JP2009146121A - Document editing device, document editing method and computer program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a device and a method that modify a document including words on different levels of abstraction to a document using words on a uniform level of abstraction. <P>SOLUTION: Words are extracted from an input document. Classes on an ontology corresponding to the extracted words are selected as corresponding classes. Classes having a high statistic representing information on the frequency of occurrence in a document corpus are selected as reference classes from neighboring classes of the corresponding classes. A document abstraction level that is the variance or mean of levels of abstraction of all the extracted corresponding classes is compared with a threshold to determine a level of deviation of word abstraction levels of the input document. If the level of deviation is high, the words are replaced with words corresponding to the reference classes or reference class information is presented. This technique can modify a document using words on different levels of abstraction to generate a document using words on a uniform level of abstraction. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a document editing apparatus, a document editing method, and a computer program. More specifically, the present invention relates to a document editing apparatus, a document editing method, and a computer program that perform processing for adjusting the abstraction level of words included in a document to make the abstraction level substantially constant.

  In document creation, the word used in the document is determined by the document creator, but words with different abstractions may be used at random depending on the knowledge bias of the document creator. For example, a document created by a specialist in the medical field, specifically a document describing a diagnosis result, includes various technical terms, but common terms are also mixed, and each word has a high level of abstraction. Often different. If the abstraction level of the words contained in the document is different, the consistency of the description is lost and the document may be difficult to understand for the reader.

  As a conventional technique for executing editing processing for unifying styles and terms by editing processing of an arbitrarily created document, for example, there are the following documents.

  In Patent Document 1 (Japanese Patent Application Laid-Open No. 07-219930), a set of documents to be learned is set, a format of these document sets is learned, and a processing target document for executing a format unification process is input. An apparatus and method for unifying document formats based on learning results are disclosed. As specific processing, for example, when a document format management table storing correspondences between headings and different notations is prepared, a unified flag is set for the headings that appear in the learning document, and a unified target sentence is input Based on this unified flag, warnings and candidates for unification are presented to support unification of styles.

  Patent Document 2 (Japanese Patent Application Laid-Open No. 2004-240859) discloses a system in which words that require specialized knowledge appear in a text in an easy-to-understand manner. For example, a paraphrase candidate is presented using a dictionary that stores paraphrases for words that appear in text entered by the user. At this time, the paraphrase candidates are presented based on the level set by the user, the phrase level, and the user's past paraphrase history.

  Patent Document 3 (Japanese Patent Laid-Open No. 2000-268304) defines a text type and a pre-edit process for the text in advance, and automatically identifies the text type when the text is input. Discloses an apparatus and method for editing text using a pre-editing process corresponding to the type. First, perform morphological analysis on the text you want to edit automatically, extract vocabulary attributes such as parts of speech, and then determine the text type from the vocabulary attribute based on the prepared identification rules, and according to the text type The text is automatically edited using the pre-editing process.

  Various processing configurations have been proposed for apparatuses that perform document editing processing in this way. In many prior arts, word conversion (unification, paraphrase, editing) is performed according to a reference document set, a preset phrase level, and a document type. However, these processes are basically conversions to other notations that are semantically synonymous, and standardization that takes into account the degree of abstraction as meaning and concepts, such as converting the entire document into a more abstract and detailed representation. It is not the structure which performs.

In a document in which consistency of description is important, it is desirable that the degree of abstraction of expression is uniform throughout the document. For example, as medical text,
“The left lung S3 has a size… with a spinous process… as a differentiation, lung cancer”
If there is a text such as
A region called “left lung S3”,
The state of spinous processes,
The disease name "lung cancer",
Comparing these notations, the disease name “lung cancer” is more abstract than other notations, and the disease name has an abstraction level corresponding to the notation of the part or state, that is,
A more detailed expression such as “primary lung cancer” is good.
Japanese Patent Laid-Open No. 07-219930 JP 2004-240859 A JP 2000-268304 A

  An object of the present invention is to provide a document editing apparatus, a document editing method, and a computer program that perform standardization in the sense that the consistency of description, that is, the level of abstraction of the entire document is constant.

  Specifically, for example, a document corpus and an ontology for a specific field that are the basis for standardization are prepared, and the statistics (number of occurrences or values corresponding to them) of classes in the ontology are calculated based on words that appear in the document corpus. To do. When a user inputs a document to the system, an ontology class corresponding to each word in the document is extracted, and the degree of abstraction is calculated using a class statistic calculated in advance. Note that a unique method is used as a method for calculating the degree of abstraction. When the abstraction level of the class corresponding to each word in the document varies, the system informs the user of the word that causes the variation in the abstraction level in the document, or the word that causes the variation in the abstraction level in the document. Editing apparatus, document editing method, and computer program are provided.

The first aspect of the present invention is:
A document editing device that executes an input document editing process;
A word extraction unit for extracting words contained in the input document;
A corresponding class extracting unit that extracts ontology classes corresponding to the extracted words as corresponding classes;
A class statistic database in which a class ID that is an identifier of a class defined on the ontology and a statistic indicating appearance frequency information in the document corpus of the class-corresponding word are associated and stored;
In a class hierarchy defined on the ontology, a reference class extraction unit that selects, as a reference class, a class having a high statistic from neighboring classes extracted by the corresponding class extraction unit;
A class abstraction degree calculation unit for calculating the abstraction degree of each of the reference class and the corresponding class;
A document abstraction level calculation unit that calculates a variance or average of abstraction levels of all corresponding classes extracted corresponding to each word in the input document as a document abstraction level;
By comparing the document abstraction level with the threshold value, the degree of variation in the abstraction level of the words included in the input document is determined. If the degree of variation is determined to be large, the word included in the input document is determined to correspond to the reference class. A document standardization processing unit for processing to replace words or presenting reference class information;
In a document editing apparatus.

  Furthermore, in one embodiment of the document editing apparatus of the present invention, the corresponding class extracting unit performs a process of extracting an ontology class having a property corresponding to an extracted word from the input document as a corresponding class. And

  Furthermore, in one embodiment of the document editing apparatus of the present invention, the class statistics database has a configuration in which TF-IDF (Term Frequency-Inverse Document Frequency) indicating appearance frequency information in the document corpus is recorded as a statistics. It is characterized by being.

  Furthermore, in one embodiment of the document editing apparatus of the present invention, the class statistic database is calculated and calculated for each document creator of TF-IDF (Term Frequency-Inverse Document Frequency) which is appearance frequency information in the document corpus. The reciprocal of the variance of [TF-IDF] corresponding to a plurality of document creators is recorded as a statistic.

