JP3444223B2 - Database registration device - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a database registration device suitable for a database search device that enables efficient and highly accurate related data search for atypical data.

[0002]

2. Description of the Related Art Conventionally, a database that enables a search using semantic information of data is a keyword that is automatically searched from generally accumulated data such as a database of technical documents and a database of library collections. Was extracted, and the degree of association between data was calculated using a synonym dictionary or the like that was created empirically in advance.

On the other hand, in the case of non-language data such as images and voices, a method is known in which the feature quantity of the data is represented in the form of a vector and the degree of association between the data is represented by the distance between the two vectors. There is. Such methods include, for example, JP-A-58-147799 and JP-A-60-1895.
85, JP 64-17182 A, JP 4
There is a method in which the Euclidean distance between feature vectors of a voice pattern or a character pattern is used as the degree of similarity, as in Japanese Patent Publication No. -15777. In addition to this, for example, as in JP-A-60-202491 and JP-A-63-780, there is a method in which the sum of the absolute values of the differences between the components of the feature vector is used as the similarity. In either method, the feature vector is a series of physical quantities that characterize the audio data and the image data, and it is premised that all the components of the feature vector are determined in advance.

In recent years, hypertext has attracted attention as a data format that handles documents, images, sounds, and the like in a unified manner. Hypertext has an information unit called a node composed of a document, an image, a sound, etc., and the nodes can be associated with each other by a link. In the field of databases, the data expressed in hypertext is used as a new medium without distinction between data types such as documents, images, and sounds.

Here, there is a demand for a method of automatically adding a link from a new node to a related node or a link from a related node to a new node when adding a new node to the hypertext system. In such a method, when the content of the node is changed, the difference information or all the information of the node is held as a node of a new version, as in Japanese Patent Laid-Open No. 3-278247, and a new node is added to the node before the change. There is a thing that a link is added from. Further, as in Japanese Patent Laid-Open No. 4-317172, in addition to a node having normal information (search target node), a node used for search (search node) is prepared and linked with a corresponding search target node. Incidentally, there is also a method in which when the search target node is changed, the link of the search node that needs to be changed is updated or added.
Finally, as in JP-A-5-20363, qualitative search conditions are described in advance in each piece of information, and the search conditions are displayed at the same time when the information is displayed, and other search conditions are selected based on the search conditions selected there. There is also a method in which information is fuzzy searched.
Note that the qualitative search condition here is a condition that can be determined by a physical quantity such as image contrast, such as the darkness of the image. Further, in the hypertext database, conventionally, regarding the display of the search result, generally, a list of links to nodes matching the search expression is displayed.

[0006]

The first problem is as follows.
Regarding the registration of data, in the related art in which a keyword is automatically extracted from the data or a keyword is freely registered by the registrant in order to define the meaning of the data, user characteristics, task characteristics, such as user interface data, If it is necessary to characterize from various viewpoints such as system characteristics, and if the appropriate vocabulary as a keyword is not clear, it is difficult to properly define a keyword that represents the meaning of the data, and therefore appropriate characterization of the data There were problems such as being difficult.

A second problem is that in the conventional method of automatically creating a link using the qualitative information added to each node when registering a new node in the hypertext database, (1 ) Problems such as not being able to deal with atypical data that allows arbitrary combinations of documents, images, sounds, etc. (2) The direction of links cannot be dynamically determined for node pairs linked to each other was there.

Therefore, in the present invention, atypical data in which documents, images, sounds, etc. are freely combined can be characterized by a single method, and data to be newly registered has already been stored. An object is to provide a database registration device that automatically links with related data.

