elements (step S102). The method of determining the continuous direction of the table elements is determined by whether or not the distance between the table elements is equal to or less than a certain threshold value. Alternatively, it may be determined that the sum of the distances in the row and column directions divided by the number of comparisons of the table elements is continuous in the smaller direction. The latter method of determination is suitable for classification of a table having items in only one of the rows or columns. In addition, it is possible to deal with a table in which the row and column items of the table are near the center. In addition, if a diagonal direction is added to the distance to be compared, a table having items on a diagonal line can be handled.
When the table element features of all the tables included in the document have been specified, the tables are classified based on the table element features (step S103). The table classification is broadly divided into three patterns: whether the table elements are continuous in the row direction, continuous in the column direction, or continuous in both rows and columns. Furthermore, special patterns such as a table in which table elements are continued every other row in the row direction, or a table in which diagonal lines are drawn diagonally may be classified as different classifications. This classification number is arbitrary. You may prescribe | regulate the maximum number of classification | category numbers, and may define the distance of the agreement degree between the table groups to classify | categorize.
Moreover, the table classification | category part 102 may calculate the similarity degree of a table element characteristic within the table group classified by the same classification, and may provide a coincidence to a user. Using this degree of coincidence, it is possible to ask the user for confirmation of a table that will have the same structure but a low degree of coincidence. The table for which confirmation is requested is a table in which the degree of coincidence deviates from a predetermined threshold, or a predetermined number from the lower degree of coincidence is displayed on the screen so that the user decides whether or not to make the same classification You can do it.

  Finally, the table structure information determination unit 103 determines the table structure of one table (representative) from each classification according to table element characteristics or designation from the user, and the remaining tables of the same classification are also based on the same rule. The table structure is determined (step S104). This process is developed for all tables in the document, and the table structures of all tables are extracted and recorded. As a representative table in each classification, a table in which the distance between table elements in a specific direction is the smallest or a table close to the average can be used.

Next, the operation of the table classification apparatus 10 in the case of the second embodiment (configuration of FIG. 2) will be described with reference to the flowchart of FIG. The added processing steps will be described, and description of other processing steps will be omitted.
In the operation of the second embodiment, the operations of step S105 and step S106 are added to the flowchart of FIG.

  After step S100, the table header specifying unit 104 specifies a table header from the table configuration information (step S105). The method of specifying the table header is to determine a row / column whose table configuration information feature is different from other rows and columns as a table header. For example, when the background color is different only in the first line, the first line is determined as a table heading.

  After the table classification unit 102 classifies the table in step S103, the table reclassification unit 105 reclassifies the table based on the table configuration information 111 and the table classification rule 114 (step S106). Even if tables with the same table element characteristics are used, it is useful to divide the table classification according to the number of columns and rows of the table, the number of columns and rows of the heading part, etc. In this case, by describing whether to reclassify the table, it becomes possible to classify the table further finely or to summarize the once classified table. It is also possible to classify the table more finely and to summarize the table afterwards. The table classification rule 114 can be arbitrarily set by the user, and can be used for required classification along the document.

Here, the calculation method of the distance of the table element in step S101 will be described in detail. The distance between table elements (distance between table elements) is composed of the distance of the number of characters, the morpheme component distance, and the character type component distance. First, the distance of the number of characters is obtained by normalizing the difference in the number of characters in the table element and dividing it by the longer number of characters. When the distance between the table element e ₁ and the table element e ₂ is D _len (e ₁ , e ₂ ), the number of characters of the table element e ₁ is l ₁ , the number of characters of the table element e ₂ is l ₂ , It is as the number 1.

  For the morpheme composition distance, the morpheme analysis is performed on the table elements, and the edit distance of the part of speech composition is calculated. For example, when the character strings of the table elements are “media signal processing trend” and “speech / audio coding”, as shown in FIG. 5, “media signal processing trend” is a noun, noun, noun, It becomes a particle / noun, and “speech / audio coding” becomes a noun / symbol / noun / noun / noun. If each part of speech is regarded as a character, both can be regarded as a character string of five characters, and the edit distance can be obtained. The editing distance between the two character strings is 2, and if the value is normalized by dividing by the longer number of characters, it becomes 2/5.