  Furthermore, in an embodiment of the document editing apparatus of the present invention, the reference class extracting unit is configured to set a predetermined number of hierarchies starting from the corresponding class extracted by the corresponding class extracting unit among the class hierarchies defined on the ontology. It is characterized in that a separation relation class is excluded from neighboring classes within [n], and a process of selecting a class having the maximum statistic as a reference class is executed.

  Furthermore, in one embodiment of the document editing apparatus according to the present invention, the reference class extraction unit extracts, as a reference class, a class having a statistic that is greater than a statistic of the corresponding class by a predetermined value (p%). The process which performs is performed.

  Furthermore, in one embodiment of the document editing apparatus of the present invention, the class abstraction level calculation unit determines the abstraction level of the corresponding class in the vicinity of the reference class based on the distance of the hierarchy from the reference class on the ontology. This is a configuration for performing processing, and is characterized in that an abstraction level setting is performed such that the abstraction level of the upper class is high and the abstraction level of the lower class is low.

  Furthermore, in an embodiment of the document editing apparatus of the present invention, the class abstraction degree calculation unit performs a process of determining the abstraction degree of the corresponding class in the vicinity of the reference class based on a difference from the number of properties of the reference class. This is a configuration in which an abstraction level is set such that a class with a small number of properties has a high level of abstraction and a class with a large number of properties has a low level of abstraction.

Furthermore, the second aspect of the present invention provides
A document editing method for executing an input document editing process in a document editing apparatus,
A word extracting unit for extracting a word included in the input document;
A corresponding class extraction step in which a corresponding class extraction unit extracts a class on the ontology corresponding to the extracted word as a corresponding class;
The reference class extraction unit selects, as a reference class, a class having a high statistic indicating the appearance frequency information in the document corpus from the neighborhood classes of the corresponding class extracted by the corresponding class extraction unit in the class hierarchy defined on the ontology. A base class extraction step;
A class abstraction level calculating step in which a class abstraction level calculation unit calculates an abstraction level of each of the reference class and the corresponding class;
A document abstraction level calculation unit that calculates a variance or average of the abstraction levels of all corresponding classes extracted corresponding to each word in the input document as a document abstraction level;
The document standardization processing unit determines the degree of variation in the abstraction level of the words included in the input document by comparing the document abstraction level and the threshold value. If the degree of variation is determined to be large, the word included in the input document A document standardization processing step of replacing a word with a word corresponding to the reference class, or presenting reference class information;
The document editing method is characterized by comprising:

  Furthermore, in one embodiment of the document editing method of the present invention, the corresponding class extracting step performs a process of extracting an ontology class having a property corresponding to an extracted word from the input document as a corresponding class. And

  Further, in one embodiment of the document editing method of the present invention, the reference class extracting step includes a predetermined number of hierarchies starting from the corresponding class extracted by the corresponding class extracting unit in the class hierarchies defined on the ontology. It is characterized in that a separation relation class is excluded from neighboring classes within [n], and a process of selecting a class having the maximum statistic as a reference class is executed.

  Furthermore, in one embodiment of the document editing method of the present invention, the reference class extracting step extracts a class having a statistic that is larger than a statistic of the corresponding class by a predetermined value (p%) and having a maximum statistic as a reference class. The process which performs is performed.

  Furthermore, in one embodiment of the document editing method of the present invention, the class abstraction level calculating step determines an abstraction level of a corresponding class in the vicinity of the reference class based on a hierarchy distance from the reference class on the ontology. This is a configuration for performing processing, and is characterized in that an abstraction level setting is performed such that the abstraction level of the upper class is high and the abstraction level of the lower class is low.

  Furthermore, in one embodiment of the document editing method of the present invention, the class abstraction level calculating step includes a process of determining an abstraction level of a corresponding class in the vicinity of the reference class based on a difference from the number of properties of the reference class. This is a configuration in which an abstraction level is set such that a class with a small number of properties has a high level of abstraction and a class with a large number of properties has a low level of abstraction.

Furthermore, the third aspect of the present invention provides
In a document editing apparatus, a computer program for executing input document editing processing,
A word extraction step for causing the word extraction unit to extract words included in the input document;
A corresponding class extraction step for causing the corresponding class extraction unit to extract a class on the ontology corresponding to the extracted word as a corresponding class;
Let the reference class extraction unit select, as a reference class, a class having a high statistic indicating appearance frequency information in the document corpus from the neighborhood classes of the corresponding class extracted by the corresponding class extraction unit in the class hierarchy defined on the ontology. A base class extraction step;
A class abstraction calculating step for causing the class abstraction calculating unit to calculate the abstraction of each of the reference class and the corresponding class;
A document abstraction level calculation step for causing the document abstraction level calculation unit to calculate the variance or average of the abstraction levels of all corresponding classes extracted corresponding to each word in the input document as the document abstraction level;
The document standardization processing unit determines the degree of variation in the abstraction level of the word included in the input document by comparing the document abstraction level and the threshold value. If the degree of variation is determined to be large, the word included in the input document A document standardization processing step that replaces the word with a word corresponding to the reference class, or presents the reference class information,
There is a computer program characterized by comprising:

  The computer program of the present invention is, for example, a computer program that can be provided by a storage medium or a communication medium provided in a computer-readable format to a general-purpose computer system that can execute various program codes. . By providing such a program in a computer-readable format, processing corresponding to the program is realized on the computer system.

  Other objects, features, and advantages of the present invention will become apparent from a more detailed description based on embodiments of the present invention described later and the accompanying drawings. In this specification, the system is a logical set configuration of a plurality of devices, and is not limited to one in which the devices of each configuration are in the same casing.

  According to the configuration of an embodiment of the present invention, a word included in an input document is extracted, an ontology class corresponding to the extracted word is selected as a corresponding class, and further, from a neighboring class of the corresponding class, in a document corpus Select a class with high statistics indicating the appearance frequency information as a reference class, calculate the variance or average of the abstraction levels of all corresponding classes extracted corresponding to each word in the input document as the document abstraction level, By comparing the calculated document abstraction level with the threshold value, the degree of variation in the abstraction level of the words included in the input document is determined. If it is determined that the degree of variation is large, the word included in the input document is determined to correspond to the reference class. Since it is configured to replace the word or present the reference class information, change the document using words with varying degrees of abstraction to make uniform extraction. It is possible to generate a use document of a word with a degree.

  Hereinafter, a document editing apparatus, a document editing method, and a computer program according to an embodiment of the present invention will be described in detail with reference to the drawings.

