JPH11120180A - Data retrieval device and storage medium recording data retrieval program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To effectively rearrange retrieved data on a two-dimensional plane at a high speed by transforming the attribute value of data to a two-dimensional coordinate and displaying the data on the two-dimensional plane of a display device, based on the transformed coordinate value. SOLUTION: The attribute values of the respective data are two-dimensionally transformed in a coordinate transformation means 5 by the algorithm of a self-organizing map and the transformed data are displayed on the two-dimensional plane by a two-dimensional plane display means 6. The distances of a specified point and the transformed two-dimensional coordinates of the respective data are calculated by a retrieval means 4, an index corresponding to the data for which the value is less than a prescribed value is taken out from a data base and sequentially stored in a storage device 7 and the number of the retrieved data is stored further. Thus, the retrieved data are effectively rearranged on the two-dimensional plane at a high speed, and even when the retrieved data are many, the required ones are easily found from them.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a data retrieval apparatus for retrieving data such as documents and images stored in a storage device such as a CD-ROM or a hard disk, and to a technique for visualizing the retrieved data.

[0002]

2. Description of the Related Art In retrieval of documents, images, and the like, conventional retrieval methods mainly include a method based on keywords and a method based on data attributes. The keyword-based method compares a keyword input by a user with a keyword group added to data in advance, and outputs matched data as a search result. This method is useful if the user does not know the keyword attached to the data.
It takes many trials and errors to get the desired data. Also, it is not suitable for ambiguous search, for example, where a designer searches for a photograph that matches his image.

A method based on data attributes is as follows. First, each data has a plurality of attributes and a value for the attribute (attribute value).
Is given. Specifically, for example, in a photo database that searches for photos based on images, each attribute is expressed as a sensibility word such as "elegance", "power", or "emotional", and the attribute value is Is given as a degree to the phrase. Considering a multidimensional space in which each attribute is regarded as a coordinate axis, each data is arranged in the multidimensional space using its attribute value as a coordinate value. Data search is performed in the form of a neighborhood search in a multidimensional space, and is classified into several methods depending on how search conditions are given.

A first search method is to specify the values or ranges of some attributes as numerical values as search conditions and output data satisfying the conditions. A second search method is to display the above-described multidimensional space on a screen, specify a point in the multidimensional space as a search condition, and output data located near the point. To specify a point in multidimensional space,
Usually, a pointing device such as a mouse is used. The third search method is to provide a keyword as a search condition and output data close to the keyword. The keyword given here is given an attribute value for each attribute in advance. Therefore, the keywords are arranged in a multidimensional space, and usually, data located within a predetermined distance from the position of the keyword in the multidimensional space is output as a search result. The keyword here is not limited to text, but includes, for example, a method of giving an image and searching for an image similar to the image.

In the second search method, since points on the screen are represented in two dimensions, when displaying a multidimensional space on the screen, the multidimensional coordinates are actually converted to two-dimensional coordinates. To display. Data that has been converted to two-dimensional coordinates is usually 2
It is displayed at a corresponding screen position in the dimensional space with some symbol indicating the existence of data. This allows the user to see the distribution of the data at a glance. Usually, a circle or a square is used as a symbol, but it is not clear from which alone what kind of data is distributed in which part of the two-dimensional space.
Therefore, for example, for some representative data given in advance, by displaying the data itself or a part of the data at the corresponding screen position,
The clarification of the data structure of a two-dimensional space is also performed. In this case, at the time of the search, the user refers to the displayed representative data, estimates the position of the target data in the two-dimensional space, and determines the position to be pointed.

As a method of coordinate transformation, there is a method of allowing a user to select two coordinate axes and projecting the plane on a plane formed by the two coordinate axes, or a method of projecting the plane on a plane obtained by principal component analysis of data. As an example of the former, FIG. 10 shows data obtained by projecting data consisting of three-dimensional attribute values shown in FIG. 9 onto a plane consisting of two coordinate axes. 10A is a projection on the xy plane, FIG. 10B is a projection on the yz plane, and FIG. 10C is a projection on the xz plane.
In this method, two of the attribute values expressed in multi-dimensions are used, and the rest are ignored, so that data arrangement reflecting the distribution of data in the multi-dimensional space is impossible.