[0009]

The database registration device of the present invention is characterized by a content registration means for a data registrant to input the contents of new data and a characteristic of the new data from a predetermined keyword set. Keyword selecting means for the data registrant to select an appropriate keyword to represent, keyword set saving means for saving the keyword set used by the keyword selecting means, and each keyword selected by the keyword selecting means. On the other hand, from the weighting input means for the data registrant to input the weighting with a real number of [0, 1], the keyword selected by the keyword selecting means, and the weighting input by the weighting input means. Feature vector generation means for generating a feature vector of new data, data storage means for storing data, The data storage in which the distance between the feature vector of the new data generated by the feature vector generation means and the feature vector added to each data stored in the data storage means is equal to or less than a predetermined threshold value. Position group extraction means for obtaining the data stored in the means as neighborhood data for the new data, and a predetermined link from the category of the neighborhood data obtained by the location group extraction means and the category of the new data By using a rule, link direction determining means for determining the direction of a link to be newly added between the neighboring data and the new data, and the link direction determining means between the neighboring data and the new data. A link is added in the determined direction, and the new data and the feature vector alignment of the new data are added. Information about the link and a node addition means for storing in said data storage means. Further, the position group extracting means stores an inter-component correlation table storing means for storing a table of the correlation degree between the components of the feature vector, the feature vector of the new data generated by the feature vector generating means, and the data storage. The undefined components of the feature vector added to each data stored in the means are defined by the components defined by these vectors and the correlation table between the components stored in the inter-component correlation table storage means. An undefined component complementing means for complementing by using the distance between the feature vector of the new data whose undefined portion is complemented by the undefined component complementing means and the feature vector of the data stored in the data accumulating means. If the inter-vector distance calculation means to be obtained and the distance between the feature vectors calculated by the inter-vector distance calculation means are less than or equal to a predetermined threshold value, the data is And the neighboring data determination means determines that the proximity data to serial new data,
And a threshold value storage means for storing a predetermined threshold value used in the determination of the proximity data in the proximity data determination means.

In the database registration device having such a configuration, it is possible to assign an appropriate keyword to atypical data composed of an arbitrary combination of text, image, voice, etc. by a single method. In addition, it is possible to automatically link to another node that is semantically related in newly creating a hypertext node.

The undefined component complementing means is necessary for obtaining the complementary value of the undefined component of the feature vector of the feature vector generated by the feature vector generating means and the data stored in advance. If there is no defined component that correlates with, the undefined component is replaced by a special symbol that will always be 0 regardless of the difference with any value, so the undefined portion can be complemented as described above. Even if there is not, the degree of association can be measured by obtaining the distance between the vectors.

[0012]

BEST MODE FOR CARRYING OUT THE INVENTION Next, an example of an embodiment of the present invention will be described in detail with reference to the drawings. Before that, a device for retrieving registered data using a database registration device of the present invention The (database search device) will be described.

Referring to FIG. 1, an example of a database search apparatus includes a query input means 1 for a user to input a desired data feature, and a query vector generation means 2 for converting an input query into a feature vector. , A data storage means 3 for storing data and data feature vectors
An undefined component complementing means 4 for complementing the undefined component included in the feature vector, and an inter-component correlation table storage means 5 storing an inter-component correlation table used by the undefined component complementing means 4.
An inter-vector distance calculation means 6 for obtaining the distance between the feature vectors of the inquiry and the saved data, and a neighborhood data determination means for determining the saved data having the feature vector at a distance less than a certain threshold with respect to the feature vector of the inquiry 7, a threshold value storage means 8 for storing the threshold value thereof, a search result display means 9 for displaying a list of the neighborhood data obtained by the neighborhood data determination means 7, and a search result display means 9. It includes a data designating means 10 for allowing the user to designate arbitrary data from the list of the neighborhood data that has been designated, and a data display means 11 for displaying the content of the designated data. Here, undefined component complementing means 4, inter-component correlation table storage means 5,
The position group extraction unit 12 is configured by the vector distance calculation unit 6, the neighborhood data determination unit 7, and the threshold value storage unit 8.

