US-12619670-B2 - Content recommendation system, content recommendation method, content library, method for generating content library, and target-input user interface

US12619670B2US 12619670 B2US12619670 B2US 12619670B2US-12619670-B2

Abstract

Provided is A content recommendation system that includes a vital-feature-amount generator that acquires chronological vital data that is vital data of a user that is continuously and chronologically sensed by a vital sensor, and generates chronological vital-feature-amount data from the chronological vital data. The content recommendation system further includes an emotion estimation calculator that generates, from the chronological vital-feature-amount data, an estimated emotion value that is an estimated value of an emotion of the user, a recommendation engine that acquires a target emotion value that is input through a user interface terminal apparatus and indicates an emotion that is a target of the user, and selects, from a content library, content used to reach the target emotion value from the estimated emotion value, and a content recommendation section that recommends the selected content to the user interface terminal apparatus.

Inventors

Mao Katsuhara

Assignees

Sony Group Corporation

Dates

Publication Date: 20260505
Application Date: 20220126
Priority Date: 20210310

Claims (20)

1 . A content recommendation system, comprising: a processor configured to: acquire first chronological vital data that is vital data of a user, wherein the first chronological vital data of the user is continuously and chronologically sensed by a vital sensor; generate first chronological vital-feature-amount data based on the first chronological vital data; generate, based on the first chronological vital-feature-amount data, a first estimated emotion value, wherein the first estimated emotion value is a first estimated value of a first emotion of the user; acquire a target emotion value based on a user input through a user interface terminal apparatus, wherein the target emotion value indicates an emotion that is a target of the user; select, from a content library, a plurality of pieces of content, wherein the plurality of pieces of content is selected to reach the target emotion value from the first estimated emotion value; generate context information regarding the user; narrow down the selected plurality of pieces of content to at least one piece of content based on the context information; and recommend the at least one piece of content to the user interface terminal apparatus.
2 . The content recommendation system according to claim 1 , wherein the processor is further configured to acquire the first chronological vital data and generate the first estimated emotion value before the acquisition of the target emotion value.
3 . The content recommendation system according to claim 1 , wherein the processor is further configured to: obtain newest vital-feature-amount data based on a change in previous vital-feature-amount data of the first chronological vital-feature-amount data; and generate a second estimated emotion value based on the newest vital-feature-amount data.
4 . The content recommendation system according to claim 1 , wherein the processor is further configured to store the at least one piece of content in the content library, the at least one piece of content is stored such that a corresponding estimated emotion value of a plurality of estimated emotion values reaches a respective target emotion value of a plurality of target emotion values, the corresponding estimated emotion value reaches the respective target emotion value based on a respective piece of content of the at least one piece of content, the plurality of target emotion values includes the target emotion value, and the plurality of estimated emotion values includes the first estimated emotion value.
5 . The content recommendation system according to claim 4 , wherein the content library is a two-dimensional matrix that includes the plurality of estimated emotion values and the plurality of target emotion values, the respective piece of content is registered in a portion of the content library, and the portion of the content library corresponds to a point of intersection of a first line corresponding to the target emotion value and a second line corresponding to the first estimated emotion value.
6 . The content recommendation system according to claim 1 , wherein the processor is further configured to: store the plurality of pieces of content in the content library, wherein the plurality of pieces of content is stored such that a corresponding estimated emotion value of a plurality of estimated emotion values reaches a respective target emotion value of a plurality of target emotion values, the corresponding estimated emotion value reaches the respective target emotion value based on a respective piece of content of the plurality of pieces of content, and at least one piece of the context information of a plurality of pieces of the context information is associated with a respective piece of content of the plurality of pieces of content; and narrow down the selected plurality of pieces of content to the at least one piece of content based on the association of the at least one piece of the context information to the respective piece of content.
7 . The content recommendation system according to claim 1 , wherein the processor is further configured to: obtain second chronological vital-feature-amount data based on the recommended at least one piece of content; generate a second estimated emotion value based on the second chronological vital-feature-amount data, wherein the second estimated emotion value is a second estimated value of a second emotion of the user; and update the content library for the user based on registration of the recommended at least one piece of content in the content library, wherein the at least one piece of content is registered such that the first estimated emotion value reaches the second estimated emotion value, and the first estimated emotion value reaches the second estimated emotion value based on the at least one piece of content.
8 . The content recommendation system according to claim 1 , wherein the processor is further configured to: obtain second chronological vital-feature-amount data based on the recommended at least one piece of content; generate a second estimated emotion value based on the second chronological vital-feature-amount data, wherein the second estimated emotion value is a second estimated value of a second emotion of the user; select, from the content library, content to reach the target emotion value from the second estimated emotion value, wherein the plurality of pieces of content one of includes or excludes the content; and recommend the selected content to the user interface terminal apparatus.