[0007] Principal component analysis is a method of statistically treating multivariate data. In this method, by finding an axis direction in which the variance of the data is maximum, and projecting the data along the axis, coordinate transformation can be performed so as to reflect the distribution of the data to the maximum. Therefore, it is very effective especially when multidimensional data is distributed in a plane. However, when the multi-dimensional data is distributed on a curved surface, or when the data is distributed with a spatial spread in three or more dimensions, the two-dimensional coordinate conversion in the principal component analysis is insufficient.

As another method of coordinate transformation, a paper published by the Information Processing Society of Japan, DBS99-22, "Abstracting mechanism of database using self-organizing map and vocabulary search", describes a nonlinear method based on a neural network. Is a transformation,
T. A method of arranging multidimensional data in two dimensions using a self-organizing map proposed by Kohonen is disclosed.

The method will be described briefly. First, the self-organizing map is an unsupervised competitive learning model without a hidden layer, which finds its topological structure from input data and usually expresses it in a two-dimensional space. Each unit of the output layer is
The units are topologically arranged in a two-dimensional space, and each unit has a vector of the same dimension as the input data. The learning is performed by bringing a unit having the closest vector to each input data and a unit in the vicinity thereof close to the input data. Here, the above-described multidimensional data is sequentially given as input data. After a certain number of learnings, each data is mapped to a unit having the closest vector, so that the data is arranged in a two-dimensional space.

When searching for the target data from the searched data, it is usual to check the data while checking the data one by one. However, if the structure of the searched data is visualized and displayed, the target data can be quickly found. The technique disclosed in Japanese Patent Application Laid-Open No. 1-195578 is a method of searching based on data attributes. However, displaying representative data corresponding to several coordinate axes based on the searched data attributes is disclosed. Is to visualize the structure of the subspace to which the retrieved data belongs. As a method of selecting the coordinate axes for displaying the representative data, the degree of dispersion of the search data on each axis is obtained, and a method having a large dispersion is selected.

Japanese Patent Application Laid-Open No. 8-339375 discloses that data is searched based on a distance from a keyword.
After displaying the searched data on the two-dimensional plane as a display on the two-dimensional plane, if the user performs an operation of moving away unnecessary data and bringing necessary data close by using a pointing device, another operation is performed in conjunction with the operation. There is disclosed a method for efficiently narrowing down data by simultaneously moving data. However, the method of displaying the search data on the two-dimensional plane here is to simply obtain the distance between the attribute of the data and the attribute of the keyword and display the data at a position proportional to the distance from the center. Is not taken into account.

In other words, at the stage of displaying the retrieved data in two dimensions, data close to the keyword is simply arranged in order from the center, and data with close attributes are displayed apart from each other, or data with far attributes are displayed separately. May be displayed nearby. Another method of statistically treating multivariate data is Quantification Class 4. The fourth type of quantification is a method of converting qualitative data given a mutual correlation into a quantity vector so as to maintain the correlation.

[0013]

In the conventional search technology,
The data obtained as a search result is simply displayed in the order of the data, or is displayed in the order of the score with some score. The user searches for necessary data by looking at the data in order. However, with the recent expansion of computerization and the increase in the size of databases, the data obtained by searching may include various viewpoints and may be numerous, and it takes time to find necessary data from the data. For example, in the search method based on the self-organizing map described above, when a large number of data are mapped to one unit, when the user designates that unit, a large number of output data must be viewed in order. It takes time and effort to find the desired data.

According to the present invention, there is provided a data search apparatus for effectively arranging a large number of data output as search results on a two-dimensional plane when the number of data is large. As a result, the structure of the searched data becomes clear, and the target data can be quickly found rather than simply arranging the searched data in order.