Inquiry input means 1 and data instruction means 1
Reference numeral 0 is an information input device such as a mouse and keyboard, and the search result display means 9 and data display means 11 are information display devices such as a VDT. Further, the data storage means 3 is composed of an external storage device such as a fixed disk, and the inquiry vector generation means 2 and the position group extraction unit 12 are programs operating on, for example, a PC (personal computer) or WS (workstation). Composed of.

The data stored in the data storage means 3 is
It is composed of data having 0 or more relations to other data, which is an arbitrary combination of documents, images, sounds, and the like, and a feature vector corresponding to the data. For example, data A
And data B, and each feature vector is v
[A], v [B], and when one of the relations of the data A has a relation to the data B, the data storage means 3
Then, it is saved in the form as shown in FIG.

Next, the operation of the database search device configured as shown in FIG. 1 will be described with reference to the drawings.

Referring to FIG. 1, it is assumed that the inquiry input means 1 selects a keyword from the user using, for example, a keyboard or a mouse, and weights each keyword with a real number of [0, 1]. For example, in FIG.
As shown in (a), it is assumed that a keyword set representing a characteristic of data such as a design example of a user interface of an arbitrary system is previously defined, and the user now selects the keyword shown in FIG. 3 (b). In order from the top, "Unselected, 0.8, 0.2, Unselected, 0.8,
0.7, unselected, 0.1, 0.2 ”are weighted. At this time, the query vector generation means 2 generates a feature vector (query vector) corresponding to the input query as follows. (X [1], 0.8, 0.2, x [2], 0.8, 0.7, x [3], 0.1, 0.2) (a) where x [1 ], X [2], x [3] represent undefined components.

Next, with respect to each of the above-mentioned inquiry vector and the feature vector of each data in the data storage means 3 to be searched, when an undefined component is included, the undefined component complementing means 4 causes the feature vector of those data. The complementary value of the undefined component is determined based on the defined component in (1) and the correlation table between the attribute values stored in advance by the component correlation table storage unit 5.

FIG. 4 shows an example of a correlation table between attribute values stored in the inter-component correlation table storage means 5. In this table, attribute values are listed in the row direction and the column direction, and the presence or absence of correlation and the degree of correlation are described at the intersections of rows and columns. When investigating another attribute value having a correlation with a certain attribute value α, the attribute value α is selected from the attribute value group listed above, and the value written at the intersection of the column and each row is examined. If the numerical value p is described, there is a correlation with the attribute value described on the left side of the line, and the degree of correlation is p, and if "-", there is no correlation. For example, when investigating another attribute value having a correlation with the attribute value "for beginners", the value of the intersection between the column of the attribute value "for beginners" and each row in the attribute value group listed at the top of the table is examined. In the case of the figure, the attribute value "For infants",
There is a correlation with "for the general public" and "for the elderly", and the correlations are 0.9, 0.3 and 0.8, respectively.

In the case of the inquiry vector (a), the undefined component complementing means 4 includes three undefined components x [1], x [2], and x [3]. Performs various processing.

Looking at the correlation table of FIG. 4, there are three correlations with the attribute value "for beginners" corresponding to x [2]: "for infants", "for general", and "for elderly". Although it is an attribute value, the attribute value for "infant" in query vector (a) is excluded because it is an undefined component x [1], and the correlation degree is the highest among the remaining "general" and "elderly". “For senior citizens”, which has a high rating, is selected. The meaning of this correlation value indicates that "an elderly person is a beginner with considerable certainty." Looking at the correlation table, the degree of correlation is 0.8, and the query vector (a) is "for the elderly".
Since the weight assigned to x is 0.2, x [2] = 0.2 · 0.8 = 0.16 is complemented.

Similarly, the "visually impaired" corresponding to x [3] has a degree of correlation of 0.7 only with "for the elderly" in the correlation table of FIG. 4, and the inquiry vector (a ), The weighting of the attribute value “for the elderly” is 0.2, so x [3] = 0.2 · 0.7 = 0.14 is complemented.

