Content evaluation device and content evaluation method

Abstract

A content evaluation device (10) has an acquisition unit (15b) that acquires biological information when content to be processed has been presented to a user. A calculation unit (15c) calculates a distance matrix that represents the correlations between the biological information. A sorting unit (15d) uses the calculated distance matrix to sort the content that corresponds to the biological information into groups. A presentation unit (15e) uses the calculated distance matrix to present the positional relationships of the content that corresponds to the biological information and the groups. A generation unit (15a) generates content that has the features of content or a group that has been designated by the user from among the presented content and groups.

Description

Content evaluation device and content evaluation method

[0001] The present invention relates to a content evaluation device and a content evaluation method.

[0002] A technology has been proposed for determining the quality of a combination while allowing for variations in content categories (see Patent Literature 1). For example, a two-class classifier has been realized that uses machine learning to distinguish between a positive set and a negative set for a combination of multiple items included in a category. Furthermore, for a new combination of items, when a category is assigned to the items, image features corresponding to the category are input, and when a category is not assigned, an average image is input, and machine learning and classification are performed on the classifier.

[0003] JP 2019-28698 A

[0004] However, with conventional technologies, it can be difficult to evaluate how each user perceives content when viewing it and generate new content based on the evaluation. For example, in conventional technologies, a classifier is trained and discriminated using a pre-collected combination of items (coordinates) as a positive set and a random combination of items as a negative set. However, impressions such as positive / negative are essentially determined by the user themselves, and this technology fails to reflect the user's own perception. Furthermore, it is difficult to generate new items that have the characteristics of multiple items or to evaluate the generated items simply by determining the combination of items.

[0005] The present invention has been made in view of the above, and aims to evaluate how each user perceives content when viewing the content, and to generate new content (having similar characteristics) based on the evaluation.

[0006] In order to solve the above-mentioned problems and achieve the object, the content evaluation device of the present invention is characterized by having an acquisition unit that acquires biometric information when content to be processed is presented to a user, a calculation unit that calculates a distance matrix that represents the correlation between the biometric information, a classification unit that uses the calculated distance matrix to classify each piece of content corresponding to the biometric information into groups, a presentation unit that uses the calculated distance matrix to present the positional relationship of each piece of content and each group that corresponds to the biometric information, and a generation unit that generates content that has the characteristics of the content or group specified by the user from the presented content and group.

[0007] According to the present invention, it is possible to evaluate how each user perceives content when viewing the content, and to generate new content based on the evaluation.

[0008] FIG. 1 is a schematic diagram illustrating a general configuration of a content evaluation device according to this embodiment. FIG. 2 is a diagram illustrating processing by a generation unit. FIG. 3 is a diagram illustrating processing by a generation unit. FIG. 4 is a diagram illustrating processing by an acquisition unit. FIG. 5 is a diagram illustrating a distance matrix. FIG. 6 is a diagram illustrating processing by a presentation unit. FIG. 7 is a diagram illustrating processing by a generation unit. FIG. 8 is a diagram illustrating a feature vector distance matrix. FIG. 9 is a diagram illustrating processing by a presentation unit. FIG. 10 is a flowchart illustrating a content evaluation processing procedure. FIG. 11 is a diagram illustrating an example of a computer that executes a content evaluation program.

[0009] Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings. Note that the present invention is not limited to this embodiment. In addition, in the description of the drawings, the same parts are designated by the same reference numerals.

[0010] [Configuration of Content Evaluation Device] Fig. 1 is a schematic diagram illustrating the general configuration of a content evaluation device of this embodiment. As illustrated in Fig. 1, a content evaluation device 10 of this embodiment is realized by a general-purpose computer such as a personal computer, and includes an input unit 11, a biological information measurement unit 11a, an output unit 12, a communication control unit 13, a storage unit 14, and a control unit 15.

[0011] The input unit 11 is realized using input devices such as a keyboard and a mouse, and inputs various instruction information such as a command to start processing to the control unit 15 in response to input operations by an operator.

[0012] The biological information measurement unit 11a measures biological information such as movements and reactions expressed in the user's behavior. For example, it is an electroencephalograph that measures brain waves on the scalp, and measures the brain waves of the user wearing the electroencephalograph in real time. Note that the biological information measurement unit 11a is not limited to an electroencephalograph, and may be, for example, an eye tracker that measures eye movement or an electrical skin resistance meter.