In the technique disclosed in Japanese Patent Application Laid-Open No. 1-195578, since a method of simply selecting a coordinate axis having a high degree of dispersion from the coordinate axes corresponding to the respective attributes, a method which reflects the data distribution well is used. A dimensional plane is not selected. Here, we also provide user assistance to determine the browsing direction to get the desired data,
The representative data corresponding to each direction is displayed, and the retrieved data is not displayed on a two-dimensional plane.

In the technique disclosed in the above-mentioned Japanese Patent Application Laid-Open No. 8-339375, searched data is arranged near the center of a two-dimensional plane in ascending order of keywords. Not considered. Therefore, in this display method, only the degree of closeness between the searched data and the keyword is shown, and the display is not arranged so that the entire structure of the searched data can be understood. The above-mentioned “Database Abstraction Mechanism Using Self-Organizing Map and Vocabulary Search” also describes a method of reconstructing a partial area of the map on a plane, which is called “map refinement”. Here, the area is detailed by applying the self-organizing map again in the selected area.

The conversion into a two-dimensional plane according to the present invention is performed every time a search is performed, and therefore, high speed is required. In the self-organizing map, it is necessary to perform learning a certain number of times in order to accurately reflect the data structure, and therefore high speed cannot be expected. Therefore, here, the self-organizing map is not used, and the principal component analysis or the quantification class 4 is used. In the method using the quantification class 4, the correlation between the data is given by the direction cosine of the angle between the vectors when each data is regarded as a multidimensional vector. Thereby, each data is regarded as qualitative data and quantified in two dimensions. Also in the quantification class 4, the conversion to two dimensions is represented as a kind of projection, similarly to the principal component analysis. However, since the data to be converted here is the searched data, it is data distributed in a narrow area in a multidimensional space, and the number thereof is limited. Therefore, it is considered that the structure of the data can be sufficiently expressed even by transformation using projection such as principal component analysis or quantification.

[0018]

According to a first aspect of the present invention, there is provided a data search apparatus comprising: an electronic database in which each data is represented by a multidimensional attribute value; Search means for searching data in the database, a display device for displaying the data searched by the search means from the database, and a pointing device for designating a position on the screen of the display device stage,
A data retrieval device comprising: a storage device for temporarily storing data output from the database; and a coordinate for converting attribute values of the data retrieved by the retrieval means and stored in the storage device into two-dimensional coordinates. It is characterized by comprising a conversion means, and displaying data on a two-dimensional plane of the display device based on the coordinate values converted by the coordinate conversion means.

According to a second aspect of the present invention, there is provided a data search apparatus according to the first aspect of the present invention, wherein the coordinate transformation means converts a data attribute value by principal component analysis. It is characterized in that it is converted into two dimensions.

According to a third aspect of the present invention, there is provided a data search apparatus according to the first aspect of the present invention, wherein the coordinate conversion means converts the attribute values of the data into quantified four types. Is converted into two-dimensional data by

According to a fourth aspect of the present invention, there is provided a recording medium on which a data search program according to the present invention is recorded. A program for executing a procedure of converting a value into two-dimensional coordinates and a procedure of displaying data on a two-dimensional plane based on the converted coordinate values is recorded.

[0022]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of a data retrieval apparatus according to the present invention will be described with reference to the drawings. FIG. 1 is a schematic configuration diagram of a data search apparatus according to the present invention. In FIG. 1, reference numeral 1 denotes a database stored in a storage device such as a CDROM or a hard disk;
Is a pointing device such as a mouse, 3 is a display device such as a CRT display, 4 is search means for executing data search, 5 is coordinate conversion means for converting attribute values of data into two-dimensional plane coordinates, 6 Represents the coordinate transformed data,
The two-dimensional plane display means 7 arranged on the two-dimensional plane is a storage device such as a RAM for temporarily storing data.

