JP2011022829A - Sensibility search engine - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a sensibility search engine for easily and highly precisely searching for a piece of work matching one's own sensibility. <P>SOLUTION: The sensibility search engine includes: a sensibility measurement image database DB 1 for storing images for measurement; a work image database DB2 for storing work images; a feature amount extraction part 4 for extracting the feature amount of the image every time a new image is stored; a feature amount database DB3 for storing a feature amount; a sensibility amount measurement part 5 for conducting a survey about the image for measurement for a new searcher and measuring the sensibility amount on the basis of its result; a sensibility amount database DB4 for storing the sensibility amount; a search part 6 for displaying work images in order determined on the basis of a correlation value between the sensibility amount and a feature amount stored in the databases DB3 and DB4; and a sensibility amount update part 7 for correcting the sensibility amount of the searcher in the direction approaching the feature amount of the work image and restoring it when one of the displayed work images is enlarged and displayed. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a sensibility search system that enables a searcher to search for a work that matches his sensibility from works registered by a work provider.

  For example, a system described in Patent Document 1 is known as a system for searching for works (contents) that match one's preference and sensitivity. In this system, when a searcher inputs an emotional and ambiguous search term (cool, cold, hot, warm, pop, plain, etc.), a color pattern corresponding to the search term is determined by referring to a database. Is done. Then, a work having the color pattern is searched from the database and presented to the searcher.

JP 2008-171243 A

  However, with this system, each time a search is performed, the searcher needs to input a sensitive search word, which is very troublesome. In addition, since an appropriate search word for expressing the work to be searched cannot be found, a work different from the desired work may be presented or the search itself may be given up.

  Accordingly, an object of the present invention is to provide a sensitivity search system that can easily search for a work that matches the sensitivity of the user with high accuracy.

In order to solve the above-described problem, a sensitivity search system according to the present invention includes:
The server is connected to the terminal of the work provider and the terminal of the searcher via a network, and is capable of transmitting / receiving data to / from each terminal, from among works registered by the work provider The sensibility search system in which the searcher can search for works that match his sensibility, wherein the server stores a plurality of measurement images used for measuring the sensibility of the searcher. An image database for measurement, a work image database for storing work images of a plurality of works to be registered sent from the terminal of the work provider, and a new measurement in the image database for sensitivity measurement and the work image database Each time a product image and a work image are stored, a feature amount extraction unit that extracts a feature amount related to a predetermined sensitivity word of the image, and the extracted feature amount A plurality of measurement images stored in the sensitivity measurement image database are displayed on a feature database stored in association with an image and a searcher's terminal that uses the system for the first time, and a questionnaire regarding the image is performed. Based on the result of the questionnaire sent from the terminal, a sensitivity amount measuring unit that measures a sensitivity amount related to the predetermined sensitivity word of the searcher, and associates the measured sensitivity amount with the searcher And when the searcher's terminal is instructed to perform a search, the sensibility amount of the searcher stored in the sensitivity amount database and the feature amount database are stored. The correlation value with the feature amount of the work image being obtained is obtained, and the work image is reduced in the order determined based on the correlation value When one work image is selected from the work images displayed on the terminal by the search unit to be displayed on the recording terminal and the searcher who has performed the search, the work image is enlarged and displayed on the terminal. A sensitivity amount updating unit that corrects and restores the sensitivity amount of the searcher stored in the sensitivity amount database in a direction approaching the feature amount of the work image. .

  According to this configuration, if the searcher answers the questionnaire only for the first time using the system, the amount of sensitivity is measured and stored in the sensitivity amount database. It is possible to automatically search for works that match one's own sensibility based on the stored sensibility without inputting words.

  In addition, according to this configuration, the searcher's sensibility is updated as needed based on the feature quantity of the work that the searcher has enlarged, that is, the work that interests him, so the search accuracy improves as he continues to use it. And you can search for works that match your sensibility with high accuracy.

  In the above configuration, the feature amount may be extracted based on hue, saturation, and brightness of the measurement image and the work image that are to be extracted.

  In the above configuration, the questionnaire is (1) one that allows the user to select a favorite image and / or a dislike image from the plurality of measurement images displayed on the searcher's terminal, or (2) An image that matches the predetermined sensitivity word may be selected from the plurality of measurement images displayed on the searcher's terminal.