[0013] The output unit 12 is realized by a display device such as a liquid crystal display, a printing device such as a printer, etc. For example, the output unit 12 presents to the user content to be processed in the content evaluation process described below, and displays the positional relationship of the content as a result of the content evaluation process, etc. In this embodiment, the output unit 12 has a function of switching and displaying multiple images, which are examples of content to be processed, at regular intervals.

[0014] The communication control unit 13 is realized by a NIC (Network Interface Card) or the like, and controls communication between the control unit 15 and external devices via telecommunication lines such as a LAN (Local Area Network) or the Internet. For example, the communication control unit 13 controls communication between the control unit 15 and a management device that manages various types of information, a measurement device equipped with the biological information measurement unit 11a, or the like.

[0015] The storage unit 14 is realized by a semiconductor memory element such as a RAM (Random Access Memory) or a flash memory, or a storage device such as a hard disk or an optical disk. The storage unit 14 stores in advance processing programs that operate the content evaluation device 10, data used during execution of the processing programs, and the like, or temporarily stores them each time processing is performed. The storage unit 14 may be configured to communicate with the control unit 15 via the communication control unit 13.

[0016] In this embodiment, the storage unit 14 stores content 14a, biometric information 14b, a group correspondence table 14c, and the like, which are generated in a content evaluation process described later.

[0017] The control unit 15 is realized using a CPU (Central Processing Unit), NP (Network Processor), FPGA (Field Programmable Gate Array), etc., and executes a processing program stored in memory. As a result, the control unit 15 functions as a generation unit 15a, an acquisition unit 15b, a calculation unit 15c, a classification unit 15d, and a presentation unit 15e, as illustrated in FIG. 1, to perform content evaluation processing. Note that these functional units may be implemented individually or in part in different hardware. The control unit 15 may also include other functional units.

[0018] The generation unit 15a generates content to be processed. In this embodiment, the content to be processed is, for example, an image. However, the content is not limited to an image and may be music, video, or the like. In this case, the content evaluation device 10 includes a speaker.

[0019] Specifically, the generation unit 15a first generates an initial image to be presented to the user first in the content evaluation process described later. For example, the generation unit 15a receives a prompt input from the user via the input unit 11 or the communication control unit 13 and generates the initial image.

[0020] 2 and 3 are diagrams for explaining the processing of the generation unit 15a. A prompt is a command (instruction) input by a user in an interactive system with a computer, and specifies the theme, general genre, category, etc. of the image to be generated as an initial image. In the example shown in FIG. 2, the image theme "icon of cat" is input as the prompt.

[0021] The generation unit 15a receives the prompt and generates an initial image. For example, the generation unit 15a generates an image using an image generation AI that uses a diffusion model called Stable Diffusion (https: / / stability.ai / stable-diffusion), a text2image technique, or the like. In the example shown in FIG. 2 , seven images related to the prompt "icon of cat" are generated, and each image is associated with a content identifier that identifies it and presented to the user via the output unit 12.

[0022] The generation unit 15a also derives a caption and a feature vector for each generated image. Here, a caption is a series of text that describes the content and features of an image. The generation unit 15a derives the caption using, for example, an image2text technique called BLIP (https: / / blog.salesforceairesearch.com / blip-bootstrapping-language-image-pretraining / ) or CLIP Interrogator (https: / / github.com / pharmapsychotic / clip-interrogator). The generation unit 15a may associate the derived caption with each generated image and present it to the user, as illustrated in FIG. 3 .

[0023] The feature vector is a multidimensional numerical vector that represents the structural and semantic features of an image, and is expressed as a set of float-type values ​​of, for example, 2 × 2048. The generation unit 15a derives the feature vector by using, for example, the encode_image function, which is part of the functions of an image generation AI called Stable unCLIP (https: / / huggingface.co / docs / diffusers / main / en / api / pipelines / stable_unclip).

[0024] The generating unit 15a may store the generated images, captions, and feature vectors in the storage unit 14 as content 14a in association with the content identifiers.

[0025] Returning to the description of Fig. 1, the acquisition unit 15b acquires biometric information when the content to be processed is presented to the user. Specifically, the acquisition unit 15b presents an initial image, which is the content to be processed, to the output unit 12, and acquires the biometric information output from the biometric information measurement unit 11a worn by the user who is gazing at the output unit 12.

