Evaluation device, evaluation method, and computer program

Abstract

Improve the accuracy of video ad evaluation. [Solution] The evaluation device for evaluating video advertisements comprises an acquisition unit and an evaluation unit. The acquisition unit is configured to acquire feature quantities relating to one or more expressive elements contained in the video advertisement. The evaluation unit is configured to input the feature quantities into a trained machine learning model and acquire evaluation values ​​for the video advertisement output by the machine learning model using one or more evaluation indicators based on the feature quantities from the machine learning model. One or more expressive elements include at least one of visual information and auditory information. One or more evaluation indicators include one or more indicators relating to advertising effectiveness.

Description

Technical Field

[0004] ,

[0006] , , , , , ,

[0005] , , ,

[0003] , , , ,

[0001] This disclosure relates to an evaluation device.

Background Art

[0002] <时 Techniques for evaluating advertisements are known. For example, Patent Document 1 discloses an information processing program that evaluates a document used in an advertisement and improves the constituent elements of the advertisement.

Prior Art Documents

Patent Documents

[0003] ;

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] Video advertisements such as TV commercials and web advertisements contain various types of expression elements. Therefore, it is difficult to accurately evaluate video advertisements by only considering the documents used in the advertisements, as in the information processing program disclosed in Patent Document 1.

[0005] One aspect of this disclosure is to improve the accuracy of evaluating video advertisements.

Means for Solving the Problems

[0006] One aspect of this disclosure is an evaluation device for evaluating a video advertisement, comprising an acquisition unit and an evaluation unit. The acquisition unit is configured to acquire feature amounts related to one or more expression elements included in the video advertisement. The evaluation unit is configured to input the feature amounts into a learned machine learning model and acquire, from the machine learning model, an evaluation value for the video advertisement output by the machine learning model using one or more evaluation indicators based on the feature amounts. The one or more expression elements include at least one of visual information and auditory information. The one or more evaluation indicators include one or more indicators related to advertisement effects.

[0007] With this configuration, video ads can be evaluated using evaluation values ​​based on at least one element of the visual and auditory information contained in the video ad. Therefore, the accuracy of video ad evaluation can be improved.

[0008] In one aspect of this disclosure, one or more expressive elements may include auditory information. In one aspect of this disclosure, one or more expressive elements may include both visual and auditory information.

[0009] With this configuration, video ads can be evaluated using evaluation metrics based on the auditory information contained within them. Therefore, the accuracy of video ad evaluation can be improved.

[0010] In one aspect of this disclosure, the acquisition unit may be configured to acquire the appearance time of one or more expressive elements in a video advertisement as a feature. The evaluation unit may be configured to input the appearance time corresponding to one or more expressive elements to a machine learning model as a feature.

[0011] In one aspect of this disclosure, the appearance time may be the sum of the time that each of the one or more expressive elements appears in the video advertisement. This configuration allows for the evaluation of video ads not only based on their expressive elements but also on the duration of those elements. Therefore, the accuracy of video ad evaluation can be improved.

[0012] One aspect of this disclosure may further include a calculation unit. One or more expression elements may include multiple elements. The calculation unit may be configured to calculate a contribution, which is the degree to which each of the multiple elements contributes to each of the one or more evaluation indicators.

[0013] With this configuration, the contribution of each expressive element included in the video advertisement can be used for evaluation. In one aspect of this disclosure, the calculation unit may be configured to identify at least one element among a plurality of elements that has a relatively high contribution.

[0014] This structure allows us to identify which expressive elements in video ads contribute to their effectiveness. Therefore, we can improve video ads. One aspect of this disclosure may be a computer-based evaluation method for evaluating video advertisements. The evaluation method includes obtaining features relating to one or more expressive elements contained in the video advertisement, inputting the features into a trained machine learning model, and obtaining from the machine learning model an evaluation value for the video advertisement output by the machine learning model using one or more evaluation metrics based on the features. One or more expressive elements include at least one of visual information and auditory information. One or more evaluation metrics include one or more metrics relating to advertising effectiveness.