On the other hand, the attribute value "for young children" corresponding to x [1] has a correlation only with the attribute value "for beginners" in the correlation table of FIG. "To" is the undefined component x [2]. Therefore,
The complementary value cannot be calculated. In such cases,
Assign ⊥. ⊥ is the distance between vectors in the latter stage,
Indicates that the component is not used in the distance calculation. That is,
When at least one of the two vector components for which a difference is to be obtained is ⊥, the difference is always set to 0. By doing so, the undefined component that could not be complemented does not affect the distance calculation in the subsequent stage, and therefore the inter-vector distance can still be used as an index of the semantic relevance of the two data. As a result of the above complement, the inquiry vector (a) becomes as follows. (⊥, 0.8, 0.2, 0.16, 0.8, 0.7, 0.14, 0.1, 0.2) (b)

In the correlation table of FIG. 4, each attribute value has a correlation with at least one other attribute value, but there is an attribute value that has no correlation with another attribute value. It is also conceivable that if such an attribute value appears as an undefined component, the complementary value cannot be obtained as before. In such cases, assign ⊥.

After that, the inter-vector distance calculating means 6 calculates the distance between the query vector (b) in which the undefined component is complemented as described above and the feature vector of each data. There are two possible ways to define this distance: (1) the sum of the absolute values of the differences for each component, or (2) the Euclidean distance. An index of the degree of semantic relevance between the query and each piece of data to be searched. Both have the same effect. Therefore, here, the distance is defined as the sum of absolute values of the differences between the components. Then, the neighborhood data determination means 7
Then, the threshold value stored in the threshold value storage means 8 is compared with the vector distance, and if the vector distance is less than or equal to the threshold value, the search target data is added to the position group Gp.

The data set in the position group Gp thus obtained is the search result for the inquiry from the user, and the list of the data contained therein and the relationship between the data is displayed by the search result display means 9 such as a display device. . In the case of hypertext, the list includes the title of each data. When the user looks at such a list and uses the data designating means 10 to designate any data in the list, the data display means 11 displays the detailed contents of the designated data.

As described above, it becomes possible to perform intuitive input without using a search expression in the user's inquiry input, and efficient and highly accurate relevance calculation using the feature vector becomes possible. Even if there are undefined components in it, it becomes possible to calculate the distance between feature vectors without impairing the meaning of the degree of semantic association between data.

Next, another example of the database search device will be described with reference to the drawings.

Referring to FIG. 5, another example of the database search device is obtained by the position group extraction unit 12 when the relation between data is defined in advance and the data is classified into several categories. Using the predetermined inter-category rule stored in the inter-category rule storage means 14 from the category to which each of the neighborhood data belongs and the category to which the data that can be reached by following the relation from the neighborhood data belong, The database search apparatus differs from the database search apparatus of the example shown in FIG. 1 in that it includes a relational data determination unit 13 that determines whether or not the search result includes data that can be reached by tracing the relationship from the neighborhood data. The relational data determination unit 13 and the inter-category rule storage unit 14 constitute a relational group extraction unit 15. Relationship group extraction unit 1
5 is implemented by a program that operates on a PC or WS, for example.

Further, the data stored in the data storage means 3 includes data having 0 or more relations to other data in which documents, images, voices, etc. are arbitrarily combined, feature vectors corresponding to the data, and It is composed of the category to which the data belongs. For example, there are data A and data B, their feature vectors are v [A] and v [B], their categories are C1 and C2, and the data A is
When one of the relations has a relation to the data B, the data is stored in the data storing means 3 as shown in FIG.

Next, the operation of the database search device shown in FIG. 5 will be described in detail with reference to the drawing, focusing on the differences from the database search device shown in FIG.