[0026] Here, Fig. 4 is a diagram for explaining the processing of the acquisition unit. In the example shown in Fig. 4, of the seven initial images shown in Fig. 2, the image with content identifier B is presented to the user, and electroencephalograms are acquired as biometric information from the electroencephalograph, which is the biometric information measurement unit 11a worn by the user.

[0027] The acquiring unit 15b stores the acquired biometric information in the storage unit 14. For example, the acquiring unit 15b acquires electroencephalogram data, which is time-series data of electroencephalogram waveforms themselves, and stores the electroencephalogram data as biometric information 14b in the storage unit 14 for each section in which an image is displayed, in association with a content identifier that identifies the image.

[0028] The acquiring unit 15b may acquire biometric information multiple times for the presentation of one type of content. The acquiring unit 15b may also present the content in a random order to reduce the effect of the order on measurement accuracy.

[0029] Returning to the description of FIG. 1 , the calculation unit 15c calculates a distance matrix that represents the correlation between pieces of biometric information. Specifically, the calculation unit 15c first filters the electroencephalogram data of the biometric information 14b to reduce the influence of noise. Next, when electroencephalogram data is acquired multiple times for one type of image, the calculation unit 15c averages the electroencephalogram data for each image and uses this as the electroencephalogram data for each image.

[0030] Furthermore, the calculation unit 15c calculates a distance matrix from the correlation coefficients between the electroencephalogram data for each content identifier. The distance matrix is ​​an index indicating how far the electroencephalogram data corresponding to each image is from other electroencephalogram data.

[0031] Here, Fig. 5 is a diagram for explaining a distance matrix. As illustrated in Fig. 5, the number of rows and the number of columns of the distance matrix each match the number of images, and the diagonal components are 0.0. For example, when Pearson's product-moment correlation coefficient is applied as the correlation coefficient r, -1≦r≦1, and the value range of the distance matrix is ​​from 0 to 1, as expressed by (1−r) / 2.

[0032] 5, the electroencephalogram data corresponding to image A and the electroencephalogram data corresponding to image B perfectly match, and the distance is calculated to be 0.00. In addition, the distance between the electroencephalogram data corresponding to image A and the electroencephalogram data corresponding to image B is calculated to be 0.58.

[0033] The processing by the calculation unit 15c is not limited to the above. For example, it is not necessary to treat all EEG data equally, and only a portion may be extracted and used. Furthermore, the EEG data may be subjected to preprocessing such as power spectral density estimation to convert the data into frequency components. For example, the calculation unit 15c may use only a portion of the frequency components, or may weight specific frequency components to strongly influence them.

[0034] Returning to the description of FIG. 1 , the classification unit 15d classifies each piece of content corresponding to the biometric information into groups using the calculated distance matrix. For example, the classification unit 15d uses a clustering method such as Ward's method to classify the electroencephalogram data into groups (clusters) according to the proximity of the distances between the pieces of electroencephalogram data. The classification unit 15d may associate the classified groups with the content identifiers of the corresponding content, and store the group correspondence table 14c in the storage unit 14.

[0035] The presenting unit 15e uses the calculated distance matrix to present the positional relationship of each content and each group corresponding to the biometric information. The presenting unit 15e also generates and presents labels for the groups using captions that describe each content.

[0036] 6 is a diagram illustrating the processing of the presentation unit. For example, the presentation unit 15e first calculates the positional relationship of each content and each group as one-dimensional coordinates using multidimensional scaling (MDS). As a result, the one-dimensional coordinates of each content are calculated so that the more similar the electroencephalogram data, the closer the content and the more dissimilar the content, and conversely, the further the content and the more dissimilar the content, the farther the content.

[0037] Furthermore, the presentation unit 15e calculates the coordinates and radius of the group so that all of the content in each group is included. For example, the presentation unit 15e calculates the average of all of the electroencephalogram data associated with the content identifiers included in each group to determine the coordinates of the group as the electroencephalogram data for the group. In this case, the presentation unit 15e may recalculate the distance matrix using a correlation coefficient between the electroencephalogram data for each content identifier and the electroencephalogram data for each group. Furthermore, the presentation unit 15e calculates the radius of the group as the absolute value of the difference between the coordinates of the group and the coordinates of the farthest content in the group.

[0038] The coordinates of each content and each group and the radius of the group are not limited to one-dimensional coordinates. The presentation unit 15e may calculate the coordinates of each content and each group and the radius of the group using a dimension reduction method such as UMAP (Uniform Manifold Approximation and Projection).