  FIG. 1 shows a block diagram of a document editing apparatus according to an embodiment of the present invention. The document editing apparatus 100 of the present invention inputs an arbitrary document to be processed from the document input unit 101, and is a document in which the abstraction level of words used in the input document is adjusted to a substantially uniform level, that is, a standardized document Is to create.

  The document editing apparatus 100 according to the present invention prepares a document corpus and an ontology in which document data of a specific field as a basis for standardization is accumulated, and based on words appearing in the document corpus, statistics of classes in the ontology (number of appearances) Or a value in accordance with it), extract ontology classes corresponding to each word in the input document, calculate their abstraction using pre-calculated class statistics, and calculate each word When the level of abstraction of the corresponding class varies, a process for notifying the user of a word that causes a variation in the abstraction level in the document and a process for automatically correcting the word that causes the variation in the abstraction level are performed.

  As shown in FIG. 1, a document editing apparatus 100 according to an embodiment of the present invention includes a document input unit 101, a document extraction unit 102, a word extraction unit 103, a corresponding class extraction unit 104, a class statistic calculation unit 105, a reference A class extraction unit 106, a class abstraction level calculation unit 107, a document abstraction level calculation unit 108, a document standardization processing unit 109, a document corpus storage unit 121 in which document data in a specific field that is the basis of standardization are collected, and an ontology are stored. The ontology database 122 includes an ontology class ID and a class statistics database 123 that holds values representing statistics in the document corpus.

  Before describing the processing using the document editing apparatus 100, each of the document corpus storage unit 121, the ontology database 122, and the class statistics database 123 will be described. In the following embodiment, a processing example related to a medical document will be described. Accordingly, the corpus and ontology used in this example are a corpus and ontology related to medical documents and terms.

  An example of a text corpus stored in the document corpus storage unit 121 will be described with reference to FIG. The document corpus storage unit 121 is a database that stores various texts. As shown in FIG. 2, an ID as a text identifier and an identifier of a text creator (such as a creator name) are set for each text.

  Next, the ontology stored in the ontology database 122 will be described with reference to FIG. An ontology means a set of knowledge classification systems and inference rules, and is a kind of concept dictionary that represents existence, and is configured as a dictionary that describes word relationships between superordinate concepts and subordinate concepts. For example, it has a hierarchical structure in which a relationship between a superordinate concept and a subordinate concept of a concept is defined by a connection relationship between nodes. The ontology describes a class corresponding to [concept], properties as class property information, attributes as class attribute information, and the like.

  FIG. 3 shows an information configuration example of one registered word described according to OWL known as an ontology description language. In the ontology database 122, registration information as shown in FIG. 3 is recorded for each class corresponding to a concept that is a registered word.

As shown in FIG. 3, the ontology includes, for example, the following information:
(A) a class corresponding to [concept],
(B) a label as class property information;
(C) Relationship information with other classes which are class attribute information (for example, parent-child relationship),
These pieces of information are registered.

[Class] is a field for recording a concept name representing the registered existence. In the class field, words and conceptual information of words are recorded as class identifiers (class IDs). In the example shown in the figure,
Class identifier (class ID) = [Epithelial Neoplasm]
It is. In addition, Epithelial Neoplasm is a medical term and means “epithelial tumor” as a disease name.
[Property] records property information of a class such as a label. Further, relation information with other classes is recorded as attribute information.

  The relationship with other classes will be described with reference to FIG. The ontology records the registration information shown in FIG. 3 in units of classes corresponding to concepts, and has a configuration that allows each class to be identified as one node and the relationship between classes to be identified as a connection relationship between nodes. As relations between classes, three relations are defined when roughly classified. With the class A201 shown in FIG. 4 as the class of interest, the class relationship with the class A201 will be described.

First, the class P211 is a superordinate concept class corresponding to the class A201.
This class P211 is a superclass (parent class) of class A201.

The class A 201 includes a class S 221 and a class T 222 as subordinate concept classes corresponding to the class A 201.
The class S221 and the class T222 are subclasses (child classes) of the class A201.

Further, classes registered as subordinate concepts of class P211 include class B212 and class C213 in addition to class A201. Class B212 and class C213 have the same superclass (parent class) = class P211 as class A201.
The class B212 and the class C213 are sibling classes (sibling classes) of the class A201.

In this way, the ontology class relationship is
(A) Super class (parent class)
(B) Subclass (child class)
(C) Sibling class (sibling class)
These class relationships are defined. Furthermore, a subclass of a subclass (grandchild class) and the like are also defined.

  Next, the class statistic database 123 will be described with reference to FIG. The class statistic database 123 is a database that holds a class ID of each class recorded in the ontology and a value representing a statistic in the document corpus. The construction process of the class statistics database 123 will be described later.

  As shown in FIG. 5, the class statistic database 123 holds a class ID of each class recorded in the ontology and a value representing the statistic in the document corpus. As described above with reference to FIG. 3, the class ID is an ID indicating the concept recorded in the field for recording the concept name of the word registered in the ontology database. The statistic is frequency information that this word appears in the document stored in the document corpus DB 121. Specifically, TF-IDF (Term Frequency-Inverse Document Frequency) = (number of occurrence of word / occurrence of word) Number of documents) is used.

  The TF-IDF is used as an index indicating the importance of words appearing in a document, for example, usefulness as a search word. For example, the TF-IDF of a word that appears frequently in one document increases. However, when it appears frequently in many documents, the value of TF-IDF becomes small. That is, in the case of a word that is generally used in a document, the value of TF-IDF tends to be small, and the value of TF-IDF tends to increase if it appears frequently only in a specific document. When used as a keyword for selection, it is effective to use a word having a high TF-IDF value.

Next, processing using the document editing apparatus 100 shown in FIG. 1 will be described. The following two processes will be described sequentially.
(A) Class statistic database construction process (b) Input document standardization process During the processes (a) and (b) above, (a) class statistic database construction process is performed as follows: This is a process of constructing a class statistic database used for standardization processing, and (b) is a process for constructing a database by executing it in advance before executing the standardization processing of input documents. Once the database is constructed, it is not necessary to perform the same processing thereafter. However, when the document corpus database 121 and the ontology database 122 are updated, it is preferable to update the class statistic database 123 according to the same process as the construction process as necessary.

(A) Class Statistics Database Construction Process First, a class statistics database construction process sequence will be described with reference to the flow shown in FIG.

  In step S101, one document is extracted from the document corpus. This process is a process in which the document extraction unit 102 of the document editing apparatus 100 shown in FIG. 1 extracts a document from the document corpus database 121.