  The constituent distance of the character type can also be calculated by the same method as the constituent distance of the morpheme. For example, in the case of “4/23” and “speech / audio encoding”, if the character type is roughly divided into Japanese, numbers, symbols, and letters, “4/23” is composed of numbers, symbols, and numbers. “Voice / Audio Coding” is composed of Japanese, symbols, and Japanese. If the edit distance is calculated with the same capacity as the morpheme composition distance, it becomes 2, and if it is normalized by dividing by the longer number of characters, it becomes 2/3. In addition, the classification method of the character type is not limited to Japanese, numbers, symbols, and English characters, but may be further divided into “Hiragana, Katakana, Kanji,…”, and symbols are included in Japanese. A rough classification may be made. The way of classification is set in advance.

Table element e ₁ and table element e ₂ of the distance D (e _{1, e 2),} the distance of the characters, the distance morphological structure, it is assumed that each of the distances of character type configuration sum over weight. When the distance of the morpheme configuration is expressed as D _mor (e ₁ , e ₂ ), the distance of the character type configuration is expressed as D _char (e ₁ , e ₂ ), and the weight is expressed as w ₁ , w ₂ , w ₃ , the distance between the table elements is It can be expressed by the following formula 2.

  The weight for each distance may be set in advance by the user. The weights may be automatically distributed based on the name and attributes of the read document.

  Next, the operation of the embodiment of the present invention will be described using specific examples. Note that the present invention is not limited to the following operation example.

A calculation example of the table classification unit 102 will be described with reference to FIGS.
In order to obtain the table element characteristics of the table as shown in FIG. 6 after extracting the table structure information, first, the number of characters, the morpheme structure, and the character type structure of each table element are combined and examined. The number of characters of each table element is as shown in FIG. 7, the morpheme configuration is as shown in FIG. 8, and the character type is as shown in FIG. In the morpheme configuration shown in FIG. 8, “name” is an abbreviation for “noun”, and “help” is an abbreviation for “particle”. In FIG. 9, the character type “day” is an abbreviation of “Japanese” and “number” is an abbreviation of “number”.
Then, the distance of the number of characters of these table elements to be combined, the distance of the morpheme configuration, and the distance of the character type configuration are calculated.
In calculating each distance, table elements are often similar in either row direction or column direction, so first calculate the distance between adjacent table elements in the row direction and column direction. By comparing which direction is closer, it is specified whether the arrangement of the table elements is row direction or column direction. As described above, the distance to the adjacent cell may be calculated in other directions.
When the table header can be specified by the table header specifying unit 104, the distance is calculated excluding the table header. If the heading cannot be specified, the heading is often in the first row or the first column. Therefore, emphasizing the distance between the table elements in the second and subsequent rows and the second and subsequent columns. You may determine the direction to arrange.

FIG. 10 shows the result of calculating the distance between the table elements in the column direction. T10 is an original table, and T11 indicates a pair of table elements to be compared (a pair for obtaining a distance).
The calculation results of the tables shown in T12 to 15 correspond to the portions corresponding to the same positions in the table of T11. T12 is the calculation result of the distance of the number of characters, T13 is the calculation result of the distance of the morpheme configuration, and T14 is the calculation result of the distance of the character type configuration. T15 is the total of the calculation results of T12 to T14.
The morphological calibration distance and the edit distance of the distance of the character type configuration are calculated by assuming that the replacement cost is 2. The weights for each distance are all calculated as 1.
Looking at the calculation result shown in T15, it is possible to identify that 0 is arranged in the A column, the B column, and the D column in the table of T10, and that the similar calculation is described in the above calculation. Can be determined.

Next, similarly to the calculation in the column direction, FIG. 11 shows the result of calculating the distance between the table elements in the row direction.
Looking at the total value of the calculation results shown in T25, 0 is arranged in the first row, but in the other rows, unlike the calculation result in the column direction shown in T15 of FIG. A large percentage appears.