[0039] Furthermore, the presenting unit 15e generates a label for the group by using the group and the caption generated for the content of each group. For example, in the example shown in Fig. 3, when image B and image E belong to the same group, the presenting unit 15e compares the captions of the images and sets the common word "Kawaii" as the label for the group.

[0040] The presenting unit 15e is not limited to extracting labels from captions as described above. For example, the presenting unit 15e may input multiple captions into a large-scale language model such as GTP-4 (https: / / openai.com / gpt-4) and estimate abstract words that encompass these captions as labels. Alternatively, the presenting unit 15e may input images of a group into a large-scale language model that can input images, such as GTP-4V (https: / / openai.com / research / gpt-4v-system-card), and generate labels. This may allow keywords to be estimated from specific perspectives, such as the style, artist, impression, or target demographic of the image, and used as labels for the group.

[0041] Alternatively, the user's preferences, emotions, sensitivities, and recalled images and videos may be extracted from the electroencephalogram data, and labels may be set accordingly. For example, images that the user likes and images that the user dislikes may be presented in advance, and the corresponding electroencephalogram data may be acquired. Then, based on the similarity between this electroencephalogram data and the electroencephalogram data obtained when the image to which the label is to be assigned is presented, it is determined whether the label of the image is "like" or "dislike."

[0042] Furthermore, image groups and group labels are not limited to being set automatically as described above, but may be editable and reconfigurable by the user.

[0043] Then, as illustrated in Fig. 6, the presentation unit 15e displays the positional relationship of each content and each group and the label of the group on the output unit 12 and presents them to the user using the coordinates of each content and each group and the radius of the group. In Fig. 6, icons representing each content and each group are arranged on a one-dimensional coordinate system, and the range of each group is indicated by an ellipse. In the example illustrated in Fig. 6, images B and E shown in Fig. 2 are classified into a group labeled "Kawaii," images B, C, and G are classified into a group labeled "Head," and images A, D, E, and F are classified into a group labeled "Sitting." It can also be seen that the electroencephalogram data when image F is presented is the most distant from the electroencephalogram data when image C is presented.

[0044] Returning to the description of FIG. 1 , the generation unit 15a generates content having the characteristics of the content or group specified by the user from among the presented content and groups. Here, FIG. 7 is a diagram for explaining the processing of the generation unit. Specifically, the user refers to the positional relationship of each content and each group and the group label illustrated in FIG. 6 to specify the desired content or the desired content group. In this case, the generation unit 15a generates new content having the characteristics of the content of this group, as illustrated in FIG. 7 .

[0045] For example, in the example shown in Figure 6, if the user specifies a group labeled "Kawaii" that includes images B and E, the generation unit 15a will generate new images H and I as additional images that have the characteristics of the images in this group labeled "Kawaii," as illustrated in Figure 7.

[0046] The generation unit 15a generates an image having the characteristics of the images in the group by using the images, captions, group labels, etc. of the group specified by the user. For example, the generation unit 15a generates a new image by merging the feature vectors of the images in the group by calculating the average value, and inputting the merged result into a de-diffusion model of an image generation AI such as Stable unCLIP. This generates an additional image that inherits the semantic and compositional characteristics of the images in the specified group.

[0047] Alternatively, the generation unit 15a generates a prompt using the captions of the images in the group. For example, the generation unit 15a generates a prompt that emphasizes the label of the group using common features of the captions of the images in the group, or generates a prompt that encompasses all captions using a large-scale language model such as GPT-4. The generation unit 15a then generates a new image by inputting the generated prompt into an image generation AI such as Stable Diffusion. This generates an additional image that inherits the semantic features of the images in the specified group.

[0048] Alternatively, the generation unit 15a may specify features to be excluded or avoided when generating a new image, thereby highlighting the features of the group. For example, the generation unit 15a may merge captions of images in the group to generate a prompt, and merge captions of images other than the specified group to generate a negative prompt. By inputting these to an image generation AI such as Stable Diffusion, an additional image is generated that inherits the features of the images in the specified group and is enhanced to distance itself from the features of the other images.

[0049] The generation unit 15a derives a caption and a feature vector for the generated additional image in the same manner as when generating the initial image. Furthermore, the generation unit 15a may store the generated image, caption, and feature vector for the additional image in the content 14a in association with a content identifier.