  Next, in step S102, morphological analysis is performed on the text of the document extracted from the document corpus, the document is separated, and a set of words is extracted. This process is a process of the word extraction unit 103 of the document editing apparatus 100 shown in FIG. An existing morpheme analysis program can be applied to the morpheme analysis. Specifically, morphological analysis is performed using, for example, a tea bowl (http://chasen.naist.jp/hiki/ChaSen/).

  Next, in step S103, for each extracted word, the correspondence with the ontology class stored in the ontology database 122 is analyzed. This processing is executed as processing of the corresponding class extraction unit 103 of the document editing apparatus 100 shown in FIG.

  In this correspondence class extraction process, for each word extracted from a document stored in the document corpus database 121, a label which is class property information recorded in ontology class correspondence data (see FIG. 3) stored in the ontology database 122. It is executed as a process for determining the presence or absence of a word that matches.

For example, the morphological analysis result of the document with document ID = 1 extracted from the document corpus is
“… / Residue / tumor / also / denial / can / don /”
For documents with text that is split
Ontology class [Neoplasm] in the ontology database 122 is recorded, and label information as property information of this class [Neoplasm]
Assume that "tumor", "tumor", and "mass" are recorded.
In this case, a recorded class having a label [tumor] similar to the word [tumor] in the document extracted from the document corpus exists in the ontology, and the following correspondence information (a) to (c) is generated. To do.
(A) Document ID extracted from the document corpus
(B) Words in the document extracted from the document corpus (c) Class name corresponding to the words That is, in the above example,
(Document ID, word, class name) = <1, tumor, Neoplasm>
In such a data format, the correspondence between the document, the word, and the class is acquired.

  In step S104, it is determined whether or not all the documents in the document corpus have been processed. If not, the process returns to step S101, and the processes in steps S101 to S103 are executed for the unprocessed document. If it is determined in step S104 that all the documents in the document corpus have been processed, the process proceeds to step S105.

  In step S105, a class statistic is calculated based on the correspondence between the extracted word and the class. This process is a process of the class statistic calculation unit 105 of the document editing apparatus 100 shown in FIG. As described above, for example, TF-IDF (Term Frequency-Inverse Document Frequency) = (number of occurrences of words / number of occurrences of words) is used as the statistic, as described above.

  As described above, TF-IDF is used as an index indicating the importance of words appearing in a document, for example, usefulness after retrieval. For example, the TF-IDF of a word that appears frequently in one document increases. However, when it appears frequently in many documents, the value of TF-IDF becomes small. That is, in the case of a word that is generally used in a document, the value of TF-IDF tends to be small, and the value of TF-IDF tends to increase if it appears frequently only in a specific document. When used as a keyword for selection, it is effective to use a word having a high TF-IDF value.

  In step S106, the class ID corresponding to the extracted word and the statistic (TF-IDF) are stored in the class statistic DB. That is, it is stored in the class statistics database 123 of the document editing apparatus 100 shown in FIG. As a result of this processing, the class statistic database 123 described above with reference to FIG. 5 is constructed.

(B) Input Document Standardization Processing Next, an input document standardization processing sequence executed using the document editing apparatus 100 of FIG. 1 will be described with reference to a flowchart shown in FIG.

  First, in step S201, a document to be processed is input from the document input unit. This process is a process of the document input unit 101 of the document editing apparatus 100 shown in FIG.

  Next, in step S202, morphological analysis is performed on the text of the document input from the document input unit, the document is separated, and a set of words is extracted. This process is a process of the word extraction unit 103 of the document editing apparatus 100 shown in FIG. For the morphological analysis, it is possible to use an existing morphological analysis program, specifically, a teacup (http://chasen.naist.jp/hiki/ChaSen/) described above.

  Next, in step S203, for each extracted word, the correspondence with the ontology class stored in the ontology database 122 is analyzed. This process is a process of the corresponding class extraction unit 104 of the document editing apparatus 100 shown in FIG. This process is a process for determining, for each extracted word, whether or not there is a word that matches the label that is the class property information recorded in the ontology class correspondence data stored in the ontology database 122 (see FIG. 3). Execute.

Similar to the processing in step S103 in the flow of FIG. 6 described above, when the ontology has a recorded class having the same label as the word extracted from the document, the following correspondence information (a) to ( c) is generated.
(A) Document ID extracted from the document corpus
(B) Word in the document extracted from the document corpus (c) Class name corresponding to the word, ie (document ID, word, class name)
In such a data format, the correspondence between the document, the word, and the class is acquired.

  In step S203, a class in which a label similar to a word included in the input document is recorded is extracted in this way. In the analysis process for the input document, in step S204, the neighborhood class is acquired according to the ontology class hierarchy (see FIG. 4) with the class corresponding to the word included in the input document as the corresponding class. The processes of steps S204 and S205 are processes of the reference class extracting unit 106 of the document editing apparatus 100 shown in FIG.

  For example, a class in the range of n layers (n is an arbitrary value of about 1 to 3) on the ontology is extracted from the corresponding class. In other words, according to the hierarchical structure described above with reference to FIG. 4, n-level classes adjacent to the corresponding class are also extracted. These extracted classes include the superclass (parent class) and subclass (child class) of the corresponding class in which the same label as the word included in the input document is recorded.

Even if the class is a neighborhood class, a class corresponding to the following condition is not selected.
(A) a class with no statistics set,
(B) a class that is “disjoint relationship” on the ontology, and a class that passes through the separation relationship class;
These classes are not selected.

An example of two classes in which a separation relationship (disjointWith) is established will be described. For example, as a class recorded in the ontology, a class indicating the name of the body part,
Class [hand],
Class [foot]
When there is, a separation relationship (disjointWith) may be defined between these classes. This means that an entity (instance) of a certain body part is a [hand] and cannot be a [foot]. That is, the two classes in which the separation relationship (disjointWith) is established are the relationships between the classes that cannot be compatible as the correct classes of both entities.

  Next, in step S205, a class having a statistic larger than the corresponding class by p% (p is an arbitrary value of about 10 to 50) and the largest statistic is extracted from the extracted neighboring classes as a reference class.

An example of reference class extraction processing will be described with reference to FIG. FIG. 8 shows the ontology class hierarchy structure stored in the ontology database 122.
Corresponding class [Carcinoma] 301 is a class extracted in step S203 as a class having the same label as the word included in the input document. The class having the hierarchical structure shown in FIG. The upper and lower two classes are shown.

  The value of [w] shown for each class is a statistic recorded in the class statistic database 123 described above with reference to FIG. 5, that is, a statistic indicating an index corresponding to the appearance frequency in the document corpus such as TD-IDF. Amount.