[0015] This evaluation method produces the same effect as the evaluation device described above. In one aspect of this disclosure, a computer program may be provided for causing a computer to perform at least part of the evaluation method described above. [Brief explanation of the drawing]

[0016] [Figure 1] This is a block diagram showing the configuration of the evaluation system. [Figure 2] This diagram illustrates the input and output of a machine learning model. [Figure 3] This diagram illustrates the relationship between past advertisements stored in the database, a list of expression elements, and evaluation data. [Figure 4] This diagram explains the expressive elements and their appearance times in the list of expressive elements. [Figure 5] This diagram illustrates the total value of the expressive elements and their appearance times in the list of expressive elements. [Figure 6] This is a flowchart representing the learning process. [Figure 7]It is a flowchart representing an evaluation process. [Figure 8] It is an example of a screen representing the result of an evaluation process. [Figure 9] It is a diagram explaining the output of a machine learning model in other embodiments.

Mode for Carrying Out the Invention

[0017] Hereinafter, exemplary embodiments of the present disclosure will be described with reference to the drawings. [1. First Embodiment] [1-1. Configuration] [1-1-1. Overall Configuration] The evaluation system 100 shown in FIG. 1 is a system for evaluating video advertisements broadcast by media such as television and the Internet.

[0018] The evaluation system 100 includes an evaluation device 1, a machine learning model 2, and a database 3. The evaluation device​​​​​​​​​​​​​The user interface 14 is a general term for interfaces that accept various input operations from the user and interfaces that output various information to the user. Examples of user interfaces 14 include keyboards, mice, touch panels, displays, etc.

[0022] The communication interface 15 is an interface capable of transmitting various types of data according to a predetermined standard. The evaluation device 1 is configured to communicate with the machine learning model 2 and the database 3 through the communication interface 15.

[0023] Machine learning model 2 is a pre-trained model using machine learning. Machine learning model 2 may also be a model trained using deep learning. As shown in Figure 2, when the machine learning model 2 receives feature quantities 41 relating to one or more expressive elements 40 included in a video advertisement, it is configured to output an evaluation value 51 calculated using one or more evaluation metrics 5 for the video advertisement based on the feature quantities 41, and a contribution score 52 corresponding to each of the one or more expressive elements 40.

[0024] One or more expressive elements 40 include at least one of the visual information and / or auditory information contained in the video advertisement. Visual information is information that appeals to the viewer of the video advertisement through sight. Examples of visual information include people such as talents and characters, and text such as catchphrases and captions. Auditory information is information that appeals to the viewer of the video advertisement through hearing. Examples of auditory information include voices such as dialogue and narration, music such as songs and background music, and sound effects.

[0025] The expressive element 40 may also be a performance or expression. For example, the expressive element 40 may be a "sizzle" that appeals to the senses through sight and sound. Examples of sizzle include ingredient sizzle, which displays the ingredients themselves along with sound effects in advertisements dealing with ingredients or dishes, and cooking sizzle, which includes the process and sounds of cooking, such as cutting or frying ingredients.

[0026] One or more evaluation metrics 5 are indicators related to the advertising effectiveness of video ads. An example of an evaluation metric 5 is an indicator related to the effectiveness of the advertising expression. Examples of indicators related to the effectiveness of advertising expression include the degree to which the appeal of the brand or product is showcased (i.e., how attractively the brand or product was portrayed), and the degree to which attention is drawn to the video ad (i.e., how eye-catching the advertising expression was).

[0027] The more appealing the video ad is, the more likely it is to make the product seem attractive to the viewer. The more attention-grabbing an ad is, the more prominent its visual presentation becomes, making it more likely to capture the viewer's attention.

[0028] The rating score 51 is a value obtained by evaluating a video advertisement using evaluation metric 5. The rating score 51 is calculated for each of one or more evaluation metric 5. The rating score 51 may be a numerical value that evaluates the video advertisement, or it may be a rank that evaluates it on a tiered basis based on the numerical value. As an example, Figure 2 shows that for a certain video advertisement, the rating scores 51 corresponding to the attractiveness and attention-grabbingness of evaluation metric 5 are "94.1" and "86.5", respectively.

[0029] As shown in Figure 3, database 3 contains past advertisement group ADG, expression element list IL, and evaluation data EV. The past ad group ADG includes multiple video ads that were broadcast in the past.

[0030] The list of expressive elements IL contains a list of features 41 related to one or more expressive elements 40 corresponding to each video ad in the past ad group ADG. The evaluation data EV includes one or more evaluation values ​​of 51 corresponding to each video ad in the past ad group ADG.