[0050] When an additional image is generated, the presentation unit 15e uses the feature vectors of the initial image, which is the previously presented content, and the additional image, which is the generated content, to add and present the additional image, which is the further generated content, in the positional relationship of the presented initial image.

[0051] Specifically, the calculation unit 15c calculates a feature vector distance matrix from the correlation coefficients between the feature vectors of each image. The feature vector distance matrix is ​​an index indicating how far the feature vectors of the additional image are from the feature vectors of the initial image.

[0052] Here, Fig. 8 is a diagram for explaining a feature vector distance matrix. Fig. 8 shows a feature vector distance matrix indicating the distances between the feature vectors of additional images H and I generated in the example shown in Fig. 7 and each of initial images A to G. In the example shown in Fig. 8, the distance between additional image H in the first row and initial images B and E in the group with the specified label "Kawaii" is relatively small, and it can be seen that additional image H inherits the features of initial images B and E.

[0053] Then, the presentation unit 15e calculates the coordinates of the additional image using the feature vector distance matrix and the positional relationship of the initial image. That is, the presentation unit 15e calculates the coordinates of the additional image by trigonometry using the known coordinates of the initial image and the distance between the additional image and the initial image. The presentation unit 15e may also obtain the coordinates of the additional image using the ratio of the distances between the images. For example, in the example shown in FIG. 8, the coordinates P of image H are Ｈ is the coordinate P of image D Ｄ , the distance D between images H and D ＨＤ , the distance D between images H and E ＨＥ is used to calculate as in the following equation (1).

[0054]

[0055] Here, Fig. 9 is a diagram for explaining the processing of the presenting unit. As shown in Fig. 9, the presenting unit 15e adds and presents additional images in the positional relationship of the initial images shown in Fig. 6. In the example shown in Fig. 9, additional images H and I, which are shown surrounded by thick frames, have been added to Fig. 6.

[0056] When an additional image is generated, a label for the group including the additional image may be generated again.

[0057] In this way, a new additional image having the characteristics desired by the user is generated. If the user is not satisfied, the user may instruct the regeneration of the additional image again. In this case, all of the additional images or a portion selected by the user are added as initial images, and the above process is repeated. This broadens the scope of evaluation targets and enables iterative generation and evaluation, making it possible to obtain results that better meet the user's desires.

[0058] Furthermore, by using the distance between the features of content to obtain the coordinates of new content, in contrast to mapping the positional relationships of content using biometric information, it is possible to obtain and illustrate the coordinates of new content without the need for biometric information. Content that inherits the features of existing content and does not significantly change its impression is expected to have similar subjective evaluations, making it easy to obtain a map that is subjectively appropriate for the user.

[0059] [Content Evaluation Process] Next, the content evaluation process performed by the content evaluation device 10 according to this embodiment will be described with reference to Fig. 10. Fig. 10 is a flowchart showing the procedure of the content evaluation process. The flowchart in Fig. 10 starts, for example, when the user performs an operation input to instruct the start of the process.

[0060] First, the generation unit 15a receives a prompt input from the user specifying a genre or the like, and generates an initial image to be presented to the user (step S1). The generation unit 15a also derives a caption and a feature vector for each generated image. The generation unit 15a may associate each generated image, caption, and feature vector with a content identifier and store them in the storage unit 14 as content 14a.

[0061] Next, the acquisition unit 15b acquires biometric information when the content to be processed is presented to the user (step S2). Specifically, the acquisition unit 15b presents an initial image to be processed on the output unit 12 and acquires the biometric information output from the biometric information measurement unit 11a worn by the user gazing at the output unit 12. For example, the acquisition unit 15b selects the initial image to be processed in random order, and repeats the process of acquiring the measurement value of the biometric information when the image is presented from the biometric information measurement unit 11a for all images.

[0062] Furthermore, when biometric information is acquired multiple times for one type of content, the calculation unit 15c averages the biometric information for each content, and sets this as the biometric information for that content.

[0063] Next, the classification unit 15d classifies each piece of content corresponding to the biometric information into groups (step S3). Specifically, the calculation unit 15c calculates a distance matrix representing the correlation between pieces of biometric information. Then, the classification unit 15d classifies each piece of content corresponding to the biometric information into groups using the calculated distance matrix. For example, the classification unit 15d classifies the pieces of content corresponding to the biometric information into groups according to the proximity of the distance between pieces of biometric information using a clustering method such as Ward's method.