  Further, in the above configuration, the updated sensitivity amount re-stored by the sensitivity amount update unit is based on a correlation value between the sensitivity amount before the searcher's update and the feature amount of the enlarged displayed work image. It is preferable to obtain the weighted average of the feature amount with a weight greater than 0 and less than or equal to 1 and the sensitivity amount before update with a weight of 1.

  According to the present invention, it is possible to provide a sensibility search system that can easily and accurately search for works that match one's sensibility.

It is a block diagram of the Kansei search system concerning the present invention. It is an example of the image stored in the image database for sensitivity measurement, and a work image database. FIG. 2 is a screen displayed on a searcher's terminal in the first embodiment, where (A) is a sensitivity measurement screen and (B) is a search result screen. It is a graph for determining the weight used by a weighted average. FIG. 6 is a screen displayed on a searcher's terminal in Example 2, where (A) is a sensitivity measurement screen and (B) is a search result screen. FIG. 6 is a screen displayed on a searcher's terminal in Example 3, where (A) is a sensitivity measurement screen and (B) is a search result screen.

  Hereinafter, preferred embodiments of a sensitivity search system according to the present invention will be described with reference to the accompanying drawings.

FIG. 1 is a block diagram of a sensitivity search system according to the present invention. As shown in the figure, the sensibility search system 1 is connected to a work provider's terminals T _A1 , T _A2 ... And a searcher's terminals T _S1 , T _S2. And a server 2 capable of transmitting / receiving data to / from each terminal. Further, the server 2, transmission and reception of data is made possible even with a terminal T _S of the system administrator. The server 2 and the terminal T _S, may be connected via a network NT. The server 2 itself may have combined the functions of the terminal T _S of the system administrator.

The server 2 includes a sensitivity measurement image database DB1 and a work image database DB2. Of these, the sensitivity measurement image database DB1, the system administrator has registered with the terminal T _S, the measurement image for measuring the sensibility of the searcher is stored. Data of the measurement images sent from the terminal T _S is stored through the transceiver 3 of the server 2 image database DB1 for sensitive measurements.

The work image database DB2 stores a work image of his / her work registered by the work provider using, for example, the terminal T _A1 . The work image data sent from the terminal T _A1 is stored in the work image database DB2 through the transmission / reception unit 3 of the server 2.

In this embodiment, eight measurement images A to H as shown in FIG. 2 and the following table are stored in the sensitivity measurement image database DB1.
That is, each image P is composed of a rectangular area Pa and a circular area Pb located at the center thereof, and each image P is colored with a different color. For example, the measurement image A is colored with the area Pa being yellow and the area Pb being red.

  In the present embodiment, for the sake of simplicity, the same images as the measurement images A to H are stored in the work image database DB2 as work images a to h of works registered by the work provider. To do. Naturally, the number of work images stored in the work image database DB2 increases as the work provider registers his / her work.

  The server 2 includes a feature amount extraction unit 4. When a new image is stored in the sensitivity measurement image database DB1 and the work image database DB2, the feature amount extraction unit 4 extracts the feature amount of the image. Here, the feature amount expresses a tendency with respect to predetermined five sensitivity words “warmth”, “softness”, “naturalness”, “brightness”, and “flashyness” as a numerical value. The feature amount extraction unit 4 extracts feature amounts based on the hue, saturation, and brightness of the image to be extracted. The feature amount takes a value of 0 to 10, and for example, in the case of an image in which a warm color such as red is often used, the “warm” feature amount tends to increase. Further, in the case of an image in which a color such as black is frequently used, the value of “brightness” tends to be low.

The feature amounts of the measurement images A to H (work images a to h) extracted by the feature amount extraction unit 4 are as shown in the following table.

  The extracted feature quantity is stored in the feature quantity database DB3 in association with the image to be extracted. In the present embodiment, the feature amount is stored in association with eight measurement images A to H registered by the system administrator and eight piece images a to h registered by the piece provider.

  The server 2 includes a sensitivity amount measuring unit 5. The sensitivity amount measuring unit 5 measures the sensitivity amount of a searcher who uses the sensitivity search system 1 for the first time. Here, the sensitivity level is a numerical expression of the searcher's preference regarding the five predetermined sensitivity words "warm", "softness", "naturalness", "brightness", and "flashy". It is. The amount of sensibility takes a value between 0 and 10; for example, in the case of a searcher who likes a reddish work, the value of “warmth” tends to increase. For searchers who prefer dark works, the “brightness” value tends to be low. In order to measure the sensitivity amount, it is necessary that a plurality of measurement images be registered by the system administrator as advance preparation.