[0031] One example of evaluation data EV is Best HIT (registered trademark). Best HIT is a fixed-point observation survey that measures the advertising effectiveness of television commercials broadcast weekly on Tokyo's key television stations. Best HIT has data that evaluates each television commercial based on indicators related to the effectiveness of advertising expression, including the degree of appeal and the degree of attention-grabbing.

[0032] The contribution score of 52 represents the degree to which each of the one or more expressive elements 40 contributes to the evaluation metric 5. In a given video advertisement, an expressive element 40 with a high contribution score of 52 to a certain evaluation metric 5 contributes a high degree to that evaluation metric 5 in that video advertisement.

[0033] The contribution score of 52 is calculated for each evaluation index 5. For example, if one or more evaluation index 5s are used, such as the first evaluation index 5a and the second evaluation index 5b, then the first contribution score of 52a is calculated as the contribution score of 52 for the first evaluation index 5a, and the second contribution score of 52b is calculated as the contribution score of 52 for the second evaluation index 5b.

[0034] For example, if the first evaluation metric 5a is the degree of attractiveness, then the first contribution metric 52a represents the degree to which the corresponding expressive element 40 contributes to the degree of attractiveness. If the second evaluation metric 5b is the degree of attention-grabbing, then the second contribution metric 52b represents the degree to which the corresponding expressive element 40 contributes to attention-grabbing.

[0035] In the example in Figure 2, for "Expression Element B," the "Contribution to Attractiveness" as the first contribution 52a is "0.556." The "Contribution to Attention-Generatingness" as the second contribution 52b is "-0.119." In other words, even if the same expression element 40 has a high contribution 52 to one evaluation index 5, it may have a low contribution 52 to other evaluation index 5s.

[0036] [1-1-2. List of Expression Elements] As shown in Figure 4, the list of expressive elements IL includes the appearance time 41a of each of the one or more expressive elements 40 in a video advertisement as a feature quantity 41 for each of the one or more expressive elements 40 in the video advertisement.

[0037] Appearance time 41a represents the time period in which the expressive element 40 appears in the video advertisement. Specifically, appearance time 41a represents the start time and end time in which the expressive element 40 appears in the video advertisement. The start time and end time are relative to the beginning of the video advertisement. Appearance time 41a may also be represented by a start time and elapsed time instead of a start time and end time. Appearance time 41a may be consecutive or non-consecutive.

[0038] In this embodiment, as an example, for appearance time 41a, the start and end times of the appearance of the expressive element 40 in a 15-second video advertisement are expressed in units of 0.1 seconds. For example, the appearance time 41a of the expressive element 40, "talent," is the time period between 0.5 seconds and 5.2 seconds and between 8.4 seconds and 15.0 seconds from the start of the video advertisement.

[0039] In Figure 4, the shaded areas for each expression element 40 visually represent the appearance time 41a. The numbers 1 through 15 above the shaded areas represent the number of seconds (rounded up to the nearest whole number) that have elapsed since the start of the video advertisement.

[0040] Alternatively, as shown in Figure 5, the appearance time 41a may be the sum of the time that each of the one or more expressive elements 40 appears in the video advertisement. In the example above, the appearance time 41a of the expressive element 40 "Talent" is 11.3 seconds.

[0041] The length of a video ad is not limited to 15 seconds; for example, it may be 30 seconds or 60 seconds. The unit of appearance time 41a is not limited to 0.1 seconds; for example, it may be 0.5 seconds or 1 second.

[0042] The expressive elements 40 and appearance time 41a may be manually associated with each video advertisement by the user, or they may be automatically associated by a computer capable of analyzing the video content.

[0043] [1-2. Processing] [1-2-1. Learning Process] In the evaluation system 100, the evaluation device 1 is configured to perform the learning process shown in Figure 6.

[0044] The learning process is the process of generating machine learning model 2 using machine learning. The following describes the processes executed by the processor 11 of the evaluation device 1 during the learning process, using the flowchart in Figure 6.

[0045] When the processor 11 receives an instruction from the user to execute a learning process through the user interface 14, it starts the process shown in Figure 6. First, in S100, the processor 11 retrieves past ad group ADG, expression element list IL, and evaluation data EV from database 3 via the communication interface 15.

[0046] Next, in S110, processor 11 generates machine learning model 2 by performing supervised learning. The machine learning model 2 has feature quantities 41 relating to one or more representation elements 40 as explanatory variables. In this embodiment, the machine learning model 2 has an explanatory variable representing the appearance time 41a of the corresponding representation element 40 as a feature quantity 41 corresponding to one or more representation elements 40.