FIG. 2 shows an example of the configuration of a database according to the data search apparatus of the present invention. The database of the present embodiment is a photograph image database. As image data,
Image data of an image is stored as a file. Each attribute is given a phrase that expresses the image of the photograph, such as "elegance", "power", or "lyrical". As the attribute value added to the data, the degree of each image in those words is given numerically in order. ing. In the database,
An image file name and an attribute value are held as a pair for each image.

FIG. 3 is a flowchart for explaining the operation of the data search device of the present invention. Hereinafter, the operation will be described with reference to FIG. In the present embodiment, a user searches data from a first two-dimensional plane displayed on a screen of a display device by specifying a position on the screen with a pointing device, as described in the related art. In the search device, the data is directly output according to the number of searched data, or the second data is displayed on the screen of the display device.
And arranging the retrieved data on a second two-dimensional plane so as to represent its structure,
It is an example to which the present invention is applied.

In step S1, the data arranged on the first two-dimensional plane is displayed on the screen of the display device 3. In the data display here, for example, the attribute value of each data is converted into two-dimensional data by the algorithm of the self-organizing map, and the data of the first two-dimensional plane formed on the screen of the display device 3 is converted. This is performed by displaying a symbol corresponding to the data at the two-dimensional coordinate position. Hereinafter, this first two-dimensional plane is referred to as a map.

By this processing, a symbol is displayed at the position where the data exists on the map, and the distribution of the data can be understood at a glance. Further, processing such as displaying the symbol in accordance with the number of data existing on the same coordinates, or changing the shape or size of the symbol may be performed. Here, the symbol may be a simple symbol such as a circle or a square. However, if a reduced image of the corresponding image is displayed, the symbol is more easily understood by the user.

FIG. 4 is a display example of a map according to the data search device of the present invention. In FIG. 4, the square is drawn in proportion to the number of data on the coordinates. In step S2, the process waits until the user specifies a position on the map with the pointing device. More specifically, when the pointing device is a mouse, it waits for a click on the map. When the position is specified, in step S3, the search means extracts data arranged in the vicinity thereof as a search result. Specifically, the distance between the designated point and the converted coordinates of each data is calculated, and an index corresponding to data whose value is equal to or less than a predetermined value is extracted from the database. The index in this case is an image file name stored in the database. However, here, the attribute value of the corresponding data is also taken out at the same time as the index from the convenience of later processing, and stored in the storage device in order.
Further, the number of searched data is also stored.
Up to this point, it can be realized by the conventional technology.

In step S4, it is checked whether or not the number of searched data is equal to or smaller than a predetermined number. In step S6, data is output with reference to the index stored in the storage device. In the present embodiment, the index is given as a file name, and the data output in step S6 is executed by opening the corresponding image file and outputting the image on the screen of the display device 3.

Next, two-dimensional display of search data in step S5, which is a feature of the present invention, will be described. Here, the attribute value of the retrieved data is
Converted to dimensional coordinate values. Thereafter, the converted data is displayed on a two-dimensional plane by the two-dimensional plane display means 6. FIG. 5 illustrates the flow. The first method of coordinate conversion is based on principal component analysis, and the second method is based on quantification type 4. They differ in the method of creating a correlation matrix and the method of conversion to two-dimensional coordinates. First, the operation of the method based on the principal component analysis will be specifically described.

First, as described above, the number of searched data, the index of each data, and the attribute value are stored in the storage device. FIG. 6 shows this. The number of attributes is now n, and the number of retrieved data is m. In step S11, the attribute values are read from the storage device, and a correlation matrix is created. Here, the attribute value of each data is X _i = (x _i1 ,... X
_in ), the correlation matrix A = (a _ij ) is defined by the following equation.

A _ij = Σ _{k = 1} ^m x _ki x _kj The creation of the above correlation matrix will be specifically described with reference to FIG. First, the numerical value '18' in the figure is the number of searched data, and thereafter, the file names and attribute values of the 18 data are stored. The first of the retrieved data has the file name '04806.jpg' and the attribute value (1,0,0,4,1,2)
In, reads the attribute value to _{X 1, X 1 = (1,0,0,4,1} ,
2) Similarly, X ₂ = (0,1,0,1,0,1) and X ₃ =
(0,1,0,2,0,3),..., X ₁₈ = (0,0,0,0,0,1).
When A = (a _ij ) is calculated from the above definition equation, it is as follows.