Referring to FIG. 5, the relational data judging means 1
3 is the position group G obtained by the neighborhood data determination means 7.
With respect to the arbitrary data A belonging to p, the relation that can be traced from the data is traced, and the arbitrary data B at the end is obtained. Then, using the categories to which the data A and the data B belong and the predetermined inter-category rule stored in the inter-category rule storage means 14, it is determined whether or not the data B is included in the data set Gr. For example, three categories, “guideline”, “design case”, and “evaluation case”, are preset,
It is assumed that Gp includes data A belonging to “design case”. The data A has a relationship with the data B, the data C, and the data D, and the categories of the respective data are the data B is “guideline”, the data C is “design case”, and the data D is “evaluation case”. And It is also assumed that the inter-category rule storage means 14 stores inter-category rules as shown in FIG. At this time, referring to FIG. 7, in the relational data determination means 13, first, Gr is G
The inter-category rule is referred to for the data A that is initialized to r = Gp and belongs to Gp. In the inter-category rules,
Since it is defined that both the relationship from the “design case” to the “guideline” and the relationship from the “design case” to the “evaluation case” are included in Gr, the data B and the data D are added to Gr. On the other hand, since the relationship from the “design case” to the “design case” is not included in Gr, the data C is not added to Gr. Therefore, for data A, data B and data D are newly added to Gr. Such processing is performed for all data belonging to Gp.

The data set Gr thus obtained is the search result for the inquiry from the user, and the search result display means 9 displays a list of the data contained therein and the relationship between the data. When the user uses the data instructing means 10 to instruct arbitrary data, the content of the instructed data is displayed by the data displaying means 11.

As a result, the user can obtain not only the degree of relevance to the inquiry but also what kind of data from the inquiry when judging whether each of the data included in the search result is the data desired by the user. It is possible to make a multifaceted decision by using together information on whether or not

Next, an embodiment of the database registration device according to the present invention will be described with reference to the drawings.

Referring to FIG. 9, the configuration of the embodiment of the database registration apparatus according to the present invention is a content registration means 1 in which a data registrant registers the content of data (node).
6, a keyword set storage means 17 in which a predetermined keyword set is stored, a keyword selection means 18 for the data registrant to select a keyword that appropriately represents the content of the node from the keyword set, and a data registrant. Is a weighting input means 19 for weighting each selected keyword by using a real number of [0, 1], a characteristic vector generation means 20 for generating a characteristic vector from a set of weighted selected keywords, and data. A data storage unit 21 that stores (nodes), a position group extraction unit 12 that obtains a set of related nodes, and a link direction for determining a link direction when adding a link between a registered node and a neighboring node. The determination means 22, the linking rule storage means 23 in which the rules used for the determination are stored, and the registered node are the data storage hands. In addition to 21 and a node addition section 24 for adding a link between the link direction determination unit direction registration node determined at 22 and neighboring nodes.

Here, the data stored in the data storage means 21 is the same as that shown in FIG. 6 described above, and the relation to other data, which is an arbitrary combination of documents, images, voices, etc., is 0. It is composed of the above-mentioned data, the feature vector corresponding to the data, and the category to which the data belongs.

The internal structure of the position group extracting unit 12 is the same as that of the position group extracting unit 12 forming the main part of the database searching apparatus shown in FIG.
As shown in FIG. 3, the correlation table between the components of the feature vector is stored in the component correlation table storage means 5, the feature vector of the new node generated by the feature vector generation means 20, and the data storage means 21. The undefined components of the feature vector added to each node are complemented by using the components defined by these vectors and the correlation table between the components stored in the inter-component correlation table storage means 5. Inter-vector distance for obtaining the distance between the undefined component complementing means 4 and the feature vector of the new node whose undefined portion is complemented by the undefined component complementing means 4 and the feature vector of the node stored in the data accumulating means 21. If the distance between the calculation means 6 and the feature vector calculated by the inter-vector distance calculation means 6 is less than or equal to a predetermined threshold value, the node is regarded as a neighbor node for the new node. Data and determines proximity data determination unit 7
And a threshold value storage means 8 for storing a predetermined threshold value used in the proximity data determination means 7 for the determination of the proximity data. Further, the undefined component complementing means 4 correlates with the undefined component necessary for obtaining the complementary value of the undefined component of the feature vector generated by the feature vector generating means 20 and the feature vector of the data stored in advance. If there is no defined component with, then the undefined component is replaced with a special symbol that is always 0, no matter what the difference.