  The measurement of the sensitivity amount in the present embodiment is performed by presenting a plurality of measurement images to a searcher who is a measurement target, and performing a questionnaire for selecting a favorite image and a dislike image from the images.

Specifically, the sensibility measuring unit 5 displays the sensibility measurement screen P1-1 (FIG. 3A) on the terminal of the searcher who is the measurement target (for example, T _S1 in FIG. 1). The sensitivity measurement screen P1-1 has an image area A10, a favorite image area A11, and a dislike image area A12. In the image area A10, eight measurement images A to H initially stored in the sensitivity measurement image database DB1 are displayed. On the other hand, nothing is initially displayed in the favorite image area A11 and the disliked image area A12.

  The searcher moves one or more favorite images among the measurement images displayed in the image area A10 to the favorite image area A11. In addition, when there is a disliked image in the measurement image displayed in the image area A10, the searcher moves the image to the disliked image area A12. If there are no images you don't like, you don't need to do this. In short, it is sufficient that at least one favorite image is selected.

  When selection of a favorite image and a dislike image is completed and the searcher presses the registration button B1, the sensitivity measurement unit 5 measures the sensitivity of the searcher according to the following flow.

(Step 1-1)
According to the following equation, the feature quantities of the selected m favorite images are added up for each sensitivity word.
Here, the symbol i is a symbol indicating that it relates to a predetermined sensitivity word (for example, i = 1 means “warm” and i = 5 means “flashy”). WL _j K _i indicates a feature amount related to a predetermined sensitivity word of the image WL _j selected as a favorite image.

For example, when measurement images A and B are selected as favorite images, the total value L ₁ related to “warmth” is 9.98 + 1.68 = 11.66.

(Step 1-2)
Next, the difference between the maximum feature value K _Max (10 in the present embodiment) and the feature values of the n disliked images selected is summed for each sensitivity word by the following equation.
Here, WU _j K _i indicates a feature amount for a predetermined sensitivity word of the image WU _j selected as the disliked image.

For example, when the measurement images C, D, and E are selected as disliked images, the total value U ₁ related to “warmth” is (10−7.89) + (10−9.53) + (10− 4.57) = 8.01.

(Step 1-3)
Subsequently, the total value S _i of L _i and U _i obtained previously is obtained by the following equation.
In the above specific example, the total value S ₁ relating to “warmth” is 11.66 + 8.01 = 19.67.

(Step 1-4)
Finally, based on the total value S _i , the searcher's sensitivity amount A _i is obtained by the following equation.
In the above specific example, the sensitivity amount A ₁ related to “warmth” is 19.67 / (2 + 3) = 3.93.

The measured sensitivity amount A _{i for} each sensitivity word of the searcher is associated with the searcher and stored in the sensitivity amount database DB4. When the measurement images A and B are selected as favorite images and the measurement images C, D, and E are disliked, the sensitivity amount database DB4 stores sensitivity amounts A _i as shown in the table below.

The server 2 includes a search unit 6. When receiving a command to search from the searcher's terminal (for example, T _S1 in FIG. 1), the search unit 6 receives all the work images a˜ stored in the sensitivity database DB4 and the feature database DB3. A correlation value with the feature quantity of h is obtained. And the search part 6 arranges and displays what reduced the work image in an order with a high correlation value on the said searcher's terminal _TS1 . In order to perform this search, as a preliminary preparation, it is necessary that the sensitivity of the searcher is measured, and that one or more work images are registered by the work provider.

  Specifically, the search unit 6 obtains a correlation value between the searcher's sensitivity amount and the feature amount of the work images a to h using the Pearson correlation according to the following flow.

(Step 2-1)
The sensitivity amount A _i of the searcher stored in the sensitivity amount database DB4 is added up by the following equation.

That is, the sum of the sensitivity amounts related to all sensitivity words is SUM_A.

(Step 2-2)
The feature quantity K _i is summed up for one of the work images a to h (for example, the work image a) stored in the feature quantity database DB3 by the following formula.
That is, the sum of the feature quantities related to all sensitivity words is SUM_K.

(Step 2-3)
Each variable (SUM_SQRT_A, SUM_SQRT_K, pSUM) is obtained by the following equation.