  Note that the separation relationship class 303 shown in FIG. 8 is a class that is in the “separation relationship (disjointWith)” described above with respect to the corresponding class 301 and is excluded from the selection target of the reference class.

Currently, the processing target class is the corresponding class [Carcinoma] 301 shown in FIG.
Number of neighborhood classes to be extracted [n]: n = 2
[P] as a statistic threshold to be selected as the reference class: p = 20%
If
From the class shown in FIG.
The reference class [Lung Carcinoma] is selected.
That is, among the plurality of classes excluding the corresponding class 301 and the separation relation class 303 shown in FIG.
The statistic [w] = 5.1 is larger than the set corresponding class by the statistic [w] = 5.1 (p is an arbitrary value of about 10 to 50),
Statistic [w] = 5.1,
The difference Δw from the statistic w = 4.2 of the corresponding class [Carcinoma] 301 is
Δw = 5.1-4.2
= 0.9
Is calculated as

On the other hand, when the statistic threshold is calculated based on the statistic w = 4.2 of the corresponding class [Carcinoma] 301,
[P]: p = 20%
Because it is the setting of
4.2 × 20% = 0.81
It becomes.
The difference Δw = 0.9 is greater than 0.81;
The statistic w = 5.1 of the class [Lung Carcinoma] is a larger value having a difference equal to or greater than the threshold p (p = 20%) than the statistic w = 4.2 of the corresponding class [Carcinoma] 301. .

The criteria class selection conditions are the following conditions 1 and 2.
(Condition 1) A class in the range of n layers (n is an arbitrary value of about 1 to 3) on the ontology from the corresponding class [n = 2 in this example], except for the separation relation class,
(Condition 2) A class whose statistic is larger than the corresponding class by p% (p is an arbitrary value of about 10 to 50) and whose statistic is the maximum [in this example, p = 20%]

  The class [Lung Carcinoma] satisfies the above conditions 1 and 2, and is selected as the reference class 302 of the corresponding class [Carcinoma] 301 as shown in FIG.

  In the example illustrated in FIG. 8, the statistic of the class [yyyyy] in the separation relation class 303 is larger than the reference class 302, but [yyyyy], [zzzzz], and [adenoma] are separated. Since it is a class and its subclass in (disjointWith), it is excluded from the extraction target.

  Next, in step S206, the abstraction level of each class (corresponding class) extracted corresponding to each word included in the input sentence and the reference class selected corresponding to the corresponding class corresponding to each word is calculated. . This process is the process of the class abstraction degree calculation unit 107 of the document editing apparatus 100 shown in FIG.

An example of class abstraction calculation processing will be described. Several methods can be applied to calculate the class abstraction, but basically,
The degree of abstraction [a] of the reference class is a = 0,
Further, the abstraction level of a neighboring class such as a corresponding class in the vicinity of the reference class is determined using one of the following methods (a) and (b).
(A) A method of calculating the degree of abstraction based on the distance of the hierarchy from the ontology reference class,
(B) A method for calculating the degree of abstraction based on the difference from the number of properties of the reference class,
Hereinafter, these two methods will be described with reference to FIGS.
explain.

(A) A method of calculating the degree of abstraction based on the distance of the hierarchy from the ontology reference class,
First, a method for calculating the degree of abstraction based on the hierarchy distance from the ontology reference class will be described with reference to FIG.

FIG. 9 shows a class hierarchy configuration including a class selected corresponding to a word included in the input unit, that is, a corresponding class 321 and a reference class 322 selected from the neighboring classes by the above-described method.
The correspondence class 321 is a class selected corresponding to the word included in the input unit, and is a class in which a label as property information corresponding to the word is recorded. The reference class 322 is the selection condition of the reference class described above, that is,
(Condition 1) Classes in the range of n layers (n is an arbitrary value of about 1 to 3) on the ontology from the corresponding class, except for the separation relation class,
(Condition 2) A class whose statistics are larger than the corresponding class by p% (p is an arbitrary value of about 10 to 50) and whose statistics are the largest,
The class is selected as a class that satisfies this condition.

  In this example, the abstraction level of each class is determined by the distance of the hierarchy from the reference class on the ontology. Specifically, when going from a certain class toward the super class (parent class), the value of abstraction is added (+1), and when moving toward the subclass (child class), the value of abstraction is minus (−1). ) To set the abstraction level of each class.

In the example shown in FIG. 9, first, the abstraction level [a] of the reference class 322 is set to a = 0. in this case,
The degree of abstraction [a] of the superclass (parent class) of the reference class 322 is a = 1
The abstraction level [a] of the subclass (child class) of the reference class 322 is a = −1
The degree of abstraction [a] of the sibling class (sibling class) of the reference class 322 is a = 0.
It becomes. In addition,
The degree of abstraction [a] of the superclass (parent class) of the reference class 322 and the superclass (parent class) is a = 2.
It becomes.
Thus, the abstraction level is set according to the class hierarchy.

  This method of setting abstraction basically follows the ontology class hierarchy. In other words, in the ontology, the level of abstraction as a word increases as it progresses in the superclass (parent class) direction, and the level of abstraction as a word increases. The level of abstraction as a word decreases as it progresses in the subclass (child class) direction. Yes, this abstraction level setting method is a method according to the ontology class hierarchy structure.

(B) A method for calculating the degree of abstraction based on the difference from the number of properties of the reference class,
Next, a method for calculating the abstraction level based on the difference from the number of properties of the reference class will be described with reference to FIG.

  As described above with reference to FIG. 3, labels as properties (property information) are set in each class of the ontology. This method is a method for setting the abstraction level of each class according to the number of properties (property information) set for this class.

  For example, the greater the number of properties set in a class, the lower the class correspondence abstraction value, and the smaller the number of properties set in the class, the higher the class correspondence abstraction value.

  Specifically, as shown in FIG. 10, for example, the abstraction level of the reference class 332 is first set to zero. The number of properties (property information) set in this reference class is four (p1 to p4).

Next, the abstraction level of each class is set according to the number of properties of the neighborhood class selected at the time of selecting the reference class.
Let the reference class have an abstraction level a = 0.
The number of properties (property information) of the base class is [4]
The number of properties of other classes is verified. Based on the number of properties [4] of the reference class, when the number of properties decreases by 1, the abstraction level a is set to plus 1, and when the number of properties increases by 1, the level of abstraction a is set to minus 1. To do.

  In the example shown in FIG. 10, for example, the correspondence class 331 is a class in which two properties are recorded, and the number of properties (property information) of the reference class is two fewer than [4], so the abstraction level a is a. = 2.

