Techniques are disclosed for curating content items. In some embodiments, a content item group generator applies rules defining inclusion and exclusion criteria to content items and associated metadata in order to assign the content items to content item groups. Given the assignments of content items to content item groups, a list generator applies a machine learning technique to generate, for each content item, a representation of the content item that includes weights associating the content item with the content item groups. The list generator then computes a weighted sum of representations of content items that a user has purchased and/or otherwise engaged with in order to generate a representation of the user that includes weights indicating affinities of the user with the content item groups. The list generator further generates one or more lists for display to the user based on the representation of the user.
BACKGROUND Technical Field (1) Embodiments of the present disclosure relate generally to computer processing of content items and, more specifically, to techniques for curating content items. Description of the Related Art (2) As digital and mobile communications have become ubiquitous, there are increasingly many options for end users to enjoy media content in addition to the traditional options of watching broadcast television (TV) and watching movies in-theater. For example, end users can now stream free and subscription content to televisions or mobile devices, rent DVDs, purchase pay-per-view rights to specific digital content, and so on. (3) Currently, there are few techniques for organizing media content for consumption by end users. One conventional approach for organizing media content is to associate media content items with metadata that indicates generic categories to which the media content items belong. For example, a collection of movie titles could be associated with metadata indicating genres, such as “comedy,” “romance,” “science fiction,” etc. to which those movies titles belong. Lists of movie titles from the different genres can then be presented for consumption by an end user. (4) One drawback of the above approach is that multiple lists can include many of same media content items, because the metadata oftentimes indicates that each media content item belongs to more than one generic category. For example, the metadata could indicate multiple genres to
1. A computer-implemented method for curating content items, the method comprising: generating, for each content item included in a plurality of content items, a representation of the content item, wherein the representation of the content item comprises a first plurality of weights associating the content item with a plurality of content item groups; generating a representation of a user that comprises a second plurality of weights associating the user with the plurality of content item groups, wherein each weight included in the second plurality of weights corresponds to a different content item group in the plurality of content item groups and is computed based on combining (i) one or more of the representations of content items that belong to the corresponding content item group and that are associated with one or more first content items that the user has at least one of purchased or engaged with, and (ii) one or more time decays that are computed based on when the user at least one of purchased or engaged with the one or more first content items; generating one or more lists of content items based on the representation of the user; and causing the one or more lists of content items to be displayed to the user. ||
11. One or more non-transitory computer-readable storage media including instructions that, when executed by at least one processor, cause the at least one processor to perform steps for curating content items, the steps comprising: generating, for each content item included in a plurality of content items, a representation of the content item, wherein the representation of the content item comprises a first plurality of weights associating the content item with a plurality of content item groups; generating a representation of a user that comprises a second plurality of weights associating the user with the plurality of content item groups, wherein each weight included in the second plurality of weights corresponds to a different content item group in the plurality of content item groups and is computed based on combining (i) one or more of the representations of content items that belong to the corresponding content item group and that are associated with one or more first content items that the user has at least one of purchased or engaged with, and (ii) one or more time decays that are computed based on when the user at least one of purchased or engaged with the one or more first content items; generating one or more lists of content items based on the representation of the user; and causing the one or more lists of content items to be displayed to the user. ||
19. A system, comprising: one or more memories storing instructions; and one or more processors that are coupled to the one or more memories and, when executing the instructions, are configured to: generate, for each content item included in a plurality of content items, a representation of the content item, wherein the representation of the content item comprises a first plurality of weights associating the content item with a plurality of content item groups, generate a representation of a user that comprises a second plurality of weights associating the user with the second plurality of content item groups, wherein each weight included in the second plurality of weights corresponds to a different content item group in the plurality of content item groups and is computed based on combining (i) one or more of the representations of content items that belong to the corresponding content item group and that are associated with one or more first content items that the user has at least one of purchased or engaged with, and (ii) one or more time decays that are computed based on when the user at least one of purchased or engaged with the one or more first content items, generate one or more lists of content items based on the representation of the user, and cause the one or more lists of content items to be displayed to the user.