The content registration means 1 shown in FIG.
6, keyword selection means 18, weighting input means 19
Is an information input device such as a mouse and a keyboard, and the keyword set storage means 17, the feature vector generation means 20, the position group extraction part 12, the link direction determination means 22, the linking rule storage means 23 and the node addition means 24 are For example, it is realized by a program that operates on a PC or WS. The data storage means 21 is an external storage device such as a fixed disk.

The operation of the database registration device having such a configuration will be described with reference to the drawings.

Referring to FIG. 9, first, the registrant creates the contents of the registered node in advance by arbitrarily combining a document, an image, and a voice, and inputs them by the content registration means 16. Further, the category of this registered node (guideline, design case, evaluation case, etc., which will be described later) is input by the content registration means 16. And the keyword selection means 1
8, an appropriate keyword is selected from the keyword set stored in advance in the keyword set storage unit 17 according to the content of the node input by the content registration unit 16. Therefore, some keywords are not selected. Here, the keyword set may be classified into some categories so that the user can easily understand it. For example, referring to FIG. 11, keyword sets (attribute values) such as “general”, “beginner”, and “visual” are
It is classified into four categories (attributes) of "age group", "system experience", "task knowledge", and "disability". Then, the registrant assigns a real number of [0, 1] to each selected keyword by the weighting input means 19 as the importance of the keyword. At this time, it may be allowed to assign a real number to the category. In this case, the real number assigned to the category is assigned to all the keywords belonging to the category. By doing so, keywords are collectively selected, and the user's time and effort at the time of inputting weights are reduced. It is also possible to assign a real number to each keyword and a real number to a category. In this case, for all the keywords belonging to the category given the real number, the product of the real number assigned to each keyword and the real number assigned to the category is calculated, and the value is reassigned to the keyword. By doing so, it is possible to reduce the time and effort required to adjust the weight when repeatedly performing the search.

The characteristic vector generating means 20 generates a characteristic vector corresponding to the registered node from the results obtained up to the preceding stage. At that time, a special symbol x indicating undefined is assigned to the keyword not selected by the keyword selecting means 18. As such x, for example, -1
A real number other than [0,1] may be used, or a character such as “x” may be used. A vector in which the real numbers assigned to the keywords are arranged in a certain order as described above is the feature vector corresponding to the registered node. For example, “general”, “beginner”, “expert”, “poor”, “normal”, “abundant”,
When eight keywords of “visual” and “limb” are selected by the keyword selecting means 18, and subsequently, the importance is set by the weighting input means 19 as shown in FIG.
The feature vector corresponding to the registration data is (x, 0.8, x, 0.9, x, 0.7, 0.3, 0.8, 0.7,
0.7, x, 0.1) (x is an undefined part).

Next, the position group extraction unit 12 calculates the distance between the feature vector of the registered node determined by the feature vector generation unit 20 and the feature vector of each node stored in the data storage unit 21, A set Gp of nodes located at a distance equal to or less than a predetermined threshold value is obtained.
That is, referring to FIG. 10, the undefined component complementing means 4 is
The feature vector of the new node generated by the feature vector generation means 20 and the undefined component of the feature vector added to each node stored in the data storage means 21 are defined by these vectors and between the components. The correlation table storage means 5 is used for complementing the correlation table between the components, and the inter-vector distance calculation means 6 is the feature vector of the new node in which the undefined portion is complemented and the data storage means 2.
If the distance between the feature vectors calculated by the inter-vector distance calculation means 6 is less than or equal to a predetermined threshold, the neighborhood data determination means 7 obtains the distance from the feature vector of the node stored in 1. The node is determined to be neighborhood data for the new node and included in the set Gp.