  This abstraction level setting method is basically a method corresponding to the setting state of ontology properties (property information). In other words, in the ontology, the word of the higher-level concept tends to increase in the degree of abstraction and the number of properties (property information) to be set, and the word of the lower-level concept decreases in the degree of abstraction and, at the same time, the property (property information). ) Tend to increase. The method described with reference to FIG. 10 is a method according to this ontology property setting configuration.

  In step S206 in the flow of FIG. 7, each class (corresponding class) extracted corresponding to each word included in the input sentence by the method described with reference to FIG. 9 or FIG. Calculate the abstraction level of the selected reference class corresponding to the class. This process is the process of the class abstraction degree calculation unit 107 of the document editing apparatus 100 shown in FIG.

  In step S207, the variance [V] of the abstraction level of all classes (corresponding classes) extracted corresponding to each word in the input document is calculated, and the calculated variance value [V] is calculated. Set as [Document abstraction]. Note that the abstraction level of the document is not limited to dispersion, and an average or the like may be used. The processing of steps S207 to S209 is executed as processing of the document abstraction degree calculation unit 108 of the document editing apparatus 100 shown in FIG.

Next, in step S208, the variance [V] is compared with a predetermined threshold [m]. In step S209,
Variance V ≧ threshold m
It is determined whether or not the above equation holds.
When the above expression does not hold, that is, when the variance [V] is less than the threshold [m], there is little variation in the abstraction level of the words included in the input sentence to be processed. That is, it is determined that the word uses words having approximately the same level of abstraction, and the process is terminated.

on the other hand,
Variance V ≧ threshold m
When the above formula is satisfied, that is, when the variance [V] is equal to or greater than the threshold value [m], variations in the abstraction level of words included in the input sentence to be processed are large. That is, it is determined that the word uses a different abstraction level, and the process proceeds to step S210.

  Step S210 is a document standardization process, which is executed by the document standardization processing unit 109 shown in FIG. As standardization processing, editing processing is performed for each word in which the reference class detected in the vicinity of the defined class corresponding to each word is replaced with the word corresponding to the reference class for the word used in the input sentence. By this editing process, the words included in the edited sentence are replaced with the words of the reference class, and are changed to sentences composed of words with the same level of abstraction.

  In addition, although it is good also as a structure which performs such an automatic edit, the reference class information selected corresponding to each word contained in an input sentence is shown to a user without performing an automatic edit, and a user is shown. A configuration may be adopted in which it is determined whether or not replacement with a word corresponding to the reference class is performed, and editing processing is executed according to the user's intention.

  As described above, the document editing apparatus according to the present invention is configured to execute processing for making the abstraction level uniform for each word included in the document, and automatically uses sentences having various abstraction levels. It is possible to generate a document (standardized document) composed of words with the same level of abstraction.

[Second Embodiment]
FIG. 11 shows the configuration of the second embodiment of the document editing apparatus of the present invention. The difference between the document editing apparatus 500 shown in FIG. 11 and the document editing apparatus 100 shown in FIG. 1 is that the class statistic calculation unit 105 in the configuration shown in FIG. It is replaced with the calculation unit 501 and the class statistic calculation unit (distribution) 502, and the data stored in the class statistic database 503 is changed.

  That is, in the second embodiment, the data stored in the class statistic database 503 is different from the previous embodiment, and is data in which variation for each document creator is excluded.

  Specifically, in the previous embodiment, the class statistic database construction processing sequence described with reference to FIG. 6 is different. In the input document standardization processing described with reference to FIG. 7, the statistics of the class statistics database 503 are applied.

  The class statistic database construction processing sequence in this embodiment will be described with reference to the flow shown in FIG.

  In step S301, one document is extracted from the document corpus. This process is a process in which the document extraction unit 102 of the document editing apparatus 500 shown in FIG. 11 extracts a document from the document corpus database 121. In this process, document creator information is also acquired.

  Next, in step S302, morphological analysis is performed on the text of the document extracted from the document corpus, the document is separated, and a set of words is extracted. This process is a process of the word extraction unit 103 of the document editing apparatus 500 shown in FIG. An existing morpheme analysis program can be applied to the morpheme analysis. Specifically, morphological analysis is performed using, for example, a tea bowl (http://chasen.naist.jp/hiki/ChaSen/).

  Next, in step S <b> 303, for each extracted word, the correspondence with the ontology class stored in the ontology database 122 is analyzed. This processing is executed as processing of the corresponding class extraction unit 103 of the document editing apparatus 500 shown in FIG.

  In this correspondence class extraction process, for each word extracted from a document stored in the document corpus database 121, a label which is class property information recorded in ontology class correspondence data (see FIG. 3) stored in the ontology database 122. It is executed as a process for determining the presence or absence of a word that matches.

  In step S304, it is determined whether or not all the documents in the document corpus have been processed. If not, the process returns to step S301, and the processes in steps S301 to S303 are executed for the unprocessed document. If it is determined in step S304 that all the documents in the document corpus have been processed, the process proceeds to step S305.

  In step S305, a class statistic is calculated based on the correspondence between the extracted word and the class. This process is performed by the document creator correspondence class statistic calculation unit 501 of the document editing apparatus 500 shown in FIG. As described above, for example, TF-IDF (Term Frequency-Inverse Document Frequency) = (number of occurrences of words / number of occurrences of words) is used as the statistic, as described above.

  However, in this embodiment, this statistic is calculated as individual data for each document creator. In other words, each document creator has a “癖” for document creation, and each author has a bias in terms of frequently used words and words that are not used. Therefore, this statistic may vary greatly depending on the creator.

  In step S306, the variance of the statistics (TF-IDF) corresponding to each document creator of each class corresponding to the extracted word is calculated, and the reciprocal of the calculated variance is used as the statistics of the class. This process is a process of the class statistic calculation unit 502 of the document editing apparatus 500 shown in FIG.

  In step S307, the class statistic is associated with the class ID corresponding to the extracted word and the [reciprocal of variance for each document creator (TF-IDF)] calculated in step S306 as the class-corresponding statistic. Store in DB.

  As a result of this processing, the data stored in the class statistics database 503 of the document editing apparatus 500 shown in FIG. 11 is recorded with statistics with little variation that are not influenced by “癖” for each creator.

  The document standardization processing in the present embodiment is the same processing as the flow shown in FIG. 7 described in the previous embodiment, but the statistic value used in step S205 in the flow in FIG. It is a different value and is not affected by the above-mentioned “癖” for each creator, etc., and has a small statistic (the reciprocal of the variance (TF-IDF) for each document creator). Selection of the reference class is executed, and document standardization with higher quality is realized.