[0064] Then, the presentation unit 15e uses the calculated distance matrix to present the positional relationship of each content and each group corresponding to the biometric information (step S4). Specifically, the presentation unit 15e calculates the coordinates and radius of the group so that all content in each group is included. For example, the presentation unit 15e calculates the average of all electroencephalogram data associated with the content identifiers included in each group to determine the coordinates of the group as the electroencephalogram data of the group. At this time, the presentation unit 15e may recalculate the distance matrix using a correlation coefficient between the electroencephalogram data for each content identifier and the electroencephalogram data for each group. Furthermore, the presentation unit 15e calculates the absolute value of the difference between the coordinates of the group and the coordinates of the farthest content in the group as the radius of the group.

[0065] Furthermore, the presenting unit 15e generates and presents a label for the group using the group and the caption generated for the content of each group.

[0066] Next, the generation unit 15a generates content or content having characteristics of the group specified by the user from among the presented content and groups (step S5). For example, the generation unit 15a generates an image having characteristics of the images of the group using the images, captions, group labels, etc. of the group specified by the user. The generation unit 15a also derives captions and feature vectors for the generated additional images.

[0067] Then, when the additional image is generated, the presenting unit 15e adds the additional image, which is the generated content, to the positional relationship of the presented initial image by using the feature vectors of the initial image, which is the content previously presented, and the additional image, which is the generated content, and presents the additional image (step S6). For example, the presenting unit 15e calculates the coordinates of the additional image by using the feature vector distance matrix and the positional relationship of the initial image, and presents the additional image by adding it to the positional relationship of the initial image.

[0068] If the user instructs regeneration of the additional image (Yes in step S7), the generation unit 15a returns the process to step S2. On the other hand, if the user does not instruct regeneration of the additional image (No in step S7), the generation unit 15a ends the series of content evaluation processes.

[0069] [Effects] As described above, in the content evaluation device 10 of this embodiment, the acquisition unit 15b acquires biometric information when content to be processed is presented to a user. The calculation unit 15c calculates a distance matrix representing the correlation between pieces of biometric information. The classification unit 15d classifies each piece of content corresponding to the biometric information into groups using the calculated distance matrix. The presentation unit 15e presents the positional relationship between each piece of content and each group corresponding to the biometric information using the calculated distance matrix. The generation unit 15a generates content having characteristics of the content or group specified by the user from the presented content and groups.

[0070] Specifically, the presenting unit 15e generates and further presents a label for the group using a caption that describes each piece of content.

[0071] In this way, the content is categorized into groups based on the similarity of biometric information and presented, and overlapping and inclusive relationships between groups are indicated, allowing the user to specify a group. This allows the system to evaluate how the user perceives the content, and based on that evaluation, generate new content with similar characteristics.

[0072] For example, by grouping and presenting biometric information corresponding to multiple contents, differences in user impressions can be clearly shown. Also, by assigning labels, it is possible to specifically show in what aspects / categories users' impressions differ. By looking at the overlap and inclusion relationships of groups, it is possible to illustrate the details of the user's subjective perception.

[0073] Furthermore, for example, image generation AI functions can be used to calculate feature vectors from multiple content items that have been objectively grouped using EEG data, and new content that inherits these features can be automatically generated. This will broaden the scope of evaluation targets.

[0074] Therefore, the content evaluation device 10 can evaluate how each user perceives content when viewing the content, and generate new content based on the evaluation.

[0075] Furthermore, the presenting unit 15e uses the feature vectors of the presented content and the generated content to present the generated content in addition to the presented positional relationship.

[0076] In this way, by using the distance between the features of content to obtain the coordinates of new content in addition to mapping the positional relationships of content using biometric information, it is possible to obtain and illustrate the coordinates of new content without the need for biometric information. Content that inherits the features of existing content and does not significantly change its impression is expected to have similar subjective evaluations, making it easy to obtain a map that is subjectively appropriate for the user.

[0077] Furthermore, the range of evaluation targets will be expanded, and iterative generation and evaluation will become possible, making it possible to obtain new content that better meets the user's needs.