If the threshold value for allowing continuity is set to 1, it can be determined that there is no continuity in the row direction. Therefore, the table classification unit 102 determines that this table is continuous in the column direction based on the calculation result. Can be classified as a table. In other words, it can be identified that the first column in the row direction is an item.
Here, the threshold value may be set to a low value (a value with a high degree of coincidence) in advance, and classification may be performed while sequentially rounding up to a predetermined number.

Next, an example of the table structure information will be described.
FIG. 12 shows the table structure information obtained by determining the table structure information in the table structure information determination unit 103 for the table of FIG. The table structure information is a set of attributes and attribute values. The table of FIG. 12 describes the table elements as they are, but the reference information of the table elements may be stored as table structure information as shown in FIG.
When the table classification unit 102 and the table header specifying unit 104 find out how the headings and the table elements are continuous, the table structure information determination unit 103 sets the table header as an attribute and sets the row or column belonging to the table header as an attribute value. The table structure information is saved. For tables for which table headings were not found, or for tables whose table element correspondence was complicated and the correspondence with table headings was not known, the user specifies the table elements that are the attributes and the table elements that are the attribute values. By doing so, the table structure information is determined. Then, the table structure is automatically created for the table having the same classification as that table according to the same rule.

  As described in the above embodiment, according to the table classification apparatus of the present invention, it is possible to automatically classify tables included in a document according to a table configuration pattern without requiring preparation in advance. It becomes. Further, the table structure information can be determined based on the same table structure determination rule by summarizing the table structures of the tables regarded as the same after classification. By processing in this way, once the table structure of the table representing each category is determined, the remaining individual table structures handled with the same category can be converted in a batch.

Each part of the table classification device may be realized using a combination of hardware and software. In a form in which hardware and software are combined, a table classification program is developed in the RAM, and each unit is realized as various means by operating hardware such as a control unit (CPU) based on the program. The program may be recorded on a storage medium and distributed. The program recorded on the recording medium is read into a memory via a wired, wireless, or recording medium itself, and operates a control unit or the like. Examples of the recording medium include an optical disk, a magnetic disk, a semiconductor memory device, and a hard disk.
If the above embodiment is described in another expression, an information processing apparatus that operates as a table classification apparatus is based on a table classification program developed in a RAM, a table classification unit, a table configuration information extraction unit, a table classification unit, It can be realized by operating the control unit as the table structure information determining unit, the table header specifying unit, the table reclassifying unit, and the storage unit.

  Although the embodiment has been illustrated and described, changes such as separation and merging of block configurations and replacement of procedures are free as long as the purpose and the function described are satisfied.

  In addition, the specific configuration of the present invention is not limited to the above-described embodiment, and changes within a range not departing from the gist of the present invention are included in the present invention.

  In addition, a part or all of the above-described embodiments can be described as follows. Note that the following supplementary notes do not limit the present invention.

[Appendix 1]
A table configuration information extraction unit that extracts information constituting individual tables from the document as table configuration information,
A table classification unit that calculates a distance between table elements between each element of each table, identifies a method of continuation of the table elements having a close distance, and classifies the individual tables based on the method of continuation;
A table classification apparatus comprising at least:

[Appendix 2]
Based on the extracted table element features or user designation, table structure information indicating the data structure of the table that is a set of attributes and attribute values of one table in each classification is determined and converted into data, and the same classification and A table structure information determination unit that converts the remaining table into data as a table structure based on the same rule,
The table classification device according to the above supplementary note, further comprising:

[Appendix 3]
The table classification unit uses a distance of the number of characters of a table element, an edit distance of a morpheme configuration, an edit distance of a character type configuration, or a combination thereof for calculating a distance between table elements. The table classification device described in the above supplementary note.

[Appendix 4]
The table classification unit determines, as a table element feature calculation rule, a method of determining that a table element is continuous in a direction in which the ratio of the value of the table element distance is equal to or less than a certain threshold, and the table element distance in the row direction. A table classification apparatus according to the above-mentioned supplementary note, characterized in that one of the methods of comparing the sum of the column direction and the total of the column directions divided by the number of comparisons of the table elements is used to determine the direction in which the table elements are continuous .