[0047] Machine learning model 2 has one or more evaluation metrics 5 as its dependent variable. In supervised learning, the explanatory variables are given as features 41 included in the list of expressive elements IL corresponding to each video ad in the past ad group ADG. The dependent variable is given as the evaluation data EV for the corresponding video ad. Machine learning model 2 is generated based on combinations of multiple expressive elements 40 included in the list of expressive elements IL, and combinations of expressive elements 40 and their appearance time 41a, etc.

[0048] After that, the processor 11 terminates the process shown in Figure 6. In this way, the processor 11 performs machine learning using various data obtained from the database 3 and generates a machine learning model 2.

[0049] [1-2-2. Evaluation Process] In the evaluation system 100, the evaluation device 1 is configured to perform the evaluation process shown in Figure 7.

[0050] The evaluation process uses machine learning model 2 to evaluate video advertisements. In other words, the evaluation process uses machine learning model 2 to predict a video advertisement rating of 51. In the following, the processes executed by the processor 11 of the evaluation device 1 during the evaluation process will be explained using the flowchart in Figure 7.

[0051] When the processor 11 receives an instruction from the user to execute an evaluation process through the user interface 14, it starts the process shown in Figure 7. First, in S200, the processor 11 obtains feature quantities 41 relating to one or more expressive elements 40 in the video advertisement to be evaluated. The processor 11 may obtain the feature quantities 41 from the storage 13, or from outside the evaluation device 1 (e.g., from the database 3) via the communication interface 15. In this embodiment, the feature quantities 41 include the appearance time 41a corresponding to each of the one or more expressive elements 40.

[0052] The feature quantities 41 in the video advertisement being evaluated may be manually extracted by the user beforehand, or they may be automatically extracted beforehand by a computer capable of analyzing the video content.

[0053] Next, in S210, the processor 11 inputs the feature quantities 41 related to one or more expression elements 40 obtained in S200 into the machine learning model 2, and predicts the evaluation value 51 of the video advertisement to be evaluated by obtaining an evaluation value 51 as the output of the machine learning model 2 based on the feature quantities 41. The processor 11 displays the obtained evaluation value 51 to the user through the user interface 14.

[0054] Next, in S220, the processor 11 uses the output of the machine learning model 2 in S210 to calculate the contribution score 52, which is the degree to which each representation element 40 contributes to the evaluation index 5. If the processor 11 inputs features 41 related to multiple representation elements 40 into the machine learning model 2 in S210, it calculates the contribution score 52 for each representation element 40. The contribution score 52 is calculated, for example, using SHAP (SHapley Additive exPlanations).

[0055] Next, in S230, the processor 11 identifies at least one representation element 40 among the representation elements 40 relating to the feature quantities 41 input to the machine learning model 2 that has a relatively high contribution 52 calculated in S220.

[0056] For example, the processor 11 identifies the single representation element 40 with the highest contribution 52 among the representation elements 40 relating to the feature quantity 41 input to the machine learning model 2. Alternatively, the processor 11 identifies a predetermined number of representation elements 40 in descending order of contribution 52.

[0057] In another example, the processor 11 identifies at least one representation element 40 of the representation elements 40 relating to the feature 41 input to the machine learning model 2 that has a contribution 52 exceeding a predetermined threshold. For example, the processor 11 identifies a representation element 40 of the representation elements 40 relating to the feature 41 input to the machine learning model 2 whose contribution 52 is a positive value.

[0058] The processor 11 displays the identified representation element 40 and the corresponding contribution 52 to the representation element 40 to the user through the user interface 14. Figure 8 shows an example of a screen displaying the video advertisement being evaluated, with the evaluation value 51 for attractiveness as an evaluation metric 5, and the top three expressive elements 40 with the highest contribution 52, as a result of the above processing, on the user interface 14 display. Figure 8 also shows the bottom two expressive elements 40 with the lowest contribution 52.

[0059] After that, processor 11 terminates the process shown in Figure 7. In this way, the processor 11 is configured to input feature quantities 41 related to the expressive elements 40 of the video advertisement to be evaluated into the machine learning model 2, obtain an evaluation value 51 from the machine learning model 2 as output, and evaluate the video advertisement to be evaluated. Furthermore, the processor 11 is configured to calculate the contribution 52 to the evaluation index 5 corresponding to the obtained evaluation value 51 for each expressive element 40, and to identify the expressive elements 40 with a relatively high contribution 52.