A ₁₁ = 1 × 1 + 0 × 0 + 0 × 0 + ... + 0 × 0 = 6 a ₁₂ = 1 × 0 + 0 × 1 + 0 × 1 + ... + 0 × 0 = 0 a ₆₆ = 2 × 2 + 1 × 1 + 3 × 3 +... + 1 × 1 = 111 In step S12, an eigenvector for the largest eigenvalue of A and an eigenvector for the second largest eigenvalue of A are obtained. To find the eigenvectors of a matrix,
Conventionally, various algorithms have been proposed, and an appropriate one is adopted here. In step S13, the attribute values of each data are projected onto a two-dimensional plane composed of two eigenvectors. Actually, since the above two eigenvectors are orthogonal, the inner product of the attribute value and each of the eigenvectors may be obtained, and these may be used as converted two-dimensional coordinate values. That is, two eigenvectors are sequentially expressed as Q ₁ =
(Q ₁₁ ,..., Q _1n ), Q ₂ = (q ₂₁ ,..., Q _2n ), and the coordinates Y _i = (u _i , v _i ) obtained by converting X _i are given by the following equations. Can be

U _i = Σ _{k = 1} ⁿ q _{1 k} x _ik v _i = Σ _{k = 1} ⁿ q _{2 k} x _{ik When} the calculation on the data in FIG. 6 is performed, first, the eigenvectors of A are Q ₁ = (0.07,0.09,0.04,0.68,0.0
2,0.71), Q ₂ = (− 0.30,0.17,0.00, −0.66, −0.06,0.
64). When Y _i = (u _i , v _i ) is obtained using this, the following is obtained.

U ₁ = 0.07 × 1 + 0.09 × 0 + 0.04 × 0 + 0.68 × 4
+ 0.02 × 1 + 0.71 × 2 = 4.23 v ₁ = −0.30 × 1 + 0.017 × 0 + 0.00 × 0−0.66 × 4−0.06 × 1
+ 0.64 × 2 = -1.72 u ₁₈ = 0.07 × 0 + 0.09 × 0 + 0.04 × 0 + 0.68 × 0 + 0.02 × 0 +
0.71 × 1 = 0.71 v ₁₈ = −0.30 × 0 + 0.017 × 0 + 0.00 × 0−0.66 × 0−0.06 ×
0 + 0.64 × 1 = 0.64 In step S14, each Y _i is stored in a buffer in the storage device. In step S15, each data is symbolized. The symbols used here are simple symbols such as circles or squares or reduced images of images, as described in step S1. However, since the number of display data is smaller than that in step S1, some text such as the file name of the image or the title of the image can be displayed as a symbol. Also, if some keyword is added to each data,
They may be used as symbols. The symbolization described above is possible with the prior art. The generated symbols are stored in a storage device if necessary.

In step S16, a second two-dimensional plane is formed on the screen of the display device 3. Specifically, parameters such as the size of the two-dimensional plane, the display position on the screen, the display magnification, and the position of the origin are set, and the frame is drawn. Also, a coordinate axis may be drawn. These parameters can be set to optimal values based on the search data for each search, or you can let the user specify them,
In the present embodiment, the parameters use fixed values that are optimally set in advance. In step S17, it reads the coordinate position Y _i from the buffer, the symbols of each data generated in step S15, displayed at a position corresponding to the Y _i of the second two-dimensional plane. Actually, the physical position on the screen corresponding to Y _i is obtained from the parameters of the two-dimensional plane, and the symbol is displayed at that position. However, since this is a method generally used conventionally, Detailed description is omitted.

The above is the flowchart of the two-dimensional display of the search data in step S5 by the principal component analysis. In an actual calculation, it is preferable to normalize X _{i so} that the average becomes zero. That is, after performing the conversion of the following equation, the generation of A, the calculation of the eigenvalue, and the conversion to Y _i are performed.