9 . The content recommendation system according to claim 1 , wherein the processor is further configured to: generate a plurality of estimated emotion values in chronological order, wherein the plurality of estimated emotion values includes the first estimated emotion value; and select, from the content library, content to reach the target emotion value from a newest estimated emotion value among the plurality of estimated emotion values, wherein the newest estimated emotion value is reached from a previous estimated emotion value of the plurality of estimated emotion values, and the plurality of pieces of content includes the content.
10 . The content recommendation system according to claim 9 , wherein the processor is further configured to store the at least one piece of content in the content library, the at least one piece of content is stored such that the newest estimated emotion value reaches the target emotion value, and the newest estimated emotion value reaches the target emotion value based on the at least one piece of content.
11 . The content recommendation system according to claim 9 , wherein the processor is further configured to store the at least one piece of content of the plurality of pieces of content in the content library, the at least one piece of content is stored such that: each of a plurality of previous estimated emotion values reaches a corresponding newest estimated emotion value of a plurality of newest estimated emotion values, wherein the each of the plurality of previous estimated emotion values reaches the corresponding newest estimated emotion value based on the at least one piece of content, and the plurality of estimated emotion values includes the plurality of previous estimated emotion values and the plurality of newest estimated emotion values; and the corresponding newest estimated emotion value reaches a respective target emotion value of a plurality of target emotion values, wherein the plurality of target emotion values includes the target emotion value.
12 . The content recommendation system according to claim 11 , wherein the content library is a three-dimensional matrix that includes the plurality of previous estimated emotion values, the plurality of newest estimated emotion values, and the plurality of target emotion values, the at least one piece of content is registered in a portion of the content library, and the portion of the content library is corresponding to a point of intersection of a first line corresponding to a previous estimated emotion value of the plurality of previous estimated emotion values a second line corresponding to the corresponding newest estimated emotion value, and a third line corresponding to the respective target emotion value.
13 . The content recommendation system according to claim 1 , wherein the processor is further configured to: calculate a probability of the user having the first emotion in a specific emotional state; and set the probability as the first estimated emotion value.
14 . The content recommendation system according to claim 13 , wherein the probability corresponds to a value that quantifies a state of the user having the emotion in the specific emotional state.
15 . The content recommendation system according to claim 13 , wherein the probability includes a first probability and a second probability, and the processor is further configured to: calculate the first probability of the user having a second emotion in a first specific emotional state; calculate the second probability of the user having a third emotion in a second specific emotional state; and generate the first estimated emotion value based on the first probability and the second probability.
16 . The content recommendation system according to claim 15 , wherein the first specific emotional state is an arousal state, the first probability is a probability of the user having the second emotion in the arousal state, the second specific emotional state is a pleasure state of valance, and the second probability is a probability of the user having the third emotion in the pleasure state.
17 . The content recommendation system according to claim 13 , wherein the user interface terminal apparatus displays a first target-input user interface, the first target-input user interface is a Graphical User Interface (GUI) displayed for the user to input the target emotion value, the first target-input user interface includes, in a single-axis direction, a first plurality of different areas, each of the first plurality of different areas corresponds to a respective probability of each of a first probability in the specific emotional state and a second probability in the specific emotional state, one of the first plurality of different areas is selectable by the user to input the respective probability of a corresponding area of the first plurality of different areas, and the respective probability is input to the user interface terminal apparatus as the target emotion value.
18 . The content recommendation system according to claim 17 , wherein the user interface terminal apparatus displays a second target-input user interface, the second target-input user interface displays a second plurality of different areas in a matrix in a biaxial direction, a first area of the second plurality of different areas corresponds to a combination of a third probability and a fourth probability, a second area of the second plurality of different areas corresponds to a combination of the third probability and a fifth probability, a third area of the second plurality of different areas corresponds to a combination of a sixth probability and the fifth probability, a fourth area of the second plurality of different areas corresponds to a combination of the sixth probability and the fourth probability, the third probability and the sixth probability are probabilities of the user having a second emotion in a first specific emotional state, the fourth probability and the fifth probability are probabilities of the user having a third emotion in a second specific emotional state, and one of the second plurality of different areas is selectable by the user to input, to the user interface terminal apparatus, as the target emotion value.
19 . The content recommendation system according to claim 17 , wherein the user interface terminal apparatus acquires the generated first estimated emotion value, and the first target-input user interface displays an object that represents the first estimated emotion value on a first area of the first plurality of different areas.
20 . The content recommendation system according to claim 19 , wherein the target emotion value is input to the user interface terminal apparatus based on the user input, the user input corresponds to swipe from the first area to a second area of the first plurality of different areas, and the second area includes the target emotion value.