Finally, an example of the hardware configuration of the information processing apparatus constituting the document editing apparatus that executes the above-described processing will be described with reference to FIG. The information processing apparatus constituting the document editing apparatus can be realized as hardware by, for example, a PC. The CPU (Central Processing Unit) 701 is a process corresponding to an OS (Operating System) or the above-described embodiment.
(A) Class statistic database construction process (b) Input document standardization process These processes are executed. These processes are executed according to a computer program stored in a data storage unit such as a ROM or a hard disk of each information processing apparatus.

  A ROM (Read Only Memory) 702 stores programs used by the CPU 701, calculation parameters, and the like. A RAM (Random Access Memory) 703 stores programs used in the execution of the CPU 701, parameters that change as appropriate during the execution, and the like. These are connected to each other via a host bus 704 including a CPU bus.

  The host bus 704 is connected to an external bus 706 such as a PCI (Peripheral Component Interconnect / Interface) bus via a bridge 705. A keyboard 708 and a pointing device 709 are input devices operated by the user. The display 710 is composed of a liquid crystal display device, a CRT (Cathode Ray Tube), or the like, and displays various types of information as text and images.

  An HDD (Hard Disk Drive) 711 includes a hard disk, drives the hard disk, and records or reproduces a program executed by the CPU 701 and information. The hard disk is used, for example, as storage means such as ontology, text corpus, class statistics, dictionary, and further stores various computer programs such as data processing programs.

  The drive 712 reads data or a program recorded on a removable recording medium 721 such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory, and reads the data or program as an interface 707 and an external bus 706. , The bridge 705, and the RAM 703 connected via the host bus 704.

  The connection port 714 is a port for connecting the external connection device 722 and has a connection unit such as USB or IEEE1394. The connection port 714 is connected to the CPU 701 and the like via the interface 707, the external bus 706, the bridge 705, the host bus 704, and the like. The communication unit 7515 is connected to the network and executes data search by communicating with an external database 801, for example.

  Note that the hardware configuration example of the information processing apparatus shown in FIG. 13 is an example of an apparatus configured by applying a PC, and the document editing apparatus of the present invention is not limited to the configuration shown in FIG. Any configuration can be used as long as the processing described in the above item can be executed.

  The present invention has been described in detail above with reference to specific embodiments. However, it is obvious that those skilled in the art can make modifications and substitutions of the embodiments without departing from the gist of the present invention. In other words, the present invention has been disclosed in the form of exemplification, and should not be interpreted in a limited manner. In order to determine the gist of the present invention, the claims should be taken into consideration.

  The series of processing described in the specification can be executed by hardware, software, or a combined configuration of both. When executing processing by software, the program recording the processing sequence is installed in a memory in a computer incorporated in dedicated hardware and executed, or the program is executed on a general-purpose computer capable of executing various processing. It can be installed and run. For example, the program can be recorded in advance on a recording medium. In addition to being installed on a computer from a recording medium, the program can be received via a network such as a LAN (Local Area Network) or the Internet, and installed on a recording medium such as a built-in hard disk.

  Note that the various processes described in the specification are not only executed in time series according to the description, but may be executed in parallel or individually according to the processing capability of the apparatus that executes the processes or as necessary. Further, in this specification, the system is a logical set configuration of a plurality of devices, and the devices of each configuration are not limited to being in the same casing.

  As described above, according to the configuration of the embodiment of the present invention, the words included in the input document are extracted, the ontology class corresponding to the extracted word is selected as the corresponding class, and the corresponding class A class with high statistics indicating the appearance frequency information in the document corpus is selected as a reference class from neighboring classes, and the variance or average of the abstraction levels of all corresponding classes extracted corresponding to each word in the input document is documented. Calculated as the degree of abstraction, and by comparing the calculated document abstraction level with the threshold value, the degree of variation in the abstraction level of the words included in the input document is determined, and if it is determined that the degree of variation is large, it is included in the input document This is a process that replaces the generated word with a word corresponding to the reference class, or presents the reference class information. The change, it is possible to generate a use document of a word with a uniform abstraction.

It is a block diagram shown about the structural example of the document editing apparatus which concerns on one Example of this invention. It is a figure explaining the example of the document corpus utilized in the document editing apparatus concerning one Example of this invention. It is a figure explaining the example of ontology utilized in the document editing apparatus concerning one Example of this invention. It is a figure explaining the example of ontology utilized in the document editing apparatus concerning one Example of this invention. It is a figure explaining the example of the class statistic utilized in the document editing apparatus concerning one Example of this invention. It is a figure which shows the flowchart explaining the construction processing sequence of the class statistic database performed in the document editing apparatus which concerns on one Example of this invention. It is a figure which shows the flowchart explaining the standardization process sequence of the input document performed in the document editing apparatus concerning one Example of this invention. It is a figure explaining the extraction process example of the reference | standard class performed in the document editing apparatus which concerns on one Example of this invention. It is a figure explaining the method of calculating an abstraction level based on the distance of the hierarchy from the reference class on the ontology performed in the document editing apparatus which concerns on one Example of this invention. It is a figure explaining the method of calculating an abstraction level based on the difference with the property number of the reference | standard class performed in the document editing apparatus which concerns on one Example of this invention. It is a block diagram shown about the structural example of the document editing apparatus which concerns on one Example of this invention. It is a figure which shows the flowchart explaining the construction processing sequence of the class statistic database performed in the document editing apparatus which concerns on one Example of this invention. It is a figure explaining the hardware structural example of the information processing apparatus which comprises the document editing apparatus which concerns on one Embodiment of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 Document editing apparatus 101 Document input part 102 Document extraction part 103 Word extraction part 104 Corresponding class extraction part 105 Class statistic calculation part 106 Reference class extraction part 107 Class abstraction degree calculation part 108 Document abstraction degree calculation part 109 Document standardization processing part 121 Document corpus storage unit 122 Ontology database 123 Class statistics database 201 Class A
211-213 Class 221-222 Class 301 Corresponding Class 302 Standard Class 303 Separation Relation Class 321 Corresponding Class 322 Standard Class 331 Corresponding Class 332 Standard Class 500 Document Editing Device 501 Document Creator Corresponding Class Statistics Calculation Unit 502 Class Statistics Calculation Unit 503 Class Statistics Database 701 CPU (Central Processing Unit)
702 ROM (Read-Only-Memory)
703 RAM (Random Access Memory)
704 Host bus 705 Bridge 706 External bus 707 Interface 708 Keyboard 709 Pointing device 710 Display 711 HDD (Hard Disk Drive)
712 Drive 714 Connection port 715 Communication unit 721 Removable recording medium 722 Externally connected device 801 Database