[0078] [Program] A program written in a computer-executable language may be created to execute the processes performed by the content evaluation device 10 according to the above embodiment. In one embodiment, the content evaluation device 10 can be implemented by installing a content evaluation program that executes the content evaluation process described above as package software or online software on a desired computer. For example, by executing the content evaluation program described above on an information processing device, the information processing device can function as the content evaluation device 10. The information processing device referred to here includes desktop and notebook personal computers. Other examples of information processing devices include mobile communication terminals such as smartphones, mobile phones, and PHS (Personal Handyphone Systems), as well as slate terminals such as PDAs (Personal Digital Assistants). The functions of the content evaluation device 10 may also be implemented on a cloud server.

[0079] 11 is a diagram showing an example of a computer that executes a content evaluation program. The computer 1000 includes, for example, a memory 1010, a CPU 1020, a hard disk drive interface 1030, a disk drive interface 1040, a serial port interface 1050, a video adapter 1060, and a network interface 1070. These components are connected by a bus 1080.

[0080] The memory 1010 includes a ROM (Read Only Memory) 1011 and a RAM 1012. The ROM 1011 stores a boot program such as a BIOS (Basic Input Output System). The hard disk drive interface 1030 is connected to a hard disk drive 1031. The disk drive interface 1040 is connected to a disk drive 1041. A removable storage medium such as a magnetic disk or optical disk is inserted into the disk drive 1041. The serial port interface 1050 is connected to a mouse 1051 and a keyboard 1052, for example. The video adapter 1060 is connected to a display 1061, for example.

[0081] Here, the hard disk drive 1031 stores, for example, an OS (Operating System) 1091, an application program 1092, a program module 1093, and program data 1094. The various pieces of information described in the above embodiments are stored in the hard disk drive 1031 or the memory 1010, for example.

[0082] The content evaluation program is stored in the hard disk drive 1031 as a program module 1093 in which instructions to be executed by the computer 1000 are written. Specifically, the program module 1093 in which each process executed by the content evaluation device 10 described in the above embodiment is written is stored in the hard disk drive 1031.

[0083] Furthermore, data used for information processing by the content evaluation program is stored as program data 1094, for example, in the hard disk drive 1031. Then, the CPU 1020 reads the program module 1093 and program data 1094 stored in the hard disk drive 1031 into the RAM 1012 as necessary, and executes each of the above-described procedures.

[0084] The program module 1093 and program data 1094 related to the content evaluation program are not limited to being stored in the hard disk drive 1031, and may be stored in, for example, a removable storage medium and read by the CPU 1020 via the disk drive 1041. Alternatively, the program module 1093 and program data 1094 related to the content evaluation program may be stored in another computer connected via a network such as a LAN or a WAN (Wide Area Network), and read by the CPU 1020 via the network interface 1070.

[0085] Although the present invention has been described above as an embodiment, the present invention is not limited to the description and drawings that form part of the disclosure of the present invention. In other words, other embodiments, examples, and operational techniques that can be made by those skilled in the art based on the present invention are all included in the scope of the present invention.

[0086] REFERENCE SIGNS LIST 10 Content evaluation device 11 Input unit 11a Biometric information measurement unit 12 Output unit 13 Communication control unit 14 Storage unit 14a Content 14b Biometric information 14c Group correspondence table 15 Control unit 15a Generation unit 15b Acquisition unit 15c Calculation unit 15d Classification unit 15e Presentation unit

Claims

1. An acquisition unit that acquires biometric information when presenting content to be processed to a user; a calculation unit that calculates a distance matrix representing the correlation between the biometric information; a classification unit that classifies each content corresponding to the biometric information into groups using the calculated distance matrix; a presentation unit that presents the positional relationship between each content and each group corresponding to the biometric information using the calculated distance matrix; and a generation unit that generates content having the characteristics of the content or group specified by the user among the presented content and the group. A content evaluation apparatus characterized by comprising:

2. The content evaluation apparatus according to claim 1, wherein the presentation unit generates and further presents a label of the group using a caption for explaining each content.

3. The content evaluation apparatus according to claim 1, wherein the presentation unit further adds and presents the generated content to the presented positional relationship using the feature vectors of the presented content and the generated content.

4. A content evaluation method executed by a content evaluation apparatus, the method including: an acquisition step of acquiring biometric information when presenting content to be processed to a user; a calculation step of calculating a distance matrix representing the correlation between the biometric information; a classification step of classifying each content corresponding to the biometric information into groups using the calculated distance matrix; a presentation step of presenting the positional relationship between each content and each group corresponding to the biometric information using the calculated distance matrix; and a generation step of generating content having the characteristics of the content or group specified by the user among the presented content and the group. A content evaluation method characterized by including:

Images