[0060] [1-3. Effects] According to the embodiments described above, the following actions and effects can be obtained. (1a) In the learning process, the processor 11 generates a machine learning model 2 by performing supervised learning. The machine learning model 2 has features 41 relating to one or more expression elements 40 as explanatory variables. The machine learning model 2 has one or more evaluation metrics 5 as the target variable. In supervised learning, the explanatory variables are given a list of expression elements IL corresponding to each video ad included in the past ad group ADG. The target variable is given the evaluation data EV of the corresponding video ad.

[0061] In the evaluation process, the processor 11 inputs feature quantities 41 related to one or more expressive elements 40 in the video advertisement to be evaluated into the machine learning model 2, and predicts the evaluation value 51 of the video advertisement to be evaluated by obtaining an evaluation value 51 as the output of the machine learning model 2 based on the feature quantities 41.

[0062] The expressive element 40 includes at least one of the visual information and auditory information contained in the video advertisement. This process allows for the evaluation of video advertisements using an evaluation score of 51, based on at least one of the visual and auditory information contained within the video advertisement. Therefore, the accuracy of video advertisement evaluation can be improved.

[0063] (1b) The machine learning model 2 may have an explanatory variable that represents the time of appearance 41a of the corresponding representation element 40 as a feature 41 corresponding to one or more representation elements 40. This processing method allows for the evaluation of video advertisements based on the appearance time 41a of the expressive element 40. Therefore, the accuracy of video advertisement evaluation can be improved.

[0064] (1c) In the evaluation process, the processor 11 uses the output of the machine learning model 2 to calculate the contribution 52, which is the degree to which the representation element 40 contributes to the evaluation index 5. Through this process, the contribution 52 of each of the 40 expressive elements included in the video advertisement can be used for evaluation.

[0065] (1d) The processor 11 identifies the representation elements 40 with relatively high calculated contributions 52 from among the representation elements 40 relating to the feature quantities 41 input to the machine learning model 2. This process makes it possible to identify which expressive elements 40 contribute to the advertising effect in a video ad. Therefore, users can improve a video ad under production, for example, by using the contribution score 52 to select or discard expressive elements 40 and adjust their appearance time 41a. Specifically, users can improve a video ad by lengthening the appearance time 41a of expressive elements 40 with a high contribution score 52, or by replacing expressive elements 40 with a low contribution score 52 with other expressive elements 40.

[0066] [1-4. Correspondence between terms] In the above embodiment, the process in S200 corresponds to an example of a process executed by the acquisition unit, the process in S210 corresponds to an example of a process executed by the evaluation unit, and the processes in S220 and S230 correspond to an example of a process executed by the calculation unit.

[0067] [2. Other Embodiments] While embodiments of this disclosure have been described above, it goes without saying that this disclosure is not limited to the embodiments described above and can take various forms.

[0068] (2a) In the above embodiment, the machine learning model 2 has an explanatory variable that represents the time of appearance 41a of the corresponding representation element 40 as a feature 41 corresponding to one or more representation elements 40. However, the feature 41 relating to the representation element 40 is not limited to the time of appearance 41a.

[0069] For example, machine learning model 2 may have explanatory variables that indicate the presence or absence of a corresponding representation element 40, as features 41 corresponding to each of the one or more representation elements 40. Alternatively, machine learning model 2 may have a categorical value relating to the distribution of one or more expressive elements 40 in a video advertisement as a feature 41 relating to one or more expressive elements 40. The categorical value may be a value that can identify at least one of the presence or absence, appearance time, and appearance time of each of the one or more expressive elements 40 in the video advertisement. The categorical value may also be a value that represents the classification (i.e., a value that represents the class to which it belongs) when the video advertisement is classified using at least one of the presence or absence, appearance time, and appearance time of each of the one or more expressive elements 40 as an indicator.

[0070] (2b) In the above embodiment, the machine learning model 2 is input with features 41 relating to one or more expressive elements 40 included in the video advertisement. However, the features input to the machine learning model 2 are not limited to these. For example, in addition to the features 41 relating to the expressive elements 40, the machine learning model 2 may also be input with features relating to the product category in the video advertisement, features relating to the appeal content of the video advertisement, and so on.