Next, the link direction determining means 22 determines, for each node of Gp obtained by the position group extracting section 12,
Determine the link direction for the registered node. At that time, the linking rule stored in advance in the linking rule storage unit 23 of FIG. 9 is applied. The linking rule is determined based on the category of the node. For example, the linking rule as shown in FIG. 13 is applied to the three categories of “guideline”, “design case”, and “evaluation case”.
Now, the registered node belongs to the "guideline", and the node A,
If B and C belong to the “guideline”, the “design case”, and the “evaluation case”, respectively, the link directions between the registered node and the nodes A, B, and C according to the linking rule of FIG. The determination is made as shown in FIG.

Finally, the node adding means 24 adds a link to (from) the registered node, its feature vector and category, and each node belonging to the position group Gp in the direction determined by the link direction determining means 22. Then, the data is stored in the node storage means 21.

As described above, at the time of registering new data, the link direction to be added between the new data and the neighbor data is determined by the category information to which the new data and the neighbor data belong and the linking rule using the category, Since a link is automatically added in that direction and new data and its link information are stored, it is possible to reduce the time and effort required for searching for a related node and adding a link when creating a new node of hypertext.

[0047]

As described above, according to the present invention, the following effects can be obtained.

According to the invention described in claims 1 and 2,
At the time of data registration, a keyword is selected by the user from a predetermined keyword set for the input data, and a feature vector of the input data is generated from the result of weighting each selected keyword. Therefore, it is possible to characterize atypical data composed of an arbitrary combination of documents, images, voices, etc. by a single method that is easy for the data registrant to understand, and reduce the time and effort required to assign appropriate keywords. .
Further, by the category information to which the new data and the neighborhood data belong and the linking rule using the category, the link direction to be added between the new data and the neighborhood data is determined,
Since a link is automatically added in that direction and new data and its link information are stored, it is possible to reduce the time and effort required for searching for a related node and adding a link when creating a new node of hypertext.

According to the third aspect of the invention, when there is no defined component necessary for obtaining the complementary value of a certain undefined portion of the feature vector, the difference with respect to an arbitrary value becomes 0. Since the undefined part is replaced with such a special symbol, the part where the undefined part exists is ignored in the vector distance calculating means in the subsequent stage. Therefore,
Even if there are many undefined parts in the feature vector, efficient relevance calculation using the feature vector can be performed.

[Brief description of drawings]

FIG. 1 is a block diagram showing an example of a database search device.

2 is an explanatory diagram of data stored in a data storage unit in FIG. 1. FIG.

FIG. 3 is a diagram showing an example of a keyword set presented to a user at the time of inputting an inquiry and an example of a keyword actually selected by the user.

FIG. 4 is a diagram showing an example of an inter-component correlation table.

FIG. 5 is a block diagram showing another example of a database search device.

6 is an explanatory diagram of data stored in a data storage unit in FIG.

FIG. 7 is an operation explanatory diagram of a relational data determination unit.

FIG. 8 is a diagram showing an example of rules between categories.

FIG. 9 is a block diagram of an embodiment of a database registration device according to the present invention.

FIG. 10 is a block diagram showing a configuration example of a position group extraction unit in the database registration device.

FIG. 11 is a diagram showing an example of a keyword set (attribute value) and an example of categories stored in a keyboard set storage means.

FIG. 12 is a diagram showing an example of a keyword (attribute value) selected by a data registrant for a new node when registering a new node and its importance (weighting).

FIG. 13 is a diagram showing an example of a linking rule.

FIG. 14 is a diagram showing an example of a link given to a registered new node.

