Back to results
US 12,720,135 B2Granted
Content recommendation method, content distribution system, recommended content estimation apparatus, and machine learning apparatus
AssigneeSeiko Epson Corporation
InventorNona Mimura
Life of this patent
Estimating the term…
Abstract
A processing device totalizes viewing logs for each of projectors and stores a totalization result in a storage device. The processing device performs clustering of a plurality of projectors based on environmental indicators, that is, illuminance, a screen size, and noise included in the totalization result. The processing device calculates, for each of a plurality of contents reproduced in the projectors belonging to groups formed by the clustering, evaluation values in the entire groups based on an evaluation value of each of the plurality of contents. The processing device determines, based on the evaluation value calculated for each of the groups, recommended contents for the projectors.
Claims4
A content recommendation method comprising:
- acquiring, for each of a plurality of viewing terminals, (i) a first environmental indicator indicating an illuminance in an environment in which a viewer views content, (ii) a second environmental indicator indicating a screen size of a screen onto which the content is projected, the second environmental indicator being determined based on a distance measured by a sensor between respective viewing terminal and the screen, and (iii) a third environmental indicator indicating a noise level in the environment in which the viewer views the content;
- acquiring, for each of the plurality of viewing terminals, an evaluation value indicating a degree of viewing quality of the content, the evaluation value being given by the viewer;
- classifying the plurality of viewing terminals into a plurality of groups according to a similarity degree of the first, second, and third environmental indicators by:
- generating a virtual three-dimensional vector space,
- placing the first, second and third environmental indicators as feature vectors that are orthogonal to each other in the virtual three-dimensional vector space, and
- classifying certain viewing terminals into a same group based on a similarity of tip positions of their respective feature vectors in the virtual three-dimensional vector space;
- and
- calculating an evaluation value in entire one group of the classified groups from a plurality of the evaluation values associated with the viewing terminals belonging to the classified one group,
- determining, based on the calculated evaluation value, recommended contents corresponding to the environment to viewers who use the viewing terminals belonging to the one group, and
- causing the respective viewing terminal to display the recommended contents corresponding to the environment on the respective screen.
7 more elements
A content distribution system that distributes one or more contents to each of a plurality of viewing terminals, the content distribution system comprising:
- a content distribution server;
- and
- a data analysis server, wherein
- the content distribution server distributes, out of a stored plurality of contents, viewing content corresponding to a request of a viewer, who uses each of the plurality of viewing terminals, to the viewing terminal associated with the viewer, and
- the data analysis server includes:
- an acquirer configured to acquire, for each of the plurality of viewing terminals, (i) a first environmental indicator indicating an illuminance in an environment in which the viewer views content, (ii) a second environmental indicator indicating a screen size of a screen onto which the content is projected, the second environmental indicator being determined based on a distance measured by a sensor between respective viewing terminal and the screen, (iii) a third environmental indicator indicating a noise level in the environment in which the viewer views the content, and (iv) an evaluation value indicating a degree of viewing quality of the content, the evaluation value being given by the viewer;
- a classifier configured to classify the plurality of viewing terminals into a plurality of groups according to a similarity degree of the first, second, and third environmental indicators by:
- generating a virtual three-dimensional vector space,
- placing the first, second and third environmental indicators as feature vectors that are orthogonal to each other in the virtual three-dimensional vector space, and
- classifying certain viewing terminals into a same group based on a similarity of tip positions of their respective feature vectors in the virtual three-dimensional vector space;
- and
- a recommender configured to calculate an evaluation value in entire one group of the classified groups from a plurality of the evaluation values associated with the viewing terminals belonging to the classified one group,
- determine, based on the calculated evaluation value, recommended contents corresponding to the environment to viewers who use the viewing terminals belonging to the one group, and
- cause the respective viewing terminal to display the recommended contents corresponding to the environment on the respective screen.
11 more elements
A recommended content estimation apparatus comprising:
- a processor programmed to:
- acquire determination data including at least (i) a first environmental indicator indicating an illuminance in an environment in which a viewing terminal is installed, (ii) a second environmental indicator indicating a screen size of a screen onto which the content is projected, the second environmental indicator being determined based on a distance measured by a sensor between the viewing terminal and the screen, (iii) a third environmental indicator indicating a noise level in the environment in which a first viewer views the content, and (iv) candidate content that is a candidate of recommended content to the first viewer who uses the viewing terminal;
- estimate an evaluation value of the first viewer for the candidate content by inputting the determination data to a learning model that learned a relation between information concerning an environment at a time when a viewer views content and information concerning an attribute of the content and an evaluation value by the viewer for the content, wherein the relation is separately learned for a plurality of groups of viewing terminals by:
- generating a virtual three-dimensional vector space,
- placing the first, second and third environmental indicators as feature vectors that are orthogonal to each other in the virtual three-dimensional vector space, and
- classifying certain viewing terminals into a same group based on a similarity of tip positions of their respective feature vectors in the virtual three-dimensional vector space;
- determine, based on the estimated evaluation value of the first viewer for the candidate content, the recommended content to the first viewer, and
- cause the viewing terminal to display the recommended content corresponding to the environment on the screen.
5 more elements
Dependencies are inferred from the claim text.