[0071] (2c) In the above embodiment, the processor 11 identifies the representation elements 40 with a relatively high contribution 52 among the representation elements 40 relating to the feature quantity 41 input to the machine learning model 2. However, the processor 11 may identify not only the representation elements 40 with a relatively high contribution 52, but also the representation elements 40 with a relatively low contribution 52. The processor 11 may identify at least one of the representation elements 40 with a relatively high contribution 52 and the representation elements 40 with a relatively low contribution 52.

[0072] In the example shown in Figure 8, in addition to the top three representation elements 40 with a high contribution score of 52, the bottom two representation elements 40 with a low contribution score of 52 are also displayed. In other words, the processor 11 identifies both the representation elements 40 with a relatively high contribution score of 52 and the representation elements 40 with a relatively low contribution score of 52, and displays them on the user interface 14's display.

[0073] (2d) In the above embodiment, the evaluation system 100 includes one machine learning model 2. The machine learning model 2 is configured to output, upon input of features 41 relating to one or more expressive elements 40 included in a video advertisement, an evaluation value 51 calculated using one or more evaluation indicators 5 for the video advertisement based on the features 41, and a contribution 52 corresponding to each of the one or more expressive elements 40.

[0074] However, one machine learning model 2 may be configured to output an evaluation value 51 calculated using one evaluation metric 5 for the video advertisement. One machine learning model 2 may also be configured to output a contribution 52 corresponding to one or more expressive elements 40 for that one evaluation metric 5.

[0075] For example, as shown in Figure 9, the evaluation system 100 may include multiple machine learning models (e.g., a first machine learning model 2a and a second machine learning model 2b) corresponding to multiple evaluation indicators 5. That is, the evaluation system 100 may include a single machine learning model 2, with each machine learning model corresponding to an evaluation indicator 5.

[0076] The first machine learning model 2a, as one machine learning model 2, may be configured to output an evaluation value 51 (94.1 in Figure 9) calculated using the first evaluation metric 5a, as one evaluation metric 5. The first machine learning model 2a may also be configured to output a first contribution 52a, as a contribution 52 corresponding to each of the one or more representation elements 40, namely "representation element A" - "representation element D", for the first evaluation metric 5a.

[0077] Similarly, a second machine learning model 2b, as one machine learning model 2, may be configured to output an evaluation value 51 (73.5 in Figure 9) calculated using a second evaluation metric 5b, as one evaluation metric 5. The second machine learning model 2b may also be configured to output a second contribution 52b, as a contribution 52 corresponding to each of the "expression element E" - "expression element H" as one or more expression elements 40, for the second evaluation metric 5b.

[0078] The second machine learning model 2b may be input with some or all of one or more representation elements 40 (represented as "representation element A" - "representation element D" in Figure 9) that were input to the first machine learning model 2a. The second machine learning model 2b may calculate an evaluation value 51 using a second evaluation index 5b based on some or all of the one or more representation elements 40 that were input to the first machine learning model 2a. The second machine learning model 2b may output a second contribution 52b corresponding to each of the one or more representation elements 40 that were input to the first machine learning model 2a.

[0079] Machine learning model 2 may be configured to include at least one other machine learning model 2. For example, as shown in Figure 9, machine learning model 2 may be configured to include a first machine learning model 2a and a second machine learning model 2b.

[0080] (2e) Multiple functions of one component in the above embodiment may be realized by multiple components, or one function of one component may be realized by multiple components. Multiple functions of multiple components may be realized by one component, or one function realized by multiple components may be realized by one component. Some of the configurations of the above embodiment may be omitted. At least some of the configurations of the above embodiment may be added to or replaced with the configurations of other above embodiments.

[0081] (2f) The present disclosure can be implemented in various forms other than the evaluation apparatus described above. For example, it can be implemented in the form of a system in which the evaluation apparatus is a component, a computer program for causing the computer to function as the evaluation apparatus, a non-transitional physical recording medium such as a semiconductor memory on which the computer program is recorded, an evaluation method, and so on.

[0082] [Technical concepts disclosed in this specification] [Item 1] An evaluation device for evaluating video advertisements, An acquisition unit configured to acquire feature quantities relating to one or more expressive elements included in the video advertisement, An evaluation unit is configured to input the features into a trained machine learning model and obtain from the machine learning model an evaluation value for the video advertisement output by the machine learning model using one or more evaluation metrics based on the features. Equipped with, The one or more expressive elements include at least one of visual information and auditory information. The aforementioned one or more evaluation metrics include one or more metrics relating to advertising effectiveness. Evaluation device.