X _i = X _i −EX, (EX = (1 / m) Σi _{= 1} ^m X _i ) FIG. 7 shows the second 2
It is a screen output example in which a two-dimensional plane is configured, and the map and the second two-dimensional plane are simultaneously displayed on a display device. FIG.
(a) is the map described in step S1, and FIG.
Is an example in which the searched data is output to the second two-dimensional plane when the * position on the map is designated in the second two-dimensional plane described in step S5 or step S17. Here, the data symbol display is simply “•”
It is done in. Therefore, in step S15, "." May be simply set as a symbol for each data.

Next, the quantification method will be described. As described above, the difference from the method based on the principal component analysis is the method for creating a correlation matrix and the method for converting to a two-dimensional coordinate. Therefore, only that part will be described. First, the quantification type 4 is a method of quantifying qualitative data given a degree of correlation between data. In the present embodiment, the degree of correlation between data is given by a direction cosine of an angle between vectors composed of data attribute values, and each data is regarded as qualitative data and quantified. Therefore, the attribute value of the data is X _i = (x _i1 , ... x _in )
Then, the degree of correlation between the data i and j is given by the following equation.

Cos θ _ij = (X _i , X _j ) / | X _i || X _j | = Σ
_{k = 1} ⁿ x _ik x _jk / (Σ _{k = 1} ⁿ x _ik ² ) ^1/2 (Σ _{k = 1} ⁿ x _jk ² ) ^1/2 First, a matrix Γ = (γ _ij ) is defined by the following equation. .

Γ _ij = cos θ _ij if i ≠ j γ _ii = 0 Next, a matrix P = (p _ij ) is defined by the following equation.

P _ij = 0 if i ≠ jp _ii = Σ _j ≠ _i γ _{ij The} correlation matrix A is defined by the following equation.

A = Γ−P The eigenvector for the largest eigenvalue of A obtained above is (1,
1, ..., 1). In step S12, an eigenvector for the second largest eigenvalue of A and an eigenvector for the third largest eigenvalue of A are obtained. This is the same as the method based on the principal component analysis. The obtained two eigenvectors are represented by Q ₁ = (q ₁₁ ,... Q _1m ) and Q ₂ = (q ₂₁ ,.
q _2m ), the coordinates Y _i = (u _i , v _i ) obtained by converting X _{i in} step S13 are given by the following equations.

When the calculation of u _i = q _1i v _i = q _2i or more is performed only for the data of FIG.
It looks like this:

Γ ₁₂ = (1 × 0 + 0 × 1 + 0 × 0 + 4 × 1 + 1 × 0 +
2 × 1) / (1 ² +0 ² +0 ² +4 ² +1 ² +2 ² ) ^1/2 (0 ² +1 ² +0 ²
+1 ² +0 ² +1 ² ) ^1/2 = 0.74 γ ₁₃ = (1 × 0 + 0 × 1 + 0 × 0 + 4 × 2 + 1 × 0 + 2 × 3) / (1 ²
+0 ² +0 ² +4 ² +1 ² +2 ² ) ^1/2 (0 ² +1 ² +0 ² +2 ² +0 ² +
3 ² ) ^1/2 = 0.80 γ ₁₈₁₇ = (0 x 0 + 0 x 0 + 0 x 0 + 0 x 3 + 0 x 0 + 1 x 3) /
(0 ² +0 ² +0 ² +3 ² +0 ² +3 ² ) ^1/2 (0 ² +0 ² +0 ² +0 ² +0 ²
+1 ² ) ^1/2 = 0.71 or less P, A, Q ₁ , Q ₂ , and Y _i can also be easily calculated by applying the above formula, and a specific calculation is omitted here. This concludes the description of the coordinate conversion method based on Quantification Class 4.