[Explanation of symbols]

1 Inquiry input method 2 Query vector generation means 3 data storage means 4 Complementary means for undefined components 5 component correlation table storage means 6 Vector distance calculation means 7 Neighborhood data determination means 8 Threshold storage means 9 Search result display means 10 Data instruction means 11 Data display means 12 Position group extraction unit 13 Relational data determination means 14 Category preservation rules 15 Relation group extraction unit 16 Content registration means 17 Keyword set storage means 18 Keyword selection means 19 Weighting input means 20 Feature vector generation means 21 Data storage means 22 Link direction determination means 23 Linking rule storage means 24 node addition method

Continuation of front page (58) Fields surveyed (Int.Cl. ⁷ , DB name) G06F 17/30 G06F 12/00 JISST file (JOIS)

Claims (3)

(57) [Claims] 1. A content registration means for a data registrant to input the contents of new data, and a data registrant selects a keyword suitable for expressing the characteristics of the new data from a predetermined keyword set. For selecting a keyword set, a keyword set storage unit for storing a keyword set used by the keyword selection unit, and a weighting with a real number of [0, 1] for each keyword selected by the keyword selection unit. A weighting input means for the data registrant to input, and a feature vector generating means for generating a feature vector of new data from the keyword selected by the keyword selecting means and the weighting input by the weighting input means. A data storage means for storing data, and a new data generated by the feature vector generation means. Of the data stored in the data storage means in which the distance between the feature vector of the data and the feature vector added to each data stored in the data storage means is equal to or less than a predetermined threshold value. From the position group extracting means to be obtained as neighborhood data with respect to the data, the category of the neighborhood data obtained by the location group extracting means and the category of the new data, by using a predetermined linking rule, the neighborhood data and the Link direction determining means for determining the direction of the link to be newly added between the new data, and add a link in the direction determined by the link direction determining means between the neighborhood data and the new data, Information about the new data, the feature vector of the new data, and the link is stored in the data storage. Database registration apparatus characterized by comprising a node addition means for storing the unit. 2. A content registration means for a data registrant to input the contents of new data, and a data registrant selects a keyword suitable for expressing the characteristics of the new data from a predetermined keyword set. For selecting a keyword set, a keyword set storage unit for storing a keyword set used by the keyword selection unit, and a weighting with a real number of [0, 1] for each keyword selected by the keyword selection unit. A weighting input means for the data registrant to input, and a feature vector generating means for generating a feature vector of new data from the keyword selected by the keyword selecting means and the weighting input by the weighting input means. And a table of the degree of correlation between the data storage means for storing data and the feature vector components. Between the component correlation table storage means, the feature vector of the new data generated by the feature vector generation means, and the undefined component of the feature vector added to each data stored in the data storage means. Undefined component complementing means for complementing by using a component defined by a vector and the correlation table between the components stored in the inter-component correlation table storing means, and an undefined portion is complemented by the undefined component complementing means An inter-vector distance calculating means for obtaining a distance between the feature vector of the new data and the feature vector of the data stored in the data accumulating means, and a distance between the feature vectors calculated by the inter-vector distance calculating means Is less than or equal to a predetermined threshold value, the data is neighborhood data determination means for determining the neighborhood data for the new data, and the neighborhood data determination means. Threshold value storing means for storing a predetermined threshold value used in the determination of the neighborhood data in the data determining means, the neighborhood data category determined by the neighborhood data determining means, and the new data category. , Link direction determining means for determining a direction of a link to be newly added between the neighborhood data and the new data by using a predetermined linking rule, and between the neighborhood data and the new data A node adding unit that adds a link in the direction determined by the link direction determining unit, and stores the new data, the feature vector of the new data, and information about the link in the data storing unit. Database registration device. 3. The undefined component complementing means, which is necessary for obtaining a complementary value of the undefined component of the feature vector generated by the feature vector generating means and the feature vector of the data stored in advance, the undefined component 3. When there is no defined component having a correlation with the component, the undefined component is replaced with a special symbol such that the difference with any value is always 0. 3. Database registration device.