Description

CROSS REFERENCE TO RELATED APPLICATIONS This application is a U.S. National Phase of International Patent Application No. PCT/JP2022/002893 filed on Jan. 26, 2022, which claims priority benefit of Japanese Patent Application No. JP 2021-038000 filed in the Japan Patent Office on Mar. 10, 2021. Each of the above-referenced applications is hereby incorporated herein by reference in its entirety. TECHNICAL FIELD The present disclosure relates to a content recommendation system that recommends content such as a piece of music to a user interface terminal apparatus, and a content recommendation method. The present disclosure relates to a content library used to select content to be recommended, and a method for generating the content library. The present disclosure relates to a target-input user interface that is a graphical user interface displayed on the user interface terminal apparatus. BACKGROUND ART Typically, when content such as a piece of music and a video is recommended, meta-information including the behavior of a user such as a shopping history or a viewing history of the user; as well as a genre, a rhythm, and a tempo of a piece of music is used to generate a recommendation list for each scene according to a general interpretation, and the generated recommendation list is presented. CITATION LIST Patent Literature Patent Literature 1: Japanese Patent Application Laid-open No. 2018-195043Patent Literature 2: Japanese Unexamined Patent Application Publication No. 2016-532360Patent Literature 3: Japanese Patent Application Laid-open No. 2018-159908 DISCLOSURE OF INVENTION Technical Problem However, the use of the above-described recommendation method based on generalities results in difficulty in knowing a true intention of a user. A recommendation against an intention of a user may result in a reduction in the level of user's confidence in a recommendation system. Consequently, there is a possibility that the service will not be used by the user. For this reason, recommendations of content depending on a psychological state of each user have been proposed (Patent Literature 1 and Patent Literature 2). However, those methods are methods for actually making a recommendation using estimation, and thus have little scientific basis. Further, content to be recommended is selected on the basis of conventional meta-information. This results in difficulty in recommending content fit for each user. In view of the circumstances described above, it is an object of the present disclosure to recommend content fit for each user. Solution to Problem A content recommendation system according to an embodiment of the present disclosure includes: a vital-feature-amount generator that acquires chronological vital data that is vital data of a user that is continuously and chronologically sensed by a vital sensor, andgenerates chronological vital-feature-amount data from the chronological vital data; an emotion estimation calculator that generates, from the chronological vital-feature-amount data, an estimated emotion value that is an estimated value of an emotion of the user;a recommendation engine that acquires a target emotion value that is input through a user interface terminal apparatus and indicates an emotion that is a target of the user, andselects, from a content library, content used to reach the target emotion value from the estimated emotion value; and a content recommendation section that recommends the selected content to the user interface terminal apparatus. The present embodiment makes it possible to reflect continuous and chronological vital data for a longer period of time. This results in being able to generate an estimated emotion value more precisely. Since the estimated emotion value is more precise, content more suitable for reaching a target emotion value from the estimated emotion value can be selected. Before the recommendation engine acquires the target emotion value, the vital-feature-amount generator may acquire the chronological vital data to generate the chronological vital-feature-amount data, andthe emotion estimation calculator may generate the estimated emotion value on the basis of the chronological vital-feature-amount data. Before the recommendation engine acquires a target emotion value, the vital-feature-amount generator acquires chronological vital data to generate chronological vital-feature-amount data, and the emotion estimation calculator generates an estimated emotion value on the basis of the chronological vital-feature-amount data. This makes it possible to reflect continuous and chronological vital data for a longer period of time, compared to when the recommendation engine acquires a target emotion value, and then, chronological vital data is acquired to generate chronological vital-feature-amount data and an estimated emotion value is generated. This results in being able to generate an estimated emotion value more precisely. Since the esti