[Appendix 5]
The table classification device according to the above supplementary note, further comprising a table heading specifying unit for specifying a heading of the table from the characteristics of the table structure and a preset heading specifying rule.

[Appendix 6]
The table classification device as described in the above supplementary note, further comprising a table reclassification unit that further classifies the table classified by the table classification unit based on the table configuration information and a preset table classification rule.

[Appendix 7]
The said table structure information determination part uses the table | surface with the shortest distance of the table elements of a specific direction, or a table | surface close | similar to an average about the table | surface in each classification | category used as a representation, The table | surface classification apparatus of the said remarks characterized by the above-mentioned.

[Appendix 8]
The table classification unit calculates similarity of table element features within the same group of tables and calculates the degree of coincidence of individual tables within the table group, and a predetermined threshold value within the table group. The table classification apparatus according to the above supplementary note, wherein the user is asked to confirm whether or not a table out of the range is included in the same classification.

[Appendix 9]
Table configuration information extraction processing for extracting information constituting individual tables from the document as table configuration information,
A table classification process for calculating a distance between table elements of each element of each table, identifying a method of continuation of the table elements having a close distance, and classifying the individual tables based on the method of continuation;
Table classification method characterized by having.

[Appendix 10]
Based on the extracted table element features or user designation, table structure information indicating the data structure of the table that is a set of attributes and attribute values of one table in each classification is determined and converted into data, and the same classification and Table structure information determination process that converts the remaining tables into data as a table structure based on the same rules,
The table classification method according to the above supplementary note, further comprising:

[Appendix 11]
The table classification process uses one of a distance of the number of characters of a table element, an edit distance of a morpheme configuration, an edit distance of a character type configuration, or a combination thereof for calculating a distance between table elements. The table classification method described in the above supplementary notes.

[Appendix 12]
In the table classification process, as a table element feature calculation rule, a method for determining that a table element is continuous in a direction in which the ratio of the value of the table element distance is equal to or less than a certain threshold is large, and the distance between the table elements in the row direction A table classification method according to the above-mentioned supplementary note, wherein one of the methods for determining the direction in which the table elements are continuous is used by comparing the sum of the column direction and the total number of column elements divided by the number of comparisons of the table elements .

[Appendix 13]
The table classification method as described in the above supplementary note, further comprising a table heading specifying process for specifying a table heading from a table structure feature and a preset heading specifying rule.

[Appendix 14]
The table classification method as described in the above supplementary note, further comprising a table reclassification process for further classifying the table classified by the table classification unit based on the table configuration information and a preset table classification rule.

[Appendix 15]
The table classification method according to the above supplementary note, wherein the table structure information determination processing uses a table in which a distance between table elements in a specific direction is the smallest or a table close to an average for the table in each classification as a representative.

[Appendix 16]
The table classification process calculates the degree of similarity of individual tables within the table group by calculating the similarity of the table element features within the same group of tables, and a predetermined threshold value within the table group. The table classification method according to the above supplementary note, characterized in that the user is asked to confirm whether or not a table that falls outside the scope is included in the same classification.

[Appendix 17]
Information processing device
Table configuration information extraction means for extracting information constituting individual tables from the document as table configuration information,
Calculate the distance between the table elements of each table element, identify how the table elements that are close to each other are continuous, and operate as a table classification unit that classifies the individual tables based on the sequence. A table classification program characterized by that.

[Appendix 18]
Information processing device
Based on the extracted table element features or user designation, table structure information indicating the data structure of the table that is a set of attributes and attribute values of one table in each classification is determined and converted into data, and the same classification and The table classification program as described in the above supplementary note, which is further operated as table structure information determining means for converting the remaining table into a table structure based on the same rule.

[Appendix 19]
The table classification means is operated so as to use one of a distance of the number of characters of a table element, an edit distance of a morpheme configuration, an edit distance of a character type configuration, or a combination thereof for calculating a distance between table elements. A table classification program as described in the above supplementary note.

[Appendix 20]
The table classification means uses the table element feature calculation rule as a rule to determine that the table elements are continuous in a direction in which the ratio of the values of the table elements is less than or equal to a certain threshold value, and the distance between the table elements in the row direction. And the sum of the column directions divided by the number of comparisons of the table elements, and the operation is performed using any one of the methods for determining the direction in which the table elements are continuous. Table classification program.