(Step 2-4)
Next, a new variable (num, den) is obtained by the following equation based on the variable obtained in the above step.
Here, n is the number of sensitivity words (5 in this embodiment).

(Step 2-5)
Finally, a correlation value R between the searcher's sensitivity amount and the feature amount of the work image is obtained by the following equation.

  When a plurality of work images are registered, the above steps 2-2 to 2-5 are executed for all remaining work images (b to h), and correlation values R for all the work images are obtained.

As an example, when the measurement image B is selected as a favorite image on the sensitivity measurement screen P1-1 and nothing is selected as a dislike image, the correlation values R of the respective work images a to h are as shown in the following table. become that way.

  As shown in the table above, the correlation value R takes a value of −1 to +1, and the closer the correlation value R is to +1, the higher the correlation between the feature value of the work image and the sensitivity of the searcher, that is, the work. Can be said to match the sensitivity (preference) of the searcher. On the contrary, the closer the correlation value R is to -1, the more the work deviates from the searcher's sensitivity.

The search unit 6 displays the reduced work images arranged on the searcher's terminal T _S1 in order from the highest correlation value R. That is, in this embodiment, as shown in the search result screen P1-2 in FIG. 3B, the work images are displayed in the order of b → f → h → e.

  As shown in FIG. 1, the server 2 includes a sensitivity amount update unit 7. The sensitivity amount update unit 7 is stored in the sensitivity amount database DB4 based on the feature amount of the work image selected and enlarged by the searcher from the work images displayed on the search result screen P1-2. The sensitivity of the searcher is updated. As a result, the sensibility amount stored in the sensibility amount database DB4 becomes closer to the sensibility of the actual searcher as you continue to use it, and you can search for works that match your sensibility with high accuracy. become able to.

  Specifically, the sensitivity amount update unit 7 updates the sensitivity amount according to the following flow.

(Step 3-1)
The sensitivity amount update unit 7 first obtains a correlation value R between the feature amount of the work image enlarged and displayed by the searcher and the current sensitivity amount of the searcher. The correlation value R is obtained by the same method as in steps 2-1 to 2-5.

(Step 3-2)
Next, for convenience of calculation, the correlation value R in the range of −1 to +1 is converted into the distance R ′ in the range of 0 to 3. The distance R ′ is obtained by the following equation.
As is clear from the above equation, when the correlation value R is high, the distance R ′ is close (close to 0). On the other hand, when the correlation value R is low, the distance R ′ becomes far (close to 3).

(Step 3-3)
The weight W corresponding to the distance R ′ is obtained using the following equation (Gauss function, see FIG. 4).
Here, σ is 0.5.

  As is clear from the graph shown in FIG. 4, the greater the distance R ′, the smaller the weight W. Conversely, the closer the distance R 'is, the closer the weight W is to 1, which is the maximum value. That is, when a work image close to the sensibility of the searcher is enlarged and displayed, the correlation value R is high and the distance R ′ is close, so the weight W is large. As will be described later, as the weight W is larger (= distance R ′ is closer), the searcher's sensitivity varies greatly before and after the update.

(Step 3-4)
Next, using the weight W, a weighted average of the searcher's sensitivity amount and the feature amount of the enlarged work image is calculated for each sensitivity word. The calculation is performed according to the following formula.
Here, A _i is the sensitivity amount before the update, A ′ _i is the sensitivity amount after the update, and K _i is the feature amount of the enlarged work image. The weight of A _i is 1, and the weight of K _i is W (0 <W ≦ 1).

For example, the weight W determined on the basis of the correlation value R is 0.5, the sensitivity amount A ₁ related to “warmth” before update is 8, and the feature value K ₁ related to “warmth” of the enlarged work image is 5 In this case, the sensitivity amount A ′ ₁ related to the “warmth” after the update is (8 × 1 + 5 × 0.5) / (1 + 0.5) = 7.

In addition, after the sensitivity amount is determined by selecting the measurement image B as a favorite image on the sensitivity measurement screen P1-1, the work image b whose correlation value R is 1.00 on the search result screen P1-2. The updated sensitivity values when the work image e having a correlation value R of 0.12 and the work image a having a correlation value R of −0.66 are enlarged and displayed, respectively, are as shown in the table below.