[0083] [Item 2] The evaluation device described in item 1, The one or more expressive elements include auditory information. Evaluation device.

[0084] [Item 3] The evaluation device described in item 1, The aforementioned one or more expressive elements include both visual and auditory information. Evaluation device.

[0085] [Item 4] An evaluation device described in any one of items 1 to 3, The acquisition unit is configured to acquire the appearance time of each of the one or more expressive elements in the video advertisement as the feature quantity. The evaluation unit is configured to input the occurrence time corresponding to each of the one or more representation elements as a feature to the machine learning model. Evaluation device.

[0086] [Item 5] The evaluation device described in item 4, The aforementioned appearance time is the sum of the time that each of the one or more expressive elements appears in the video advertisement. Evaluation device.

[0087] [Item 6] An evaluation device described in any one of items 1 to 5, It also includes a calculation unit, The aforementioned one or more expressive elements include multiple elements, The calculation unit is configured to calculate the degree to which each of the plurality of elements contributes to each of the one or more evaluation indicators. Evaluation device.

[0088] [Item 7] The evaluation device described in item 6, The calculation unit is configured to identify at least one element among the plurality of elements that has a relatively high contribution. Evaluation device.

[0089] [Item 8] A computer-based evaluation method for evaluating video advertisements, To obtain feature quantities related to one or more expressive elements included in the aforementioned video advertisement, The process involves inputting the aforementioned features into a pre-trained machine learning model, and obtaining from the machine learning model an evaluation value for the video advertisement output by the machine learning model using one or more evaluation metrics based on the aforementioned features. Includes, The one or more expressive elements include at least one of visual information and auditory information. The aforementioned one or more evaluation metrics include one or more metrics relating to advertising effectiveness. Evaluation method.

[0090] [Item 9] A computer program that causes a computer to execute the evaluation method described in item 8. [Explanation of Symbols]

[0091] 1...Evaluation device, 11...Processor, 2...Machine learning model, 40...Representation element, 41...Features, 41a...Appearance time, 5,5a,5b...Evaluation index, 51...Evaluation value, ADG...Past advertisement group, IL...List of representation elements, EV...Evaluation data.

Claims

1. An evaluation device for evaluating video advertisements, An acquisition unit configured to acquire feature quantities relating to one or more expressive elements included in the video advertisement, An evaluation unit is configured to input the features into a trained machine learning model and obtain from the machine learning model an evaluation value for the video advertisement output by the machine learning model using one or more evaluation metrics based on the features. Equipped with, The one or more expressive elements include at least one of visual information and auditory information. The aforementioned one or more evaluation metrics include one or more metrics relating to advertising effectiveness. Evaluation device.

2. An evaluation apparatus according to claim 1, The one or more expressive elements include auditory information. Evaluation device.

3. An evaluation apparatus according to claim 1, The aforementioned one or more expressive elements include both visual and auditory information. Evaluation device.

4. An evaluation apparatus according to claim 1, The acquisition unit is configured to acquire the appearance time of each of the one or more expressive elements in the video advertisement as the feature quantity. The evaluation unit is configured to input the occurrence time corresponding to each of the one or more representation elements as a feature to the machine learning model. Evaluation device.

5. An evaluation apparatus according to claim 4, The aforementioned appearance time is the sum of the time that each of the one or more expressive elements appears in the video advertisement. Evaluation device.

6. An evaluation apparatus according to claim 1, It also includes a calculation unit, The aforementioned one or more expressive elements include multiple elements, The calculation unit is configured to calculate the degree to which each of the plurality of elements contributes to each of the one or more evaluation indicators. Evaluation device.

7. An evaluation apparatus according to claim 6, The calculation unit is configured to identify at least one element among the plurality of elements that has a relatively high contribution. Evaluation device.

8. A computer-based evaluation method for evaluating video advertisements, To obtain feature quantities relating to one or more expressive elements included in the aforementioned video advertisement, The process involves inputting the aforementioned features into a pre-trained machine learning model, and obtaining from the machine learning model an evaluation value for the video advertisement output by the machine learning model using one or more evaluation metrics based on the aforementioned features. Includes, The one or more expressive elements include at least one of visual information and auditory information. The aforementioned one or more evaluation metrics include one or more metrics relating to advertising effectiveness. Evaluation method.

9. A computer program for causing a computer to execute the evaluation method described in claim 8.

Images