Claims (15)

A document editing device that executes an input document editing process;
A word extraction unit for extracting words contained in the input document;
A corresponding class extracting unit that extracts ontology classes corresponding to the extracted words as corresponding classes;
A class statistic database in which a class ID that is an identifier of a class defined on the ontology and a statistic indicating appearance frequency information in the document corpus of the class-corresponding word are associated and stored;
In a class hierarchy defined on the ontology, a reference class extraction unit that selects, as a reference class, a class having a high statistic from neighboring classes extracted by the corresponding class extraction unit;
A class abstraction degree calculation unit for calculating the abstraction degree of each of the reference class and the corresponding class;
A document abstraction level calculation unit that calculates a variance or average of abstraction levels of all corresponding classes extracted corresponding to each word in the input document as a document abstraction level;
By comparing the document abstraction level with the threshold value, the degree of variation in the abstraction level of the words included in the input document is determined. If the degree of variation is determined to be large, the word included in the input document is determined to correspond to the reference class. A document standardization processing unit for processing to replace words or presenting reference class information;
A document editing apparatus characterized by comprising: The corresponding class extraction unit includes:
The document editing apparatus according to claim 1, wherein an ontology class having a property corresponding to an extracted word from the input document is extracted as a corresponding class. The class statistics database is
2. The document editing apparatus according to claim 1, wherein a TF-IDF (Term Frequency-Inverse Document Frequency) indicating appearance frequency information in the document corpus is recorded as a statistic. The class statistics database is
TF-IDF (Term Frequency-Inverse Document Frequency), which is appearance frequency information in the document corpus, is calculated for each document creator, and the reciprocal of the calculated [TF-IDF] variance corresponding to a plurality of document creators is used as a statistic. The document editing apparatus according to claim 1, wherein the document editing apparatus has a recorded configuration. The reference class extraction unit includes:
In the class hierarchy defined on the ontology, the separation relation class is excluded from the neighboring classes within the preset number of hierarchies [n] starting from the corresponding class extracted by the corresponding class extracting unit, and the statistic is maximized. The document editing apparatus according to claim 1, wherein a process of selecting the class as a reference class is executed. The reference class extraction unit includes:
2. The document editing apparatus according to claim 1, wherein a process of extracting a class having a statistic that is larger than a statistic of the corresponding class by a predetermined value (p%) and having a maximum statistic as a reference class is executed. The class abstraction degree calculation unit
It is configured to perform a process of determining the abstraction level of the corresponding class in the vicinity of the reference class based on the distance of the hierarchy from the reference class on the ontology, and the abstraction degree of the lower class with the higher abstraction level of the upper class 2. The document editing apparatus according to claim 1, wherein an abstraction level is set to a low value. The class abstraction degree calculation unit
The configuration is such that the abstraction level of the corresponding class in the vicinity of the reference class is determined based on the difference from the number of properties of the reference class, and the number of properties is large with the high abstraction level of the class having a small number of properties. 2. The document editing apparatus according to claim 1, wherein an abstraction level is set so that the abstraction level of the class is a low value. A document editing method for executing an input document editing process in a document editing apparatus,
A word extracting unit for extracting a word included in the input document;
A corresponding class extraction step in which a corresponding class extraction unit extracts a class on the ontology corresponding to the extracted word as a corresponding class;
The reference class extraction unit selects, as a reference class, a class having a high statistic indicating the appearance frequency information in the document corpus from the neighborhood classes of the corresponding class extracted by the corresponding class extraction unit in the class hierarchy defined on the ontology. A base class extraction step;
A class abstraction calculating unit that calculates the abstraction of each of the reference class and the corresponding class;
A document abstraction level calculation unit that calculates a variance or average of the abstraction levels of all corresponding classes extracted corresponding to each word in the input document as a document abstraction level;
The document standardization processing unit determines the degree of variation in the abstraction level of the words included in the input document by comparing the document abstraction level and the threshold value. If the degree of variation is determined to be large, the word included in the input document A document standardization processing step of replacing a word with a word corresponding to the reference class, or presenting reference class information;
A document editing method characterized by comprising: The corresponding class extraction step includes:
The document editing method according to claim 9, wherein the ontology class having a property corresponding to the extracted word from the input document is extracted as a corresponding class. The reference class extracting step includes:
In the class hierarchy defined on the ontology, the separation relation class is excluded from the neighboring classes within the preset number of hierarchies [n] starting from the corresponding class extracted by the corresponding class extracting unit, and the statistic is maximized. The document editing method according to claim 9, wherein a process of selecting a class as a reference class is executed. The reference class extracting step includes:
The document editing method according to claim 9, wherein a process of extracting, as a reference class, a class that is larger than a statistic of the corresponding class by a predetermined value (p%) and has a maximum statistic. The class abstraction level calculating step includes:
It is configured to perform a process of determining the abstraction level of the corresponding class in the vicinity of the reference class based on the distance of the hierarchy from the reference class on the ontology, and the abstraction degree of the lower class with the higher abstraction level of the upper class The document editing method according to claim 9, wherein the abstraction level is set to a low value. The class abstraction level calculating step includes:
The configuration is such that the abstraction level of the corresponding class in the vicinity of the reference class is determined based on the difference from the number of properties of the reference class, and the number of properties is large with the high abstraction level of the class having a small number of properties. The document editing method according to claim 9, wherein the abstraction level is set so that the abstraction level of the class is a low value. In a document editing apparatus, a computer program for executing input document editing processing,
A word extraction step for causing the word extraction unit to extract words included in the input document;
A corresponding class extraction step for causing the corresponding class extraction unit to extract a class on the ontology corresponding to the extracted word as a corresponding class;
Let the reference class extraction unit select, as a reference class, a class having a high statistic indicating appearance frequency information in the document corpus from the neighborhood classes of the corresponding class extracted by the corresponding class extraction unit in the class hierarchy defined on the ontology. A base class extraction step;
A class abstraction calculating step for causing the class abstraction calculating unit to calculate the abstraction of each of the reference class and the corresponding class;
A document abstraction level calculation step for causing the document abstraction level calculation unit to calculate the variance or average of the abstraction levels of all corresponding classes extracted corresponding to each word in the input document as the document abstraction level;
The document standardization processing unit determines the degree of variation in the abstraction level of the word included in the input document by comparing the document abstraction level and the threshold value. If the degree of variation is determined to be large, the word included in the input document A document standardization processing step that replaces the word with a word corresponding to the reference class, or presents the reference class information,
A computer program characterized by comprising:

Images