  Since the feature amount of the work image b and the sensitivity amount of the searcher before the update are exactly the same, even if the image b is enlarged, the sensitivity amount stored in the sensitivity amount database DB4 is completely before and after the update. It does not change. When the work image e having a correlation value R of 0.12 is enlarged and displayed, the sensitivity amount corrected in the direction approaching the feature amount of the work image e is stored. In addition, the work image a having a correlation value R of −0.66 is significantly different from the sensibility of the searcher, and the weight W is low. Therefore, even if the work image a is enlarged and displayed, the sensitivity level is almost the same before and after the update. It does not change.

  As described above, according to the Kansei search system according to the present invention, the searcher only needs to answer the questionnaire once for the first time using the system. It is possible to automatically search for a work that matches one's own sensibility based on the measured and stored sensibility.

  In addition, according to the sensitivity search system according to the present invention, the searcher's sensitivity amount is updated as needed based on the work image enlarged by the searcher, that is, the feature amount of the work of interest, so that the searcher continues to use it. The search accuracy will improve, and you can search for works that match your sensibility with high accuracy.

  In the present embodiment, sensitivity measurement by the sensitivity measurement unit 5 is performed by a method different from that in the first embodiment. The other configuration is the same as that of the first embodiment, and thus the description thereof is omitted here.

  The sensibility measuring unit 5 displays the sensibility measurement screen P2-1 (FIG. 5A) on the terminal of the searcher who is the measurement target. The sensitivity measurement screen P2-1 has an image area A20 and a favorite image area A21. In the image area A20, eight measurement images A to H initially stored in the sensitivity measurement image database DB1 are displayed. On the other hand, nothing is initially displayed in the favorite image area A21.

  When the searcher moves one or more favorite images among the measurement images displayed in the image area A10 to the favorite image area A21 and presses the registration button B2, the sensitivity amount measuring unit 5 The sensibility of the searcher is measured according to the flow.

(Step 4-1)
According to the following equation, the feature quantities of the selected m favorite images are added up for each sensitivity word.
Here, the symbol i is a symbol indicating that it relates to a predetermined sensitivity word (for example, i = 1 means “warm” and i = 5 means “flashy”). WL _j K _i indicates a feature amount related to a predetermined sensitivity word of the image WL _j selected as a favorite image.

For example, when measurement images A and B are selected as favorite images, the total value L ₁ related to “warmth” is 9.98 + 1.68 = 11.66.

(Step 4-2)
Based on the total value L _i , the sensitivity amount A _i of the searcher is obtained by the following formula.
In the above specific example, the sensitivity amount A ₁ related to “warmth” is 11.66 / 2 = 5.83.

The measured sensitivity amount A _{i for} each sensitivity word of the searcher is associated with the searcher and stored in the sensitivity amount database DB4. When the measurement images A and B are selected as favorite images, the sensitivity amount database DB4 stores sensitivity amounts A _i as shown in the table below.

  In the present embodiment, sensitivity measurement by the sensitivity measurement unit 5 is performed by a method different from that in the first and second embodiments. Since other configurations are the same as those in the first and second embodiments, description thereof is omitted here.

  The sensibility measurement unit 5 displays the sensibility measurement screen P3-1 (FIG. 6A) on the terminal of the searcher who is the measurement target. The sensitivity measurement screen P3-1 has an image area A30, an image selection area A31, and a sensitivity word switching button B31. When the sensitivity word “warmth” is selected by the sensitivity word switching button B31, the image selection area A31 is an area for moving an image that the searcher feels warm. When the sensitivity word “soft” is selected, the image selection area A31 is an area for moving an image that the searcher feels soft. That is, in this embodiment, the searcher selects an image that feels that the selected sensitivity word is matched while switching the selection of the sensitivity word with the sensitivity word switching button B31.

  In selecting an image, it is sufficient that at least one image is selected for any one sensitivity word. From the viewpoint of accurately measuring the sensitivity of a searcher, one or more images are included for all sensitivity words. Preferably it is selected.

  When the searcher finishes selecting an image and presses the registration button B32, the sensitivity amount measuring unit 5 measures the sensitivity of the searcher according to the following flow. In the following, measurement images A and B are selected as “warm” images, measurement image C as “soft” images, measurement image D as “natural” images, and measurement images as “bright” images. The case where the measurement image F is selected as the “flashy” image E will be described.