The operation of the two-dimensional display of the search data has been described above. FIG. 8 is a display example of a two-dimensional plane in actual search data, in which an image title is displayed as a data symbol. In this figure, the map is omitted.
From the figure, it can be seen that the images that are determined to be sensuously close are arranged close to each other, and are effectively arranged so as to express the data structure as a whole. In addition, coordinate transformation by the above-described principal component analysis or quantification type 4 requires iterative calculation to find the eigenvectors of the matrix.
It is fast because only two eigenvectors with large eigenvalues need to be calculated.

In the above description, only the search data is arranged on the two-dimensional plane. However, it is also possible to output the corresponding search data by specifying the position of the two-dimensional plane. Specifically, in step S2, 2
As a process when a point on the dimensional plane is designated, a routine for outputting search data within a certain distance from the designated point may be added. As the operation, the two-dimensional plane coordinate values Y _i for each search data stored in the buffer, the distance between the coordinate values of the specified point is calculated and about it is equal to or less than a predetermined value of the data, corresponding to the index Open the image file to be output and output it to the screen.

The embodiment of the present invention has been described above.
It has been confirmed that when the number of searched data is large, the data can be arranged on the two-dimensional plane quickly and effectively, thereby facilitating finding the target data.

In the present embodiment, an image search is described as an example, but the data search device of the present invention is not limited to this. Further, in the present embodiment, as a search method, a position on the map is designated and data in the vicinity thereof is extracted. However, the coordinate conversion means of the present invention can similarly perform coordinate conversion on data extracted by another search method. . Therefore, the data search device of the present invention does not limit the search method.

In the recording medium of the present invention, a data retrieval program executed by the retrieval means 4, coordinate transformation means 5, and two-dimensional plane display means 6 enclosed by a dotted line in FIG. 1 is recorded.

[0050]

According to the present invention, searched data is relocated on a two-dimensional plane at high speed and effectively. This makes it easy to find necessary items from a large amount of search data.

[Brief description of the drawings]

FIG. 1 is a schematic configuration diagram of a data search device of the present invention.

FIG. 2 is a configuration diagram of a database according to the data search device of the present invention.

FIG. 3 is a flowchart for explaining the operation of the data search device of the present invention.

FIG. 4 is a display example of a map according to the data search device of the present invention.

FIG. 5 shows search data 2 according to the data search device of the present invention.
The flowchart of a dimension display.

FIG. 6 is a diagram showing a state of searched data stored in a storage device according to the data search device of the present invention.

FIG. 7 is an example of output to a two-dimensional plane of the display device according to the data search device of the present invention.

FIG. 8 is a diagram illustrating a data search device according to the present invention;
An example in which coordinate-converted data is output to a two-dimensional plane as a title of an image as a symbol.

FIG. 9 is an example of multidimensional data for explaining a method of projecting multidimensional data from two coordinate axes onto a plane according to the related art.

FIG. 10 shows an example of data arrangement in a conventional method of projecting multidimensional data from two coordinate axes onto a plane.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Database 2 Pointing device 3 Display device 4 Retrieval means 5 Coordinate conversion means 6 Two-dimensional plane display means 7 Storage device

Claims (4)

[Claims] An electronic database in which each data is represented by a multi-dimensional attribute value; a search means for searching data in the database; and a display for displaying data searched by the search means from the database. A data retrieval device comprising: a device; a pointing device for designating a position on the screen of the display device; and a storage device for temporarily storing data output from the database. A coordinate conversion means for converting the attribute values of the data stored in the storage device into two-dimensional coordinates, based on the coordinate values converted by the coordinate conversion means, on a two-dimensional plane of the display device. A data search device for displaying data. 2. The data search apparatus according to claim 1, wherein said coordinate conversion means converts the attribute value of the data into two-dimensional data by principal component analysis. 3. The data search device according to claim 1, wherein said coordinate conversion means converts the attribute value of the data into two dimensions by quantification. 4. A computer for retrieving data from a database, a procedure for converting attribute values of the data obtained by the retrieval into two-dimensional coordinates, and A recording medium on which a data search program for executing a procedure for displaying data is recorded.