[Appendix 21]
Information processing device
The program for table classification as described in the above supplementary note, wherein the program is further operated as a table header specifying means for specifying a table header from a table configuration feature and a preset header specifying rule.

[Appendix 22]
Information processing device
The table classification program according to the above supplementary note, which is further operated as a table reclassification unit for further classifying the table classified by the table classification unit based on the table configuration information and a preset table classification rule.

[Appendix 23]
The table structure information determining means is operated so as to use a table in which each distance between table elements in a specific direction is the smallest or a table close to the average for the table in each classification as a representative. Table classification program.

[Appendix 24]
The table classification means calculates a degree of similarity of table element features within a group of tables with the same classification, calculates a degree of coincidence of individual tables within the table group, and sets a predetermined threshold value within the table group. The table classification program as set forth in the above supplementary note, wherein the table classification program is operated so as to ask the user to confirm whether or not to include a table out of the range in the same classification.

  INDUSTRIAL APPLICABILITY The present invention can be used to convert table information of documents such as specifications and Web pages including many tables into documents at once and to give meanings.

10 Table classification device (table classification device)
101 Table configuration information extraction unit (table configuration information extraction means)
102 Table classification part (table classification means)
103 Table structure information determination unit (table structure information determination means)
104 Table heading specifying part (table heading specifying means)
105 Table reclassification part (table reclassification means)
110 Storage unit (storage means)
111 Table structure information 112 Table structure information 113 Heading identification rule 114 Table classification rule

Claims (10)

A table configuration information extraction unit that extracts information constituting individual tables from the document as table configuration information,
A table classification unit that calculates a distance between table elements between each element of each table, identifies a method of continuation of the table elements having a close distance, and classifies the individual tables based on the method of continuation;
A table classification apparatus comprising at least: Based on the extracted table element features or user designation, table structure information indicating the data structure of the table that is a set of attributes and attribute values of one table in each classification is determined and converted into data, and the same classification and A table structure information determination unit that converts the remaining table into data as a table structure based on the same rule,
The table classification device according to claim 1, further comprising:   The table classification unit uses a distance of the number of characters of a table element, an edit distance of a morpheme configuration, an edit distance of a character type configuration, or a combination thereof for calculating a distance between table elements. The table classification device according to claim 2.   The table classification unit determines, as a table element feature calculation rule, a method of determining that a table element is continuous in a direction in which the ratio of the value of the table element distance is equal to or less than a certain threshold, and the table element distance in the row direction. 4. The table according to claim 3, wherein any one of the methods of comparing the sum of the column direction and the total of the column directions divided by the number of comparisons of the table elements and determining the direction in which the table elements are continuous is used. Classification device.   The table classification device according to any one of claims 1 to 4, further comprising a table header identification unit that identifies a table header from a table configuration feature and a preset header identification rule. .   The table reclassifying unit for further classifying the table classified by the table classifying unit based on the table configuration information and a preset table classification rule, according to any one of claims 1 to 5. Table sorter.   The table structure information determination unit uses a table in which a distance between table elements in a specific direction is the smallest or a table close to an average for the table in each classification as a representative. The table classification device according to one item.   The table classification unit calculates similarity of table element features within the same group of tables and calculates the degree of coincidence of individual tables within the table group, and a predetermined threshold value within the table group. The table classification apparatus according to claim 1, wherein the table is determined to be included in the same classification for a table that is not included in the table. Table configuration information extraction processing for extracting information constituting individual tables from the document as table configuration information,
A table classification process for calculating a distance between table elements of each element of each table, identifying a method of continuation of the table elements having a close distance, and classifying the individual tables based on the method of continuation;
Table classification method characterized by having. Information processing device
Table configuration information extraction means for extracting information constituting individual tables from the document as table configuration information,
Calculate the distance between the table elements of each table element, identify how the table elements that are close to each other are continuous, and operate as a table classification unit that classifies the individual tables based on the sequence. A table classification program characterized by that.

Images