(Step 5-1)
The searcher's sensitivity amount A ₁ related to “warmth” is obtained by averaging the feature amounts related to “warmth” of images (measurement images A and B) selected as “warm” images. Since the feature values regarding the “warmth” of the measurement images A and B are 9.98 and 1.68, respectively, the sensitivity amount A ₁ is 5.83.

(Step 5-2)
The searcher's sensitivity amount A ₂ related to “softness” is obtained by averaging the feature values related to “softness” of the image (measurement image C) selected as the “soft” image. Since only the measurement image C is selected as the “soft” image, the sensitivity amount A ₂ is 6.72, which is the same as the feature amount related to the “softness” of the measurement image C.

(Steps 5-3 to 5-5)
Hereinafter, similarly, searchers' sensitivity amounts A _{3 to} A ₅ regarding “naturalness”, “brightness”, and “flashy” are obtained.

(Step 5-6)
The measured sensitivity amount A _{i for} each sensitivity word of the searcher is associated with the searcher and stored in the sensitivity amount database DB4. In the case of the above specific example, the sensitivity amount database DB4 stores sensitivity amounts A _i as shown in the table below.

  The preferred embodiments of the sensitivity search system according to the present invention have been described above, but the present invention is not limited to these configurations.

  For example, the update of the sensibility amount by the sensibility amount update unit 7 is not performed based on the correlation value between the searcher's sensibility amount and the feature amount of the enlarged work image, but simply calculates the sensitivity amount and the feature amount. The averaged value can be used as a new sensitivity amount. In short, the searcher's sensibility amount may be corrected in a direction approaching the feature amount of the enlarged work image.

DESCRIPTION OF SYMBOLS 1 Sensitivity search system 2 Server 3 Transmission / reception part 4 Feature quantity extraction part 5 Sensitivity quantity measurement part 6 Search part 7 Sensitivity quantity update part DB1 Kansei measurement image database DB2 Work image database DB3 Feature quantity database DB4 Sensitivity quantity database

Claims (5)

The server is connected to the terminal of the work provider and the terminal of the searcher via a network, and is capable of transmitting / receiving data to / from each terminal, from among works registered by the work provider , A sensitivity search system in which the searcher can search for works that match his sensitivity,
The server
An image database for sensitivity measurement storing a plurality of measurement images used for measuring the sensitivity of the searcher;
A work image database storing work images of a plurality of works to be registered, sent from the terminal of the work provider;
A feature quantity extraction unit that extracts a feature quantity related to a predetermined sensitivity word of the image each time a new measurement image and work image are stored in the sensitivity measurement image database and the work image database;
A feature amount database for storing the extracted feature amount in association with the image;
The searcher's terminal that uses the system for the first time displays the plurality of measurement images stored in the sensitivity measurement image database, conducts a questionnaire regarding the image, and sends the questionnaire sent from the terminal A sensitivity amount measuring unit that measures a sensitivity amount related to the predetermined sensitivity word of the searcher based on the result;
A sensory amount database for storing the measured sensory amount in association with the searcher;
When an instruction to perform a search is made from the terminal of the searcher, the sensitivity amount of the searcher stored in the sensitivity amount database and the work image stored in the feature amount database A search unit that obtains a correlation value with a feature value and displays the reduced image of the work image on the terminal in an order determined based on the correlation value;
When one work image is selected from the work images displayed on the terminal by the searcher who has searched, the work image is enlarged and displayed on the terminal and stored in the sensitivity amount database. A sensitivity amount updating unit that corrects and re-stores the sensitivity amount of the searcher in a direction approaching the feature amount of the work image;
Kansei search system characterized by having   The sensibility search system according to claim 1, wherein the feature amount is extracted based on a hue, a saturation, and a brightness of the measurement image and the work image that are to be extracted.   3. The questionnaire according to claim 1, wherein the questionnaire is to select a favorite image and / or a dislike image from the plurality of measurement images displayed on the searcher's terminal. Sensibility search system.   3. The questionnaire according to claim 1 or 2, wherein the questionnaire selects an image that matches the predetermined sensitivity word from the plurality of measurement images displayed on the searcher's terminal. Sensibility search system.   The updated sensibility amount re-stored by the sensibility amount update unit is obtained based on the correlation value between the sensibility amount before the searcher's update and the feature amount of the enlarged displayed work image, from 0 5. It is obtained by a weighted average of the feature quantity with a weight that is greater than or equal to 1 and less than the sensitivity quantity before the update with a weight of 1. 5. Kansei search system.